01 Aerospace & Electronic Systems

Published on June 2017 | Categories: Documents | Downloads: 135 | Comments: 0 | Views: 3913

of 380

Content

AEROSPACE EXPERT SYSTEMS

303

AEROSPACE EXPERT SYSTEMS Since the Wright brothers ﬁrst took ﬂight, the progress of mankind in the domain of ﬂight has been nothing short of spectacular. The progress of technology, faster aircraft, instrument ﬂight, and increased air trafﬁc resulted in establishing a governmental infrastructure to control air trafﬁc. During the late 1920s air trafﬁc control was established as a profession. Rules, procedures, standards, and facilities were designed to provide for safe and orderly trafﬁc patterns and required separation between the aircraft (1). Stakeholders of the airspace system include airline operators, airport authorities, weather specialists, and air trafﬁc controllers (ground, tower, en-route, ﬂow), as well as the passengers. In addition to civilian users, including the airlines and general aviation, the system must accommodate military and space ﬂight activities. The demand for air transport has steadily increased since introduction of jet aircraft during the late 1950s. The safety and effectiveness of the national airspace system (NAS) depends on the performance of the air trafﬁc management (ATM) personnel—the employees of the Federal Aviation Administration (FAA). The increasing complexity of the system and proliferation of computing equipment have generated an urgent need to explore the possibility of supporting the human component of the system with tools and techniques based on the concepts and methods of artiﬁcial intelligence (AI) (2). Intelligence is deﬁned as ‘‘the ability to learn or understand from experience, ability to acquire and retain knowledge’’ (3). Applied AI is about programming computers to perform tasks previously assumed to require human intelligence. The usefulness of AI is the measure of its success. The key issue is realization that knowledge must be represented explicitly in terms of its nonalgorithmic contents. The computer program acts upon it by deduction and reasoning applying various search algorithms. There is a need to create software products representing an artiﬁcial expertise—a container for limited-domain knowledge. This development is particularly important in the case when the pool of available experts is limited (or about to be limited in the future). As has been noted in other works, ‘‘an AI system must be capable of doing three things: (a) store knowledge; (b) apply the knowledge stored to solve problems; and (c) acquire new knowledge through experience. An AI system has three key components: representation, reasoning, and learning’’ (3a). ARTIFICIAL INTELLIGENCE—CONCEPTS AND APPROACHES The difﬁculties associated with learning have led people to use other methods for augmenting knowledge bases. Expert J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

304

AEROSPACE EXPERT SYSTEMS

systems (ES) are such an example, where the compiled knowledge and experience of a human expert are used in lieu of having the system develop its own experience, duplicating that of the expert. This is not to say that learning cannot be part of an ES. The methods and techniques of AI are well suited for applications not amenable to standard, procedural, problem-solving techniques. Examples of such applications are where the available information is uncertain, sometimes erroneous, and often inconsistent. In such a case, using quantitative algorithmic calculations may not lead to the solution, whereas use of plausible and logical reasoning may. The approach taken by AI leads generally to a nonoptimal, but acceptable, solutions reached by using rules of thumb and logical inferencing mechanisms. For such an approach, the system is represented by a factual description in the form of chunks of meaningful data (knowledge) related to the system state and by the relationships among those data. An external, domain-independent inferencing mechanism makes it possible to draw new conclusions from existing knowledge resulting in changes and updates of the knowledge base. The AI discipline concerned with these issues is called knowledge representation. There are various paradigms of how to represent human knowledge: predicate calculus, production rules, frames and scripts, and semantic networks. The selected representation scheme must express all necessary information, support efﬁcient execution of the resulting computer code, and provide a natural scheme for the user. AI is concerned with qualitative, rather than quantitative, problem solving. Thus the selected knowledge representation and the used tools must be able to (a) handle qualitative knowledge, (b) allow new knowledge to be created from a set of facts, (c) allow for representation applicable to not only a speciﬁc situation but also to general principles, and (d) capture complex semantic meaning and allow for metalevel reasoning (reasoning about the knowledge itself, not just the domain). A distributed-intelligence system (DIS) is the concept of a system operated by a machine and managed by a human. The human operator is involved in planning, making decisions, and performing high-level functions, whereas the machine portion of the system executes most of the system’s regular operational functions, collects and stores data, and handles routine decision situations with a limited number of options (4). Such an approach requires further research in the area of man–machine interface and physiopsychological aspects related to the stress and anxiety factors. Decision support systems (DSS) are computer systems that advise human operators or are automated systems that make decisions within a well-deﬁned area. The systems are used where similar decision processes are repeated, but where the information to decide upon may differ. Some DSS are known as expert systems (ES): they imitate human expert behavior. Decision procedures of an expert are analyzed and transformed into rules and subsequently implemented into the system. The ES is a computer program providing solutions to problems normally requiring a human expert with an appropriate domain knowledge and experience. The experts are employed to solve problems requiring planning or decision making. They frequently use rules of thumb—heuristics based on experience, analogies, and intuitive rationale to explain the behavior associated with their area of expertise. Development

of ES requires identiﬁcation of the thought process describing how the human expert solves a speciﬁc problem. There are three steps in this identiﬁcation (1) knowledge representation, (2) knowledge acquisition, and (3) knowledge processing (5). To imitate a human expert, a successful ES must reach the solution based on the predeﬁned body of knowledge, despite incomplete or uncertain data; explain and justify how the solution was reached; and communicate with the user and/or environment, thus acquiring new expertise. Some other advanced properties include the ability of natural language communication to support an effective user interface and easy knowledge acquisition. The ability to determine relevance, by either giving a referral or degrading gracefully at the expertise boundary, is another important human characteristic, as are common sense reasoning and breaking rules. The issue of effective acquisition of new knowledge and reorganization of expert knowledge base is closely related to machine learning—another important AI discipline closely associated with cognitive science. Besides the aerospace industry, other successful implementations of ES in such areas as medical diagnosis, manufacturing, and military command and control have shown potential for further expansion. However, the analysis of such systems is, in most cases, time-neutral. That is, most expert systems are not able to look into the future and predict future system behavior. Reasoning on a given problem is applied in the current time instant. Only an integrated DSS, combining the potential of simulation and the reasoning ability of ES, can be used in time-dependent applications (6). Various knowledge acquisition schemes have been developed to extract expert knowledge. Interview techniques and observation of humans performing expert tasks have been used to elicit knowledge. If the problem domain is well deﬁned, the approach leads to a rule-based system, where the expertise is represented as a set of if–then rules. An inferencing mechanism searches the rule base, drawing inferences and asserting/retracting the facts about the problem at hand. For the expert operating a complex system, it is easier to explain general experiences rather than identify speciﬁc rules. This observation led to emergence of case-based reasoning, where various operational cases are acquired and modeled into the system knowledge base. Using the process of induction the ES infers new rules based on the current facts and the experiences from previous similar cases. The system then produces advice and modiﬁes its knowledge base (7). Uncertainty is an inseparable part of any decision process. It may be caused by insufﬁcient understanding of the problem area, by missing or uncertain measurements and observations, or by nondeterministic causal relations. The emergence of Bayesian networks allows researchers to model the problem area with its built-in uncertainties. This approach can result in a directed graph, where the vertices represent the variables of the problem area, and a directed edge between two vertices means the state of one variable inﬂuences the state of the other. The magnitude of this inﬂuence is represented by a conditional probability (8). Another approach to uncertainty is the application of fuzzy set theory, which allows researchers to assign a range of quantitative variables to create qualitative entities subsequently handled by the ES (9).

AEROSPACE EXPERT SYSTEMS

NATIONAL AIRSPACE SYSTEM The FAA has been modernizing the NAS in a massive project initiated in the early 1980s. Several papers (see reading list) related to the application of AI methods and techniques to air trafﬁc control (ATC) in the NAS of the twenty-ﬁrst century describe attempts to formalize the ATC knowledge base and build systems assisting controller operations and training. Most of the systems use a simulation component designed to predict the future state of the system. The FAA provides ATC and the trafﬁc ﬂow management (TFM) functions, collectively referred to as ATM, designed to maximize air trafﬁc throughput and minimize delays while maintaining a safe operating environment. The overall ATM system is designed to give equitable access for all NAS users, while ensuring safe separation of aircraft from each other and from terrain/physical obstacles and restricted airspace. Physical and operational constraints of the NAS include availability of airport runways, severe weather conditions, equipment outage, and heavy trafﬁc demands. These constraints limit the options to accommodate user preferences for ﬂight times, routes, and altitudes. TFM is the ATM function that focuses on managing NAS resources, mitigating contention for scarce capacity, and disseminating information about the anticipated restrictions and demand predictions. TFM is intended to allocate capacity to NAS users in an equitable fashion and allow them to maintain operational control. The ATC functions of the ATM include real-time separation and sequencing of aircraft en-route, during arrival and departure at the terminal, and ground control on the airport surface. En-route controllers in 20 centers of the continental United States provide services for aircraft between the departure and arrival phases of ﬂight at air route trafﬁc control centers. Services include separating aircraft, monitoring trafﬁc ﬂow, implementing trafﬁc management initiatives provided by TFM, issuing trafﬁc and weather advisories, coordinating special use airspace, and providing emergency assistance. The terminal controllers in more than 50 larger metropolitan airport areas provide ATC services to aircraft traveling between an airport ground control and the en-route environment. The terminal area includes one or more airports controlled out of a terminal radar approach-control facility. The services include aircraft separation, sequencing, trafﬁc advisories, and alerts, signiﬁcant weather advisories, and radar vectoring for arriving, departing, and through trafﬁc. Airport ground control is responsible for all vehicles operating on taxiways, aprons, and gate areas. Advanced ATM functionality is being developed to support real-time information exchange between ATC controllers and airlines, facilitating collaborative decision making among airspace users. Advanced ATC decision support tools are also being developed to increase safety, efﬁciency, and ﬂexibility as the NAS system evolves toward free ﬂight. These advanced functions must be integrated into the ﬂow, en-route, terminal, and tower/ground facilities as part of a coherent national system in order to deliver their full beneﬁts (see Web resources). AIR TRAFFIC FLOW MANAGEMENT OPERATIONS The Air Trafﬁc Control System Command Center (ATCSCC) is located in Herndon, Virginia. The current role of the center

305

is to manage the allocation of NAS resources and limit airborne delays. These objectives are accomplished by implementing TFM initiatives: ground delay program (GDP), ground stop program (GSP), miles/minutes-in-trail restriction (MITR), and severe weather avoidance program (SWAP). The center is staffed by experienced specialists with an extensive knowledge of ATC procedures and familiar with the impact of weather conditions and possible TFM initiatives on NAS performance (10). There are a wide variety of computer tools based on available aircraft data to support the specialist operations. Data can be displayed in both character and graphic formats, showing, for instance, all aircraft scheduled in a speciﬁc sector within a speciﬁc timeframe, or all aircraft scheduled to arrive at a speciﬁc airport. The hourly arrival demand for an individual airport may be displayed and printed. Weather conditions are displayed graphically, including areas of limited ceiling and visibility, precipitation, expected thunderstorms, and jet streams. There is easy access to alphanumeric local ground weather reports, altitude proﬁles, and brieﬁngs from radar observation. The trafﬁc management specialist responds to the weather situation and to requests from the major airports in the cluster. In cases when an airport acceptance rate is anticipated to decline (deteriorating weather conditions, airport conﬁguration change) the ﬂow controller may consider implementation of the GDP for that airport. The program can be implemented for any combination of the en-route centers, from adjacent to the entire system. The scope of the program, in terms of duration and affected areas, is based on the current situation and determined as the result of the controller’s knowledge and experience. The GPD software recomputes departure times, and estimates predicted delays. When the computation predicts acceptable delays in the system, the specialist sends the new schedule to the centers and the airlines for implementation. In the case when an airport is unable to operate or experiences severely reduced capacity with already long delays and surplus trafﬁc, the specialist may order the GSP for the ﬂights destined to the affected airport. Both GDP and GSP affect only the aircraft scheduled for later departure. Any action is coordinated with all interested parties before implementation. The shift supervisor has the ﬁnal authority on whether or not the proposed plans are implemented. A regional trafﬁc management unit may request MITR in cases of reduced acceptance rate of the arrival sector caused by weather, trafﬁc volume, or stafﬁng problems. The situation is analyzed and coordinated by the area cluster specialist, and the outcome is conveyed to the affected centers. The role of ﬂow control is limited to a mediation between two adjacent centers. Severe weather conditions en-route may force a center to request more forceful measures as a signiﬁcant rerouting of trafﬁc. A separate cluster of ﬂow personnel manages the implementation of SWAP rerouting. The position is equipped with an additional workstation with a database of airport conﬁgurations under different weather conditions and the preferential routes among all major airports. The main role is to provide coordination for the new routing. There is a signiﬁcant amount of domain knowledge involved in TFM activities. For example, some of the airports require implementation of a GDP/GSP for the entire system, whereas others may restrict the program to the adjacent cen-

306

AEROSPACE EXPERT SYSTEMS

ters only, based on incoming trafﬁc patterns and past system responses. Rerouting must consider the capacity of the affected sector to handle the additional trafﬁc. Decisions result from the anticipated situation and knowledge of the system operation and its behavior. Therefore, most decisions are based on the ﬂow specialist’s personal experience. Expert Systems for Air Traffic Flow Management The FAA Operations Research Service developed SMARTFLO— a knowledge based system. This prototype computer program provides the TFM specialists with a user-friendly support tool capable of suggesting a set of required trafﬁc management initiatives for a speciﬁed time horizon (11). The system reasoning capability was based on the available information (inquiring for more information if necessary), and the base for decision and the line of reasoning and was presented to the user. The system includes encyclopedic knowledge about NAS operations. The basic activities in ES development are knowledge representation, knowledge acquisition, and feasibility prototype. The TFM system is described in terms of its concepts and activities. Among the concepts deﬁned are pacing airport, affected area, weather, volume/demand, TFM initiatives, preferred route, rerouting, and arrival distribution to the airport. TFM activities include acquiring information about the system and weather and subsequently processing the information generating the required TFM initiatives. A GDP initiative is proposed, using heuristics based on the ﬂow controller’s knowledge. The basis is the same: The action is considered when anticipated demand exceeds the expected capacity. However, the numerical values used (e.g., rate, time interval), the area involved, or freedom to use other options (GSP, MITR) are left to the specialist. The initial knowledge-acquisition phase included observation of daily operations and interviews with ATCSCC personnel. The analysis of daily reports including specialist logs, weather reports, listing of trafﬁc management initiatives for the day, and estimates of system performance (delays) provided an additional source of knowledge. As the result of these activities, a knowledge representation scheme (facts and rules) was selected and a prototype was implemented. CLIPS, an ES tool developed by NASA, was used as the inferencing mechanism. Subsequently, the Sun workstation-based graphic interface was added to complete the initial implementation. The advanced prototype of the system, implemented on a Hewlett Packard HP9000 workstation, used TAE+ graphic interface tool with active agents and the object-oriented paradigm. The objects contain information about their relationships with other objects, identiﬁcation, source information, deﬁnition and description, lists of relations, rules, and constraints deﬁning their behavior. The classes of object in SMARTFLO are jobs, agents, resources, operations, models, events, systems, devices, sites, and environment. The SMARTFLO advanced prototype graphic windows allow the user to select the airport/center, enter data, query for system information, and watch suggested TFM initiatives being generated. A user-friendly consultation allows the user to ask ‘‘What should we do?’’, and ‘‘Why?’’. In addition to generating the suggested TFM initiatives and playing what-if scenarios, the system includes an electronic encyclopedia with extensive information about ATCSCC operations.

INTELLIGENT TRAINING Computer-aided instruction (CAI) is a common use of computers in education and training. CAI tools incorporate well-prepared course materials and lessons plans into routines optimized for each student. However, conventional CAI tools are limited to either electronic page-turners or drill-and-practice monitors, severely limiting the overall effectiveness of the system in a situation where declarative knowledge is sought. The incorporation of AI techniques into CAI spawned the creation of intelligent tutoring systems (ITS) capable of modeling the student learning process, drawing conclusions from student problem solving behavior, and modifying the sequence in which material is presented to the student (12). ITS is intended to help individual students identify their speciﬁc weaknesses and rectify them effectively and to be sensitive to the student’s preferred style of learning. The objective of some researchers is to produce entirely autonomous ITS based on pedagogical expertise and the principles in the domain knowledge. The major blocks of a modern simulation-based ITS are (a) simulator, (b) domain expert, (c) student model, (d) evaluator, (e) scenario generator, (f) training manager, and (g) user interface. The simulator represents the real-life system for which the student is being trained. The domain expert contains the body of knowledge that should be presented and taught to the student. It is also used for evaluation of student performance and the overall learning progress. To achieve these objectives, most systems generate and store all feasible solutions to the problems in the same context as the student, so that their respective answers can be compared. The student model contains knowledge about the student’s understanding of the material. This knowledge is extremely important in the decision making process affecting the choice of subsequent tutoring strategies. The evaluation module is used to evaluate the student performance based on the situation assessment derived from the simulation status. The scenario generator is used to generate realistic training scenarios appropriate for the student. The core of the system is the training manager, containing the knowledge about teaching methods. The training manager, based on the current evaluation, selects the next scenario component from the scenario generator monitoring the student’s performance. The module uses a decision making process based on teacher experience and assessment of past training sessions. Finally, the success of an ITS depends signiﬁcantly on user-interface quality. All of the components are coordinated by a communication entity referred to as a blackboard. Any ITS module can place information on the blackboard making it available to all other modules. An ITS is often in the form of a computer-based problem solving tutor, a coach, a laboratory instructor, or a consultant. For the development of an ITS for aerospace (ATC specialists, pilots, astronauts, and airline dispatchers), the most suitable tutoring strategy seems to be a combination of coaching and guided-discovery learning. The student is in full control of the activity for which the tutoring is provided—a simulated version of a real system. Simulation is used because it provides an attractive motivational context for discovery learning. The coaching task of the ITS is to foster the learning inherent in the activity itself by emphasizing existing learning opportunities and by transforming failures into learning experiences.

AEROSPACE EXPERT SYSTEMS

The potential of coaching is evident when ITS is embedded in an interactive, real-time simulation. It must be noted that many simulation-based ITS are strong in their simulation capabilities, but rather weak in delivering an effective tutoring strategy. Air Traffic Control Intelligent Training System The ATC-ITS prototype system (13) was designed to support ATC training by providing the student with an interactive, simulated ATC domain by creating effective training scenarios with exercises matching the student performance level, by coaching the student through the scenarios, and by evaluating student performance. The ﬁrst major task is to build a knowledge base representing the domain expertise of an en-route air trafﬁc controller. The principal methods used are observations, informal and formal interviews, operational task analysis, and simulations. A key element is the advice provided by air trafﬁc controllers with both operational experience and en-route training expertise (14). One aspect of ITS development focuses on the expert system and its interface with a realistic simulation. The other aspect deals with the training methods, controller characteristics, student capabilities and skills, development of training strategies, and assessment techniques for student performance evaluation. The ATC-ITS provides training for en-route ATC students. It is assumed that the trainee has control over aircraft within a speciﬁed training sector. The trainee is responsible for hands-off procedures including receiving incoming and transferring outgoing aircraft. The transfers and clearances must conform to the sector standard operating procedures. The route through the sector must closely follow the ﬂight plan. The students must conform to ATC communication standards and phraseology, secure an efﬁcient and orderly movement of aircraft through the sector, and provide appropriate responses to pilot requests. The trainee is primarily responsible for ensuring proper aircraft separation within the controlled sector. All the activities are monitored and evaluated by the ITS. The simulation, implemented in object-oriented MODSIM language, provides a realistic, real-time, en-route air trafﬁc sector simulation displaying aircraft targets with data blocks, ﬂight paths, and ﬂight strips, as well as controller-to-pilot and controller-to-controller communications. ATC-ITS graphics represents an en-route radar scope. It displays sector boundaries, airways, ﬁxes, restricted areas, targets, data blocks in different forms (limited, full, separation violation, handoff initiated, handoff accepted, radio failure, emergency, etc.). The graphics display responds to mouse operations such as handoff initiation, handoff acceptance, data block move, and altitude or route change for selected aircraft. The communication window simulates the radio communication between the controller and simulated pilots. The air trafﬁc expert controller is an implementation of a rule-based ATC expert model. It can provide expert knowledge relative to ATC decision making. The knowledge base consists of facts and rules. The situation facts are assessed from the simulation module and the user input. The facts represent aircraft status, sector situation, controller workload, pilot and adjacent controller requests, emergency events, and so on. The action facts reﬂect the required action to be imple-

307

mented by the controller as a direct response to the given situation. The rules reﬂect the controller experience in tandem with ATC operational procedures and regulations. The system uses a forward-chaining mechanism with its production rules where the situation facts are the rule premises and the action facts are the rule conclusion. The CLIPS inferencing mechanism uses the Rete algorithm (15) to search the rule base. Assertion or retraction of the facts to or from the knowledge base constitutes the principal knowledge control handling mechanism (16). The ATC rules represented in the knowledge base are used by CLIPS inferencing to arrive at the ‘‘best’’ decision. Firing the rules leaves a trace, which provides the user an explanation about the line of reasoning leading to the speciﬁc recommended action. The prototype system was developed on a Sun workstation with a Sunview windowing system. Expert Systems for Pilots ES in aerospace are often designed to aid the operator of the system, including piloting vehicles. Typically the use of an ES in an operational environment such as piloting can be oriented to supporting the pilot or assessing the pilot. Often, ES that are capable of assessment incorporate some form of feedback or critique to the pilot, further enhancing their usefulness. Pilots perform three fundamental activities while ﬂying. The ﬁrst, and primary task, is control. This relates directly to the pilot’s ability to manage the attitude of the aircraft about the three axes of the aircraft. ES that support control include stability and inertial reference systems. The second activity the pilot performs is navigation. This is control of the movement of the aircraft from point A to point B. Pilots are supported by ES such as autopilots and navigation computers. The third activity of the pilot is the need to communicate information relating to the aircraft. ES can provide data link and information fusion based on the pilot’s current and future activities. Knowledge acquisition for these types of systems can be very difﬁcult, costly, and time consuming. Techniques to improve the process include multiple model integration, indexing, and multiple input summarization. Examples of systems supporting the pilot are Pilot Associate (17) and Hazard Monitor (18). Both systems are based on real-time data acquisition from the aircraft avionics. The system processes the data to create knowledge chunks, used as knowledge base by ES. Control Pilots generally understand that aircraft are dynamically unstable. This is true, even when tendencies toward instability are rendered invisible by avionics designed to override those innate problems. As a result, the human ability to control aircraft attitude is often suspected in control-related incidents. Computing systems supported by expert knowledge about the engineering and dynamics of the vehicle are often employed in these situations. While these may not be purely and exclusively expert systems, they provide an excellent, if somewhat oversimpliﬁed, example. Flight stability ES use a theoretical approach, coupled with a dynamic input, to create a response algorithm. The use of an optimal control model provides the system with a baseline of engineered performance. The ES

308

AEROSPACE EXPERT SYSTEMS

designer would typically use a rule-based approach to implement this model. Dynamic input is provided by both the pilot and the ﬂight computers and by inertial reference computers (if available), allowing a control model to be exercised. Often, a sophisticated formula or Bayesian loop is used to control the limits of the autopilot system. Engineers are concerned with the pilot’s induction of out-of-phase or undampened input to the system, and thus, in some aircraft, such as those produced by Airbus Industrie, the autopilot system will actually retain and enforce control input limits made by the pilot.

the ground to the aircraft. Based on a set of rules developed by industry, a priority is assigned and managed to ensure the timely, accurate, and intelligible delivery of data between the ground station and the aircraft. These ES function on the premise that navigation and environmental information, such as weather reports, may or may not be relevant to the particular aircraft receiving the broadcast. Using rules and realtime information residing in the aircraft itself, the ES monitors, acquires, and posts information for the pilot based on the criticality of the information already in the queue.

Navigation

Pilot Knowledge Acquisition

Navigating an aircraft is sometimes simple and sometimes complex. As with control, a number of ES-type technologies exist to ensure that the task of navigating the aircraft is safe and effective. When coupled with the control computers, the autopilot can take on the task of determining position and guiding the aircraft to its intended destination by returning commands to the autopilot system. This would imply that aircraft are comprised of complex ES networks, which is the case in many advanced turbojet and most military aircraft. Some corporate and general aviation aircraft have similar although less expert systems. When a pilot of any aircraft is operating in conditions that are favorable for seeing and navigating the aircraft by looking out the window, the use of such systems is less critical. In these circumstances, ES are often employed to monitor and assess pilot navigation performance rather than actively navigate the aircraft. In conditions where the pilot relies heavily on ES to navigate the aircraft, such systems are designed to mimic the decision process and control input that would be provided by the pilot. It is common to have an ES that uses both inductive and deductive structures to provide the input to the control algorithms. Most often the system design is reduced to having a known position and a set of control rules derived for the primary task of controlling (discussed previously) which combine to create a desired effect—the future location. This process is implemented in an ES using deductive reasoning.

Methods of capturing knowledge of both control and navigation activities are varied. Certain things are known about the aircraft that are derived from the engineering process. Other systemic effects, such as the role of the pilot, are less certain, and there is a need to capture expert knowledge. One such method is the use of models that present reasonable representations of the expert. Multiple model integration is used to reduce the need to explicitly deﬁne the knowledge for all cases and creates speciﬁc rules that ﬁre in general conditions (environment is deﬁned heuristically). This method employs both detailed and general knowledge acquisition and modeling, while yielding high conﬁdence in the rules that ﬁre. Piloting is well suited for such implementations, because the use of procedural knowledge to induce rules can be used to meet the need for speciﬁcity, whereas the general environmental conditions may be described using generalizations. The use of concept mapping is another method of reducing knowledge acquisition problems in complex situations. Concept mapping allows specialized knowledge in the form of heuristics and problem-solving methods to be explicitly associated by the knowledge users with static facts and general knowledge (19). Speciﬁc to concept mapping is the use of two techniques: ﬁrst, the use of combing multiple input whereby the experts have collectively generated a summary map of the knowledge required for the particular domain; the second technique is that of indexing, which results in the development of themes and key concepts that emerge from the relationships generated in the summary mapping process.

Communication When exploring pilot communication activities, a number of different communications take place where ES are employed. Pilots receive information from the aircraft systems in the form of displays, and send information to each other and to others on the ground. A remarkably clear pattern-of-information needs exists during a large percentage of the time pilots are ﬂying. Using this pattern, ES designers have implemented systems that anticipate and provide the information needed when it is needed. Typical systems include the automated information and crew alerting systems used to monitor aircraft systems, detect trends and anomalies in the system, and alert the crew to the problem. These are truly ES, in that they gather data and, rather than merely responding to it, they analyze it, consider alternative responses, and then initiate action. These ES are found on most transport and military aircraft and are developed using engineering data to derive functional limits, which in turn support both rule-based and inputdriven algorithms. Other forms of ES, which support pilots by managing information, are used to communicate data from

Evaluation of Pilot Performance ESs most often emerge as the result of capturing knowledge of one or more experts in a format that facilitates computational analysis. Typically the analysis is focused on directing the system to carry out certain rules and implement routines that have been programmed into the system. Often the goal of the ES designer is to achieve a human-like behavior from the system. One of the more difﬁcult tasks of the ES designer is to implement a system that, in real-time or immediately following use, can assess the performance of the human component of the system. This is different from simply developing an ES that can inﬂuence the operator during its use. The ES needed to critique human behavior is required to have a set of analysis capabilities that not only relate to the rules and networks used in the system, but a set of principles that have their roots in the psychology of human actions. A number of systems exist that provide such an ability to assess pilot performance. These systems are developed to aid in predicting human performance, critique performance, and

AEROSPACE EXPERT SYSTEMS

quantify performance for use in a variety of future activities such as curriculum design, monitoring and controlling quality, or systems engineering. Conducting knowledge acquisition for assessment requires the classiﬁcation of the ‘‘expert’’ relative to the system user. Pilots operate in a dynamic environment and the variables that comprise the reasoning and observable behavior of pilots is often very abstract or difﬁcult to deﬁne. One of the best examples of an ES used in assessing pilot performance is the U.S. Navy effort to design and implement a model of the pilot in ﬂight-test maneuvers. The pilot model is a combination of skill-based models and rules-based models (multiple model integration) whereby the skill-based portion provides a feedback path and the rule-based portion provides the inference and feed-forward path (20). This ES employs the use of the Rasmussen’s commonly recognized model of human error comprised of the hierarchical categories of knowledge-based, rule-based, and skill-based behaviors thought to lead to human error. Using a quasi-linear model, the system is capable of quantifying skill-based pilot behavior. To resolve discrete decision tasks, the use of a fuzzy-logic scheme is employed, which supports rule-based behavioral assessments and, when combined with the skill-based models, results in inference rules that can be used to derive large classes of pilot tasks. Predicting Performance In a complex system such as an aircraft, the need exists to select the best possible match of pilot and vehicle. To do this, ES are used in ground-based test environments to provide a dynamic environment that adapts to the behavior of the pilot and challenge the pilot based on the correctness of the prior action. These systems utilize psychological principles of human attention and motor skill to create a multiprocessing requirement in the human mind. ES capable of resolving these challenging environments then operate in the background of the test apparatus to provide a predictive model to the expected behavior of the pilot, who is often behind the ES in resolving the problem. The predictive nature of these systems is implemented using both rules-based and neural network structures. The ES controls the timing and the difﬁculty of the task based on a Bayesian process that involves the users input and changing heuristic models that form the initial foundation of the models. The ES also uses a series of timerelated rules that are derived as the system is exercised to control the tempo of the primary task—ﬂying the aircraft. Secondary tasks are driven by the ES control algorithms and the pilots input. Results are compared to the past historical results of others who have been used to establish the baseline for comparison. Quantifying Performance Quantifying performance can be aiding by using ES to standardize measurement and classiﬁcation of pilot performance. An ES used in quantiﬁcation of pilots in training uses a rulesbased approach and an optimal model to infer a score on a pilot’s ability to maneuver the aircraft, given a standard. This standard is used to deﬁne the baseline (quantify the optimal performance) and a simple network of all outcome collected in the past performances is used to deﬁne the pilots expected performance. Using a deductive process, the data collected in the pilot’s maneuvering of the aircraft is then reduced to a set

309

of rules where the antecedent is compared with the optimal performance antecedent and the analysis is performed based on quantitative data. This greatly improves the ability of the ES to acquire new data and weight the data according to the best known performance of the pilot prior to the event under study. The resulting analysis can drive a rules-based model that will then narrow the set of variables, and identify the rules that ﬁred, leading to a reﬁned set of maneuvers that need attention in future training. Another ES used to quantify performance is an induction engine developed for studying accident data. It can be said that most accidents involve pilot error. These accidents, when reduced to data, which has been processed using a qualitative approach, can lead to key rules derived through induction. While this method does not result in immediate ES utility, the rules are necessary for the construction of the network of events that preceded the accident. In applying these rules to the performance of the pilot in a simulated environment, the ES is able to rank the pilot’s performance by quantifying the rues which, in the past, had a high probability of ﬁring if certain pre-accident events were present. Knowledge Acquisition Is Based on the Target User Since the expertise of the user is a known factor in the successful implementation of an ES, expertise is therefore relative to the design strategy. The modeling and design of the ES must be in concert with the level of expertise to be assessed. Building systems that can be ﬂexible and adapt to such requirements require the ES designer to apply elective automation design theories. Typically ES are limited to use by one of three classes of users. The ﬁrst, the novice, ﬁnds the most system utility at a level where rote or procedural knowledge is sufﬁcient, engages the highest use of automation when available, and seeks simplicity in the interface. Second, the journeyman ﬁnds utility in the middle range of the system, partially utilizing manual and automated functions, and will tolerate most levels of complex interfaces with some training and practice. Last, the expert is capable of self-instruction on the system since they operate the system using principles and experience-based knowledge. Experts will not tolerate limited ﬂexibility in the system’s functional implementation. EXPERT SYSTEMS AND SIMULATION IN OTHER AEROSPACE APPLICATIONS The close relationship between aviation and simulation dates from the very beginning of aviation. The widespread use of digital computers and software-based simulations have broadened the scope of aviation simulation. By analyzing computer simulation in aviation and aerospace, we can identify the following main application areas: (a) ﬂight dynamics simulation, (b) ﬂight simulators for training, (c) simulation of air trafﬁc, and (d) simulation of aerospace systems to support control and decision making. Computer simulation is a technique that provides for imitation of real-life situations using computers. Models and simulation have been used to facilitate understanding of the realworld phenomena. Computer simulation is the discipline of designing and executing a model of an actual system with the subsequent analysis of the simulation output (21). In the complex world of aviation and aerospace, building the correct

310

AEROSPACE EXPERT SYSTEMS

model is essential for the simulation designer to have a good grasp of the domain. Conversely, managers responsible for aerospace projects requiring simulation should be aware of the advantages and shortfalls of simulation techniques and methods. A careful approach to the design, implementation, veriﬁcation, and validation of the model must be taken. The simulation experiment must be well designed, and feedback from the real-world system needs to be used as often as possible. Abundant research proposes using a knowledge-based approach to support the simulation. There are many ways that the simulation and ES can be combined: ES can be used as an intelligent front-end to simulation. ES may invoke the simulation run to generate the required system data. Simulation may be used to update timerelated variables in the ES. Finally, simulation can be a technique for ES testing (22). The AI techniques have enhanced the historically strong relationship between aerospace training and simulation. Using AI may reduce the impact of subjective, impatient, overloaded, and error-prone human instructor links in the training process. Efforts are being made to design systems that incorporate techniques of AI and use the full power of dynamic simulation, including animation and graphics. An extensive application of these techniques gives one possible answer to the problem of effective training in situations resembling real-world conditions. Flight Dynamics Simulation The main feature of the ﬂight dynamics simulation is that the aircraft model representing the handling characteristics of the airframe, engines, and the control systems is encoded in the computer. The ﬂight dynamics simulation is based on a rigorous quantitative mathematical model expressed in terms of continuous differential equations. Research on interfacing such quantitative simulation of the aircraft in ﬂight with a qualitative simulation, in an attempt to support decision making, has been presented in (23). The system extracts quantitative data from a mathematical model of aircraft ﬂight dynamics and uses fuzzy inductive reasoning on the qualitative model to recognize the ﬂight accidents. Fuzzy Reasoning (or Fuzzy Logic) is based on the theory of Fuzzy Sets pioneered by Zadeh (9). It extends the conventional logic introducing the concept of partial truth—truth values between ‘‘completely true’’ and ‘‘completely false.’’ Fuzzy Reasoning attempts to mirror the imprecision of the real world by providing a model for human reasoning in which even the truth is not an absolute but rather a matter of degree. Fuzzy Logic has emerged as a key methodology in the conception, design and deployment of intelligent systems. Flight Simulators The ﬂight dynamics model and simulation is the ﬁrst step for creating a ground-based ﬂight simulator, designed to reproduce the behavior of an aircraft in ﬂight. The simple simulators are static ﬂight procedure trainers with the ﬂight instruments driven by the simulation engine. Medium-range simulators use high-ﬁdelity graphics representing a dynamic out-of-the-window view with appropriate terrain database. The high-end simulators use a six-dimensional mechanical platform to imitate the motion cues of the ﬂight. The trainee operating the simulator controls experiences the responses of

the model, displayed on the instrument panel and graphic screens similar to those in the real aircraft. Additional elements to be imitated are the communication system (radio), environmental system (e.g., engine noises), weapon system (in the case of military aircraft), etc. The multidisciplinary nature of a ﬂight simulator requires specialized design skills and fundamental knowledge. The problems related to integrating components representing aeronautical, mechanical, electrical, and computer engineering are of primary signiﬁcance. Simulation of Air Traffic Discrete simulation of air trafﬁc has been used extensively to imitate ﬂow of aircraft through the airspace. Applications range from the movement of the aircraft on the ground to the en-route and oceanic ﬂow. The models describe aircraft behavior in all phases of the ﬂight. Using well-established methods of discrete dynamic simulation, the air trafﬁc models can determine a range of system parameters such as time delays, waiting queues, capacity, and occupancy. They may be used to test different ATC procedures, airspace conﬁguration, airport development, and changes in trafﬁc demand and patterns. Combined with animated high-ﬁdelity graphics, the discrete simulation can be used also as a tool for training and human-factor research. Such models are used as tools in airport planning and design in relation to both aircraft and passenger ﬂow. Often considering individual aircraft is not practical in aerospace strategic models geared toward TFM. The continuous approach where the network of air routes is simulated with ﬂow rate in the links may be used. Techniques of simulation and operation research are used to determine capacities of airways, airports, location of hubs, airline schedules, etc. They are used both for analysis and design as well as for the actual TFM. One of the comprehensive simulation systems is Total Airspace Airport Modeler (TAAM), developed by The Preston Group, combining real-time simulation and an interactive graphic user interface with AI elements. The system supports the construction of aircraft performance data, airport and airspace elements, procedures and strategies, deﬁnition of trafﬁc schedules, and rules for aircraft conﬂict resolution. TAAM simulates the actual movement of all aircraft through the airspace, collecting data on sequencing, delays, and potential airborne conﬂicts. A simplistic ES is used to resolve potential conﬂict between the simulated aircraft (24). ATCoach, developed by UFA, Inc., and TRACON Pro, developed by Wesson International, are two examples of an offthe-shelf simulation system designed to provide a comprehensive training for air trafﬁc controllers. Both systems include elaborate simulation with user-friendly interface, including speech capability and realistic graphics. The knowledge base of the ATC operation is used to support the training component. In ATCoach, the focus of ES is on monitoring the training session and providing domain guidance to the student (25). In TRACON Pro, the AI techniques support planning of airspace events using procedural techniques as researched by Wesson (26). EXPERT SYSTEMS IN AVIATION AND AEROSPACE SYSTEMS Model-based reasoning is another AI approach, which bases system behavior on the behavior of the system subcompo-

AEROSPACE EXPERT SYSTEMS

nents as represented by the frame-based model. Knowledgebased autonomous test engineer (KATE), developed for the National Aeronautics and Space Administration (NASA) by Boeing Space Operations (27), is a generic software shell for performing model-based monitoring, fault detection, diagnosis, and control. The four subsystems are (1) simulation, (2) monitoring, (3) diagnosis, and (4) control. The system originated in the mid 1980s as a tool to support the operation of the launch processing system. KATE was particularly designed to check sensor operation for the Space Shuttle liquidoxygen loading system. The system is based on a model of the sensor structure and diagnoses sensor failures. By separation of the system structure from the component functions, a more generic tool was designed. During the early 1990s, the system started its operational application monitoring the tanking data. The system was redesigned and implemented in C⫹⫹ programming language using popular Motif windowing environment on a UNIX workstation to serve as a part of the vehicle health management system. Yet another facet of ES application is in the area of planning and scheduling. One example of such application is the automatic cockpit crew scheduling developed by Japan Airlines and NEC (28) The system is designed to prepare monthly schedules for ﬂight crews. The system knowledge is represented in frames and rules. The system’s distributed architecture allows it to run inferencing on slave computers, with the master computer serving as a cooperative inference area and the monitor of data integrity. The backtracking technique is used to break a deadlock when the crew assignment can not be found. Another example is an ES tool to support shift duty assignments for an airport staff (29). The rulesbased system combines forward-chaining inference and constraints-relaxation techniques. It produces a timetable starting with the initial assignment and continuing through the iterative improvement process. The prototype has been tested in airport operations.

FUTURE TRENDS ES already plays a vital role in the safety and effectiveness of complex systems. Their future in aerospace includes autonomous vehicles in both military and passenger aircraft; cooperating ES, such as those that would provide separation of aircraft in ﬂight; ATC systems that improve the safety and efﬁciency of airspace use and of airports; and, to some extent, training systems that deliver individualized lessons to students. The need to capture knowledge regarding the human operator in the aerospace system is clear; however, the ability to accurately and effectively describe that knowledge in today’s complex systems is becoming less practical using old techniques. The future of ES design will focus on practical knowledge engineering techniques that use the target system as a means of collecting information and creating knowledge about the users. In such a systemic approach, knowledge engineering will evolve to include knowledge about the human, human systems interfaces, and the systemic effects on human operators interpretation of the system feedback. The use of such developing technologies as neural networks and ES that adapt will be more prominent than in systems in use today. The adaptive system is capable of both induction and adapta-

311

tion—principles of elective automation that expose the need for the designer to consider the individual variance of each potential system operator—thereby providing a system that can have various (inﬁnite) combinations of automated support. As the ES designer elicits knowledge to support the design process, the use of new allocation techniques will arise. This new ES design approach is driven primarily by the need to resolve a design conﬂict involving the expertise levels of the user. Experts exhibit more bias than novices, declarative knowledge forms a better foundation for self-critique, and experts adopt practices in lieu of formal procedures. This shift in the user’s mental attitude creates a need to build an ES that will allow the user to transition from novice to expert over time, while recognizing the shift in bias based on decay of formal knowledge (declarative and procedural) in favor of experience and practices. Failure to allocate functions correctly in this transition-capable system is a breech of the design principle of ﬂexibility in elective automation and will result in substandard human–system performance as both levels. The need to build transitory systems stems from the computer community itself similar to adaptive design principles. The desire to build small-scale systems has existed for decades; however, practical aspects of building a microsystem, such as complexity of design, cost, and production engineering have been seen as limitations. The microprocessor-facilitated size reduction of computing equipment to a level that made applications practical in typical human-system interfaces. This revolution in size has allowed rapid (and sometimes careless) implementation of technology into human–system interfaces. These implementations have changed the way humans understand and interact with systems and, as the generation of users that has grown up without such aids leaves the design workforce, system boundaries adjust to an artiﬁcial, and possibly arbitrary new set of limitations that constrain the next-generation designer. This is the foundation of the need for sound knowledge engineering and functional allocation. New systems are capable of more and the assumption that humans are therefore not capable or encouraged to do as much in the system as in the past will gradually become a pseudo-standard for designers. This will lead to the extinction of human capabilities unless the system design is tempered with an understanding that humans are capable of certain tasks, desire to perform certain tasks, and that allocation of these tasks to ES when humans are capable of them is changing the system dynamics. The correct allocation depends on a number of factors including reliability, cost, and efﬁcacy of the allocation, preference, and systemic risk mitigation. ES that are currently in use, such as those described in this article, will tend to proliferate in the aerospace community. Synergy among these ES is likely to lead to cooperating ES that seek out and elicit knowledge from each other. This introduces another growing area where the future holds a paradox for designers and implementers. The need to assess the efﬁcacy of information are paramount to human decision making. Primary factors affecting the efﬁcacy of the information is the medium, the source, and the timeliness of that information. Relevance of the information to the problem at hand enters as a secondary criterion when there is ample information pertaining to the problem. The abundance of infor-

312

AEROSPACE EXPERT SYSTEMS

mation (contrasted with the dearth of information), deﬁnes the ends of the spectrum in the paradox. The designers job is to foresee the sources, deﬁne the timeliness attributes, and deliver the information to the system inference engine so that the process will continue with a high level of conﬁdence and, yet within a reasonable period of time as deﬁned by the knowledge engineers and the users. At this point it becomes obvious that the designer has set about controlling the ES process and, thus, has limited the system to the boundaries of efﬁcacy that the designer(s) foresee as reasonable. To overcome this contradiction, the ES of the future will use a combination of Bayesian and induction processes to match heuristic models and speciﬁc knowledge seeking to reach theoretical absolute certainty, while settling for something less to ensure that the process continues. Limits of time will be systemically controlled using system-derived rules. For example, an ES can learn to tell time due to the ratio quality of the data when these data are input over a considerable amount of time. ES have high conﬁdence in the longitudinal approach to using ratio data in developing knowledge. ES are considered to be reasoning when either the data type or the duration of the data input is less than continuous and extended. The use of quantitative data can improve the use of short-duration inputs, but the less preferred data types (nominal, ordinal, and to some extent interval) create certainty problems for today’s ES. This is likely to be the next major improvement in expert systems in aerospace applications—the ability for the system to reason using data that are meaningful to human processing, but not yet reliable for computing. BIBLIOGRAPHY

10. A. Kornecki, Artiﬁcial intelligence for air trafﬁc, IEEE Potentials, 13 (3), 11–14, 1994. 11. A. Kornecki, SMARTFLO—Knowledge based system to support central ﬂow operations, in Proceedings of the 37th Annual Air Trafﬁc Control Association Conference, 1992, pp. 862–869. 12. T. Diefenbach, D. Carl, and M. Towhidnajad, Intelligent tutoring and air trafﬁc control training, in Proceedings of the 37th Air Trafﬁc Control Association Conference, 1992, pp. 489–494. 13. A. Kornecki et al., Intelligent tutoring issues for ATC training system, IEEE Trans. Contr. Syst. Technol., 1 (3): 204–211, 1993. 14. V. Galotti and A. Kornecki, Knowledge engineering for an air trafﬁc expert system, in Proceedings of the 36th Annual Air Trafﬁc Control Association Conference, 1991, pp. 207–211. 15. J. Giarratano and G. Riley, Expert Systems, Principles and Programming, Boston: PWS-Kent Publishing, 1989. 16. A. Kornecki, Building of an air trafﬁc control expert system, in M. Hamza (ed.),Proceedings of the IASTED International Symposium on ES, Anaheim, CA: Acta Press, 1989, pp. 217–219. 17. J. M. Hammer and R. Small, Intelligent interface in an associate system, in W. B. Rouse (ed.), Human Technology Interaction in Complex Systems, Greenwich, CT: JAI Press, 1995. 18. E. J. Bass, S. Ernst-Fortin, and R. Small, Knowledge base development tool requirements for an intelligent monitoring aid, in Proceedings of the 10th FLAIRS, 1997, pp. 412–416. 19. D. Snyder et al., Knowledge acquisition of tactical air-to-ground mission information using system mapping, in Proceedings of National Aerospace and Electronics Conference, New York, IEEE, 1992, pp. 668–674. 20. M. R. Anderson, C. Clark, and G. Dungan, Flight test maneuver design using a skill- and rule-based pilot model, in Proceedings of International Conference on Systems, Man and Cybernetics, New York: IEEE, 1995, pp. 2682–2687. 21. P. Fishwick, Simulation Model Design and Execution—Building Digital Worlds, Englewood Cliffs, NJ: Prentice Hall, 1995.

1. M. S. Nolan, Fundamentals of Air Trafﬁc Control, Belmont, CA: Wadsworth 1993.

22. R. O’Keefe, Simulation and expert systems—a taxonomy and some examples, Simulation, 46 (1): 10–16, 1986.

2. A. L. Elias and J. D. Pararas, Potential use of artiﬁcial intelligence techniques in air trafﬁc control, Transportation Research Circular, TRB, National Research Council, Washington, DC, AI Workshop Report, 1985, pp. 17–31.

23. A. de Albornoz and F. E. Cellier, Building intelligence into an autopilot—using qualitative simulation to support global decision making, Simulation, 62: 354–363, 1994.

3. Webster’s New World Collegiate Dictionary, 3rd ed. New York: Simon & Schuster, 1996. 3a. S. Haykin, Neural Networks: A Comprehensive Foundation, New York: Macmillan, 1994; A. P. Sage (ed.), Concise Encyclopedia of Information Processing in Systems and Organizations, New York: Pergamon, 1990. 4. L. L. Smith, The distributed intelligence system and aircraft pilotage, in AI and Simulation, San Diego: Simulation Councils Inc., SCS, 1985, pp. 26–28. 5. A. Kornecki, Simulation and AI in aviation training, in G. W. Zobrist and J. V. Leonard (eds.), Progress in Simulation, vol 2, New York: Ablex Publishing, 1994, pp. 91–122. 6. J. G. Moser, Integration of AI and simulation in a comprehensive decision-support system, Simulation, 47 (6): 223–229. 7. Y. Nakatani, M. Tsukiyama, and T. Fukuda, Case-based reasoning and decision aid for engineering design, in Proceedings of the World Congress on Expert Systems, New York: Pergamon Press, 1991, pp. 369–376.

24. Anonymous, TAAM—Total Airport Airspace Modeller Product Proﬁle, Richmond, Victoria, Australia: The Preston Group, 1991. 25. J. A. Scardina, P. Y. Ryberg, and A. Gerstenfeld, Future ATC automation aids based upon AI technology, Proc. IEEE, 77: 1625– 1633, 1989. 26. R. B. Wesson, Problem solving in the world of an air trafﬁc controller, doctoral dissertation, UT&A, University Microﬁlms International, Ann Arbor, MI, December 1977. 27. C. O. Pepe et al., KATE—a project overview and software description, report, Boeing Aerospace Operations, Kennedy Space Center, FL, March 1992. 28. K. Onodera and A. Mori, Cockpit crew scheduling and supporting system, in Proceedings of the World Congress on Expert Systems, New York: Pergamon Press, 1991, pp. 1–10. 29. K. P. Chow and C. K. Hui, Knowledge based approach to airport staff rostering: a case study, in Proceedings of the World Congress on Expert Systems, New York: Pergamon Press, 1991, pp. 46–53.

Reading List

8. R. O. Duda, P. E. Hart, and N. Nilsson, Subjective Bayesian methods for rule-based inference systems, in Proceedings of National Computer Conference, AFIPS, 1976, pp. 1075–1082.

J. Liebowitz, Introduction to Expert Systems, Santa Cruz, CA: Mitchell, 1988.

9. L. Zadeh, The role of fuzzy logic in the management of uncertainty in expert systems, Fuzzy Sets Syst. (11), 199–227, 1983.

E. Wagner, Artiﬁcial Intelligence and Tutoring Systems, Los Altos, CA: Morgan Kaufman, 1987.

AEROSPACE SIMULATION T. I. Oren, Artiﬁcial intelligence and simulation, AI Applied to Simulation, 18 (1), 3–8, 1986. A. M. Wildberger, Integrating an expert system component into a simulation, AI Papers, 20 (1), 132–135, 1988. R. H. Michaelsen, D. Michie, and A. Boulanger, The technology of expert systems, BYTE, 10: 303–312, April 1985. F. Hayes-Roth, P. Klahr, and D. J. Mostow, Knowledge acquisition, knowledge programming, and knowledge reﬁnement, in P.Klahr (ed.), The Rand Corporation, R-2540-NSF, 1980, Reading, MA: Addison-Wesley, 1986, pp. 310–349. A. Gerstenfeld, Simulation combined with cooperating expert systems: an aid for training, screening, plans and procedures, J. ATC, 30: 33–35, 1988. D. Spencer, Development environment for an ATC expert system, in Transportation Research Circular, TRB, National Research Council, Washington, DC, AI Workshop Report, 1985, pp. 32–37. C. A. Shively, AIRPACK: Advisor for intelligent resolution of predicted aircraft conﬂicts, Transportation Research Circular, TRB, National Research Council, Washington, DC, AI Workshop Report, 1985, pp. 58–64. A. Gonzalez et al., Simulation based expert system for training air trafﬁc controllers, in M. B. Fishman (ed.), Advances in Artiﬁcial Intelligence Research, Greenwich, CT: JAI Press, 1989, pp. 295–308. R. Steeb et al., Distributed problem solving for air ﬂeet control: framework and implementation, in P. Klahr (ed.), The Rand Corporation, N-2139-ARPA, 1984, Reading, MA: Addison-Wesley, 1986, pp. 391–432. P. McKinnon, Living with artiﬁcial intelligence, J. ATC, 29: 23–25, 1987. Web Sites Federal Aviation Administration http://www.faa.gov MITRE Center for Advanced Aviation System Development http:// www.caasd.org National Aeronautics and Space Agency Ames Research Center— Advanced Air Transportation Technologies http://aatt.arc.nasa.gov Massachussets Institute of Technology—Lincoln Laboratory http:// www.ll.mit.edu AI resources http://www.cs.reading.ac.uk/people/dwc/ai.html

ANDREW J. KORNECKI JAMES W. BLANCHARD Embry-Riddle Aeronautical University

AEROSPACE INDUSTRY. See AIR TRAFFIC.

313

AEROSPACE SIMULATION

313

tokens represented military units, were constructive simulations. The sand table has been computerized. It now approximates the mechanics of vehicles and even the cognitive processes of troops. Computer representations of processes ranging from water management to bacterial growth to hypersonic ﬂow are all constructive simulations. Virtual simulation employs live players in a simulated environment. There are still other simulations in which inanimate objects, for example, engines, sensors, control systems, or even entire missiles or unmanned aircraft are operated and tested in a virtual environment. This article addresses virtual simulation, as it applies to the ﬂight crews of aerospace vehicles. Regardless of the purpose of the simulation, the subject is the techniques for creating an effective virtual environment for the human pilot. Simulators are widely used for training. Complete pilot training in a virtual simulator is not practical. A simulator suitable for this purpose, classiﬁed by the Federal Aviation Administration (FAA) as level D, is much more expensive than a trainer aircraft. Level D simulators are produced only for very expensive aircraft and are used for, among other things, transition training of airline pilots to new types of airliners. On the other hand, supplementary use of simulators in ﬂight training has long proved useful. Training pilots to ﬂy by reference to instruments only has been accomplished since World War II by combining ﬂight time with simulator time. Simulators offer some unique training advantages:

AEROSPACE SIMULATION Whenever one process is represented by another, a simulation is in progress. A terminology developed recently by the military includes three categories of simulation: live, constructive, and virtual. In live simulation, actual equipment is operated by live crews. Practicing engine out procedures in an airplane with good engines or training instrument procedures while ﬂying in good weather are live simulations. So are war game exercises played with aircraft and tanks. Constructive simulation replaces both equipment and crews by symbols. The classical sand-table exercises, where

• Reduction of risk. • Reduced environmental impact. • Saving of Time. The simulation can be limited to the maneuver being trained. There is no need to perform a preﬂight check of an aircraft, go through engine start procedure, taxi to the runway, and ﬂy to the practice area before training can begin. No time is wasted on returning, landing, and taxiing back after the ﬂight. The simulator can be reset to repeat a maneuver. For instance, when training landing approaches, the simulator can be reset after each approach, putting it in a position to start another approach. In live training the airplane must be ﬂown around to the initial position, which may take anywhere from 3 min. to 15 min. • Control of Weather. No time is lost due to bad weather. Yet adverse weather conditions can be conjured on demand. • Training Analysis. The simulator can be ‘‘frozen’’ for a discussion between trainee and instructor, then continue to ﬂy from that position. • Repeatability. Flight histories can be recorded and replayed. Beyond individual and crew training, the military uses virtual simulation for collective training. Entire units are trained while both sides of a battle are simulated. Collective training is accomplished by a technology known as distributed interactive simulation (DIS), which involves communications between large numbers of virtual simulators located at separate sites. Each simulator includes in the virtual environment it creates the vehicles represented by other simulators. The ultimate goal is a virtual battleﬁeld on which live, vir-

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

314

AEROSPACE SIMULATION

tual, and constructive simulations can interact. The advantages of DIS (exploited already in the Gulf War of 1991) include:

Display system

yy ;;

Pilot

• Cost. In collective training, modest simulators replace expensive vehicles. There is additional savings in logistics.

Control loader

• Environmental Impact. Live exercises tend to tear up the environment, damage property, and cause loss of life.

• Mission Rehearsal. An attack can be rehearsed in the actual site while that site is still in enemy hands. • Debrieﬁng. The mission can be replayed in simulation for analysis and lessons learned. Potential civilian use of DIS has been identiﬁed in the area of air trafﬁc control. The advantage of training in places and tasks that are not yet within reach is not limited to military mission rehearsal. The astronauts practiced the lunar landing in a simulator before getting a chance to perform it live for the ﬁrst time. Quite apart from training, simulation is a powerful engineering tool. Tentative designs of new ﬂight vehicles are evaluated by experienced test pilots in virtual simulation. Much of the iterative design process by trial and error can take place in an engineering simulator before a prototype is built.

PRINCIPLES AND LIMITATIONS A pilot manipulates the ﬂight controls in response to sensory perceptions. A virtual simulator replicates the physical stimuli that induce the sensory perceptions. These perceptions, or cues, fall into several categories: • Instrument indications. • Visual. This refers to cues obtained by looking outside the vehicle. Instrument indications, even though observed visually, are addressed separately. • Motion. This refers to sensations due to the pilot being moved bodily. Visual indications of motion are included in the category of visual cues. • Tactile. Cues induced by the feel of the ﬂight controls. • Auditory. Cues inferred from the sound of the engine, of the airﬂow, and other sources. The methods of simulating sound in a virtual simulator are no different from the ones used in reproducing music and will be discussed no further. There is an arbitrary demarcation between sound and vibration, the latter being considered a motion cue. At some future time, the technology may be available to induce sensory perceptions by a direct link to the subject’s nervous system. The present article addresses only the creation of perceptions by the normal use of the subject’s sensory organs. Visual cues are produced by displays presented to the pilot; motion cues are created by moving the simulator cab.

Motion platform

Control imputs

• Secrecy. Movement of units, which the enemy is likely to detect, is avoided.

Image generator

Instrument generator

Loader controller

Math model

Washout filter

Figure 1. The pilot of a virtual simulator closes several control loops by providing control inputs to the math model in response to cues. The instrument, visual, motion, and tactile cueing systems are illustrated. All are fed state information by the math model.

Figure 1 illustrates the subsystems that create the various cues. Subsequent sections discuss each subsystem. The engineering premise of virtual simulation is the principle of physical equivalence—that is, that identical physical stimuli induce identical sensations and elicit identical responses. Human sensory organs are subject to the laws of physics like any other sensors. Physical replication of stimuli will ensure replication of cues. This is the basis of the present article. The physical nature and accuracy of the replicated stimuli is addressed in objective terms, as might be measured by laboratory instruments. For instance, terms such as ‘‘resolution,’’ ‘‘adaptation,’’ are used in the optical sense as they would apply to a camera. Even when the stimuli are perfect, the physical approach is open to challenge on psychological grounds, because the pilot knows that the ﬂight is not real. Actually, the physical stimuli produced by virtual simulators are imperfect, and one is faced with assessing the response of a human subject to incorrect and even contradictory cues. It is impossible to describe virtual simulation without alluding to physiological and cognitive processes. However, our quantitative discussion will be in purely physical terms. MATHEMATICAL MODEL An air or space vehicle is a mechanical system. A virtual simulator constructs the state history of this system electronically. The computation must be carried out in real time. In the context of virtual simulation, this means that the computation must keep up with the time history being computed. At one time, only analog computers were capable of realtime computations. Analog computers are largely limited to linear manipulations, which made it necessary to linearize the dynamic equations. The use of linearized equations lingered even with the advent of digital computers, initially because of their limited capacity and later because of habit. At the present writing, even modest computers are capable of

AEROSPACE SIMULATION

integrating the equations of motion of many aerospace vehicles in real time. It is easier to program the full equations than to linearize them. The ﬂavor of a typical mathematical model in a virtual simulator may best be conveyed by an overview of the equations governing a rigid vehicle. A rigid body is a six-degreeof-freedom system. The variables of state are 씮

xe

씮

ve q 씮

웆b

Position of CG in earth cartesian system Velocity of CG in earth cartesian system Orientation expressed as a unit quaternion Angular velocity in body coordinate system

3 components 3 components 4 components 3 components

These variables are subject to the following equations of motion:

x˙e = ve mx˙e = F q˙ = 12 qω b Jω ˙ b + ω b × (J ω b) = M where m is the mass of the vehicle, J is the moment of inertia ជ and M ជ are the force and the moment (a 3 ⫻ 3 matrix), F applied to the vehicle. Orientation can be expressed by specifying the heading, pitch attitude, and bank. These three angles, a variation on the ones introduced by Euler to study the spinning top, are called Euler angles. This is the preferred formalism for human consumption. However, Euler angles are unsuitable for virtual simulation because they develop singularities at (and lose accuracy near) the orientations of facing straight up or down. The preferred way of expressing orientations internally in a computer is as unit quaternions. Quaternions are four components entities, which may be viewed as the sum of a number and a vector. Quaternions obey the normal algebraic rules of addition and multiplication with the product of two vectors being given by V = U × V − U · V U Under these rules, quaternions form a ring. All nonzero quaternions are invertible. A well-known theorem due to Euler states that any two orientations can be bridged by a single rotation. Let the rotation from the reference orientation to the current orientation be characterized by the axis unit vector eˆ and the angle 움. Then the current orientation may be represented by the unit quaternion q = cos 12 α + eˆ sin 12 α This representation has no singularities and maintains uniform accuracy over the entire (curved and compact) three-dimensional space of orientations. However, the constraint 兩q兩 ⫽ 1 must be enforced against truncation errors. Actually,

315

quaternions represent the group SU2 rather than the rotation group SO3, and they cover the space of orientations twice. This detail is of no consequence in simulating a rigid body. The equations of motion, above, must be integrated numerically. Using advanced–retarded Euler integration with a time step ⌬t, this is acomplished by the procedure void step(void) 兵 Airloads(); t ⫹⫽ dt; Ve ⫹⫽ Ae*dt; Xe ⫹⫽ Ve*dt; Omegb ⫹⫽ Jin*(Mb ⫺ Omegbˆ(J*Omegb)*dt; q ⫹⫽ (q*Omegb)*(0.5*dt); q ⫽ q/abs(q); 其;

This is actual C⫹⫹ code, making use of the user-deﬁned types (classes) of vector, matrix, and quaternion. The global variables of state are declared as

void step(void) 兵 double t; vector Xe, Ve, Ae, Omegab; matrix J; quaternion q; 其; The symbol ˆ denotes the vector product. Arithmetic operations are overloaded for the user-deﬁned types. Thus * denotes the product of numbers; of a number by a vector, a matrix or a quaternion; of a matrix by a vector; of two matrices; or of two quaternions. The compiler determines the correct operation based on context. For the product q*Omegb (a quaternion by a vector), the compiler converts the vector to a quaternion and employs quaternion multiplication. The overloaded operations of addition and multiplication of vectors, matrices, and quaternions are deﬁned in appropriate header ﬁles (1). The procedure Airloads() computes the earth acceleration Ae and the body moment Mb. Aerodynamic computations are usually based on tables of coefﬁcients and on the local ﬂow ﬁeld. Often, steady-state aerodynamics for the instantaneous state is used even in transient conditions (adiabatic assumption). Computational ﬂuid dynamics (CFD) is, at this writing, incapable of real-time performance. Methods of integration more accurate than Euler’s are often employed. The powerful Runge–Kutta methods are not suitable when control inputs are sampled only once per step. However, the Adams–Bashforth methods that infer trends from previous steps have been used to advantage. In many cases, describing the vehicle as a rigid body is not adequate. Examples include helicopters, where ﬂapping and ﬂexing of rotor blades is important, and large aircraft and space structures, where structural modes interact with the control dynamics. In these cases, additional state variables and additional equations of motion are brought into play. The engine and other systems require modeling, too. TIMING ISSUES The computation cycle including the sampling of control inputs, the supporting calculation of forces and moments, the

316

AEROSPACE SIMULATION

integration over a time interval ⌬t, and the output to the instrument, visual, motion, and tactile cueing systems is called a simulation frame. All the computations for the frame must be accomplished within the time period ⌬t. Timing may be accomplished by clock-generated interrupts at an interval of ⌬t. The interrupt starts the frame. Once the frame is complete, computation is suspended until the next interrupt. This method ensures precise timing but, inevitably, wastes some capacity. Another approach is to run the frames continuously and adjust ⌬t to agree with real time. This ensures the smallest possible ⌬t while maintaining real time on the average, although individual frames may vary slightly. The time step used in integrating dynamic equations must not be excessive, in the interest of accuracy. Models of ﬂexible and articulated vehicles place additional burden on the host computer, due not only to the additional degrees of freedom but, more signiﬁcantly, to the higher frequencies that come into play. The rule of thumb is that the frame rate must be at least ten times the typical frequency of the system being modeled. Frame rates for modeling rigid vehicles are typically between 30 and 60 frames per second (fps). However, for helicopter rotors, frame rates as high as 120 fps are common. The frame rates of different subsystems of a simulator need not be the same. Even when the dynamic computation requires 120 fps, the visual display may be adequate at 60 fps or even 30 fps, while the motion system and control loader may run signiﬁcantly higher frame rates, sometimes as high as 5000 fps. The frame rates of subsystems must be commensurate when precise interrupt-driven synchronization is implemented. Another timing issue involves the interval between control input and observable feedback. The key concepts here are (2,3): • Latency—the excess delay of simulator response over ﬂight vehicle response • Transport delay—the delay between control input and simulator response, including computation time but excluding any modeled delay The transport delay is easier to determine, because it does not require access to the ﬂight vehicle. If the math model is perfect and reproduces the delay inherent in the vehicle exactly, then the transport delay is equal to the latency. The principle of physical equivalence requires zero latency. It is impossible to have the transport delay at zero, because computations do take time. Some compensation is achieved by not modeling the propagation time of control signals in control rods, wires, and hydraulic lines (at the speed of sound in the particular medium). Still, control responses in virtual simulators are typically delayed. The pilot expects feedback to control inputs. If this feedback is delayed, the pilot may be induced to increase the input. A delay in any cue will tend to exaggerate the control inputs. In the context of harmonic inputs and disturbances, the delay is translated into a phase lag and it limits the frequency of disturbances that can be controlled. The FAA accepts a latency of 150 ms for airplane simulators (2) and 100 ms for helicopter simulators (3) for level D certiﬁcation. Practical experience indicates that simulators subject to this amount of delay are effective. The helicopter value, 100 ms, is representative of the state of the art at this

writing. Current simulators of high performance military aircraft also keep the transport delay to less than 100 ms. Apart from the amount of the delay, there is the issue of the relative delay of different cues. The relative timing of visual, aural, and motion cues is important. Cues received out of order may cause simulator sickness—a condition where an experienced pilot becomes nauseated in the simulator. COCKPIT DISPLAYS Flight and engine instruments are the cueing devices that are easiest to implement in a virtual simulator. The Link trainers of WW II fame used analog computers to drive needles in electrically actuated replicas of airspeed indicators, altimeters, tachometers, and other airplane instruments. The devices were used to teach control of an airplane by sole reference to instruments, which made visual displays unnecessary. The Link devices had rudimentary motion capability of questionable ﬁdelity. The task was to train a pilot deprived of visual cues to disregard motion cues and react to instrument readings only. A number of postwar ﬁxed-base devices whose general architecture was the same as that of the Link device accomplished the same end. They were useful in teaching instrument ﬂight and maintaining instrument proﬁciency and were accepted by the FAA for training and currency credits. With the advent of microprocessor technology, even low end simulators became digital. Computer graphics made it possible to use graphical images of instruments in place of hard replicas. The ﬁrst FAA-accepted device to exploit this capability was the Minisimulator IIC, which came on the market in 1981. The IIC used most of its computational throughput to create the crude two-dimensional graphical representation of the instruments. But graphics techniques soon improved, and graphically displayed cockpit instruments became commonplace in actual cockpits as well as in simulators. In addition to instruments, many modern cockpits include other displays. Some, like moving maps and horizontal situation displays, are two-dimensional. Others, such as low-lightlevel TV (LLTV) and forward-looking infrared (FLIR), offer a view of the three-dimensional outside scene. The three-dimensional graphic displays are computed by the same methods as visual displays discussed in the next section. IMAGE GENERATION Creating a visual display of the outside scene is by far the most computationally demanding task in a virtual simulator. Early image generators (IG) used analog methods. A television camera would ‘‘ﬂy’’ under computer control over a miniature scene or an aerial photograph. Early digital image generators offered night scenes with only discrete points of light visible. The technology soon advanced to dusk and eventually to daylight scenes. Data about the three-dimensional environment in which the ﬂight takes place is kept in a database. Terrain and other objects are described as ‘‘wireframes’’ delimited by polygons. Each polygon is endowed with color and/or texture. There have been efforts to create an open database format; at this writing, the formats in use are mostly proprietary.

AEROSPACE SIMULATION

Ψ

Object Screen

Θ

Eyepoint

Figure 2. The three-dimensional scene is transformed into a twodimensional graphic on the image plane by projecting along rays that meet at the eyepoint.

The screen image is a projection of the three-dimensional scene on an imaginary screen by rays converging at the intended viewer’s eye (Fig. 2). Different shapes of the two-dimensional display are in use. However, for simplicity, this discussion addresses a rectangular screen which is placed to subtend a pre-selected ﬁeld of view (angle ⌿ by angle ⌰ in Fig. 2). When the image is presented in the simulator (next section), it should cover an equal portion of the pilot’s ﬁeld of view. The methods employed by image generators for ﬂight simulation are similar to the ones used in other computer graphic applications that produce perspective views of three-dimensional objects. The speciﬁc tools used by manufacturers of IGs are usually proprietary. The overall approach is best illustrated by the OpenGL language (a public domain offshoot from the proprietary IrisGL) (4). OpenGL supports a transformation formalism based on 4 ⫻ 4 matrices. These represent not only the Euclidean group (translations and rotations) but also afﬁne and projective transformations (5). This formalism supports the projection shown in Fig. 2. It can also create an image so that it appears correct when projected from one point and viewed from another. This is pertinent with front projections, since the projector and the pilot’s eye cannot be collocated. The transformation between the projector image and the viewed image is known as ‘‘distortion correction.’’ A more complex instance of distortion correction arises with spherical screen. High-end image generators perform that transformation, too. The projection in Fig. 2 represents ‘‘one channel’’ of an image generator. The image generator may offer several channels. A wide ﬁeld of view may be created as a mosaic of several adjacent or partly overlapping channels. Still, the ﬁeld of view in a simulator is usually restrictive in comparison with the aircraft. A typical channel might drive a raster display of 1280 pixels by 1024 pixels and cover a ﬁeld of view of 40⬚ ⫻ 30⬚. This choice makes each pixel subtend 1.9⬘ (1.9 minutes of arc) and effectively limits the simulator pilot to 20/40 vision. Physical

317

equivalence would dictate that 1⬘ be resolvable, corresponding to the 20/20 vision required of airmen. However, 2⬘ or even 3⬘ resolution is representative of current simulator practice. Many arguments can be raised to rationalize the contradiction between accepting a 20/40 or 20/60 simulator while insisting on 20/20 vision for the pilot—for example, that the performance of most individual tasks does not really require 20/20 vision and that most of the collective training in simulators is for night and adverse weather conditions. In reality, this policy is driven by supply and demand. A 20/20 simulator would be exorbitantly expensive, whereas humans with 20/20 vision are plentiful. The display for our typical channel consists of 1,310,720 pixels. The image generator must specify the color of each pixel. At 32 bits per pixel, the ‘‘local buffer’’ comes to 5.24. A double buffer is required for smooth operation: the image generator redraws the picture in a hidden buffer, leaving the one being displayed undisturbed. Once the updated picture is complete, the buffer pointers are switched. Thus a channel requires 10.5M of memory. A depth buffer is also required. It is also called ‘‘z buffer,’’ because, by convention of computer graphics, the coordinate system is oriented so as to make the depth (the distance from the viewer) the z coordinate. The z buffer is a scratch pad that the image generator keeps for itself. The depth of the surface that generated the pixel stored in the local buffer is kept in the z buffer. For each pixel, the image generator goes over the database and, for each polygon, determines whether that polygon is intersected by the ray corresponding to the given pixel. If so, its depth z is computed and compared with the value in the z buffer. If the new object is closer, the pixel is rewritten to represent it; otherwise it is not. The buffers are initialized to a background color in the local buffer and ‘‘inﬁnity’’ in the z buffer. The z buffer occupies another 5.24M. This amount of memory has become commonplace. But the task of reworking it at the required rate is still a challenge. The picture, which appears to be moving continuously, is computed as discreet images. Each image is traced over the screen in a manner similar to other computer displays. Two rates govern: • The refresh rate, at which the screen is retraced • The update rate, at which the content of the picture is updated A refresh rate of 60 Hz or higher eliminates ‘‘ﬂicker.’’ The refresh rate also sets a practical bound on the update rate. Even when dynamic computations are carried out more often, the additional information cannot be displayed visually. However, the update rate can be lower than the refresh rate. The smoothest motion is obtained when update and refresh are synchronized—that is, when the refresh rate is divisible by the update rate. For a refresh rate of 60 Hz, this consideration would allow an update rate of 60 Hz, 30 Hz, 20 Hz, 15 Hz, 12 Hz, . . . . The update rate that is required in a simulator varies with the task being simulated. Most demanding are the tasks that involve rapid change in the scene. This occurs during rapid angular motion of the vehicle. In the case of head- or helmetmounted displays (next section), rapid change in the scene can be caused by brisk head movements. An update rate of 60 fps is adequate for most purposes. Lower rates are accept-

318

AEROSPACE SIMULATION

able in many cases. Sensitivity to update rate varies with the individual subject. Image generators are required to perform several additional tasks: • Moving Models and Articulated Parts. Other vehicles must be displayed at changing locations. In some cases, articulated parts, such as the turret on a tank, must be seen moving. • Terrain Elevation and Slope. It is the image generator that has direct access to the model of the terrain. The host requires the terrain elevation and possibly slope for use in the ownship ground contact model. It is up to the image generator to supply these. • Color Variations. The same scene may need to be rendered in different colors to represent day, night, or sensor images. • Text and Symbology. At times, it is desired to have the image generator create the text and symbols contained in a heads-up display (HUD) and superimpose them on the scene. All of the above have negligible impact on image generator performance. The cost of rendering the polygons making up the moving models and parts is the same whether they are moving or not. The added burden for all tasks, above, is on the communications of the IG with other computers. Other functions, such as accurate simulation of light sources with correct rendering of shadows and highlights, are demanding of the IG. Correct simulation of limitations to visibility by mist or smoke is likewise expensive. On the other hand, a crude representation of the same effects is helpful in that it eliminates the labor of rendering the obscured objects. DISPLAY SYSTEM The two-dimensional projection of Fig. 2 must be presented to the simulator pilot within the original viewing angles ⌿ and ⌰. This may be accomplished by a small image nearby or a larger image further away (Fig. 3). It may be a ‘‘real image’’ projected on a screen or traced on a CRT or a ‘‘virtual image’’ created by optics. A real image is limited by practical considerations to be within a few meters from the eyepoint. A virtual image can be as far away as desired and even inﬁnitely far. With the pilot’s eye at the eyepoint, all the images in Fig. 3 create the same impression on the retina, with the same resolution. But there are signiﬁcant differences: • Accommodation. The pilot’s eye must accommodate optically to the distance at which the image is located rather than the real-world distance of the objects it contains. Should the pilot need corrective lenses to aid in accommodation, these would not necessarily be the same in the simulator as in ﬂight. • Parallax. Even when seated, the pilot’s upper body and head is free to move to some extent. As the eye moves, nearby objects (e.g., the cab environment) change their apparent position relative to objects that are further away. With the simulator display this would be governed by the distance of the image rather than the distance to the objects it represents. Objects in the image will not

Ψ

Eyepoint

Screen

Θ

Screen

Screen

Figure 3. Image planes at varying distances from the viewer create the same impression on the retina, with the same resolution. But accommodation, parallax, and stereopsis effects differ and betray a close by image for what it is—a small, ﬂat picture.

move relative to each other. Should the pilot’s eye deviate from the nominal eyepoint, the perspective would become distorted. During forward ﬂight this would create the impression of a spurious sideways component of motion. • Stereopsis. When the pilot’s two eyes observe the same image from slightly different vantage points, the two retinal impressions differ. This difference is the raw material for stereopsis, which determines apparent distance. The distance so determined is that of the image rather than of the objects it represents. The stereopsis cue might conﬂict with other cues—for example, perspective cues and cues based on the size of familiar objects. These effects are most pronounced with a small, nearby display, such as a monitor screen. They ﬂag the image as a small, ﬂat picture. A human being can transcend this detail when appreciating art. To some extent, one can transcend it during training of speciﬁc tasks. Screen displays as close as one meter have been used successfully and accepted well by experienced pilots. However, to attempt physical equivalence, one must do better. This is where the display system comes in. A screen projection is a signiﬁcant improvement over a monitor screen. The image may be projected either from the front of the screen or, with a suitable screen, from the rear. Back projection has the advantage that the projector is out of the way of the pilot and cab structure. It is possible to place the projector so as to avoid distortion and the need for distortion correction. A larger image placed, typically, three meters away is easier to perceive as real. The accommodation is only 0.3 diopter from inﬁnity. Parallax with nearby objects, such as the cockpit structure and instruments, is approximately correct. Inﬁnity optics is a more effective solution. The image is optically placed inﬁnitely far away. Accommodation is exactly

AEROSPACE SIMULATION

correct for distant objects as is parallax with the cab environment. To avoid color fringes, inﬁnity optics must employ mirrors rather than lenses. A collimator, illustrated in Fig. 4, is a common example. The monitor is set at 90⬚ to the pilot’s line of sight. A ‘‘beam splitter’’ semireﬂective glass plate, set at 45⬚, reﬂects the monitor screen into the concave spherical mirror. The pilot views the mirror through the beam splitter. The monitor face is at the mirror’s focal point (half radius as measured along the broken optical path). Light originating from a point on the monitor comes out of the mirror as a parallel pencil of rays, putting the image out at inﬁnity. A collimator typically covers the ﬁeld of view of one channel. Three channels may be combined by a battery of three collimators set at an angle to each other. Such batteries are designed with ‘‘overﬁll.’’ This means that the pictures in adjacent monitors overlap. When the pilot’s head moves, parts of the scenery that were near the edge of one collimator are now seen in the other. This way, the three collimators offer a seamless combined view. The collimated image at inﬁnity can be seen only when the viewer’s eye is within the fairly narrow collimated beam. Collimators act as funnels with an opening to the distant scene. Eyepoint movement does not distort the scene, but excessive movement blocks it. Two pilots cannot share a collimator. They must be given two separate collimators even when the same IG channel drives both with an identical image. Supplying more than one crewmember with a wide ﬁeld of view is impractical because of mechanical interference of the systems of collimators. Collimators cannot match the ﬁeld of view offered by a spherical dome that encloses the pilot and makes a borderless projection screen. But sharing of a dome or screen projection by two crew members is problematic. The basic image at inﬁnity is the same, but the distortion correction is different for the two eyepoints.

Video monitor

m

ire

fle

ct

i ve

pl

at

e

Spherical mirror

Se

Figure 4. A collimator serving as inﬁnity optics. The monitor faces down. The screen is reﬂected into a concave spherical mirror by a diagonal semi-reﬂective glass plate. The pilot views the mirror through the plate. The mirror creates an image located at inﬁnity.

319

Projector Spherical screen Spherical mirror

Motion

m platfor

Actuator

Figure 5. A six-post motion platform is capable of six DOF motion. The platform carries a simulator cab and a display system with wideangle inﬁnity optics. The display system employs back projection on a spherical screen which the crew views reﬂected in a large spherical mirror.

Figure 5 shows an elegant solution: an inﬁnity optics system that can serve several crewmembers and provide them with a correct, wide-angle outside view regardless of their position in the cockpit. The picture is back-projected by a number of projectors (only one is shown) onto a spherical screen. The simulator crew views this display through a large concave spherical mirror. The screen and mirror are concentric with their radii matched to put the screen at the focal surface of the mirror as viewed from the cab. The mirror creates a virtual image located out at inﬁnity that can be seen from anywhere in the cab. Neither the projected image nor the one viewed through inﬁnity optics offers correct stereopsis, parallax, or accommodation for objects that are not far away. This is signiﬁcant for operations where nearby objects play a role, including aerial refueling, spacecraft docking, and maneuvering helicopters near terrain and objects. Stereopsis can be achieved by offering separate images for the two eyes. When this is done, the stereo cue is expected to overpower the accommodation cue and the parallax cue with which it is not consistent. Three-dimensional images that are inherently correct in stereopsis, accommodation, and parallax for any viewer and for multiple viewers at the same time can be produced by holography. But holography requires creation of an interference pattern with resolution of the order of the wavelength of visi-

320

AEROSPACE SIMULATION

ble light (in the order of 10⫺8 m). This capability is not yet available in real time. Separate images for the two eyes (or for that matter, for two crew members) can be offered with projection systems and inﬁnity optics systems by use of polarized light or of electronically timed shutters. In the former case, two separate images are projected on the screen using mutually orthogonal polarization. The pilot views the display through polarizing lenses, so that each eye sees only one image. In the latter case, the two images alternate. The pilot views the display through electronically timed liquid crystal shutters. These block each eye when the image intended for the other is projected. Head (or helmet)-mounted displays (HMD) offer separate collimator-like display systems for the two eyes. The HMD requires head tracking to determine the instantaneous orientation of the eyepoint. Head movement can sweep a narrow ﬁeld of view over a much wider ﬁeld of regard. These systems typically induce the pilot to substitute head movement for eye movement, and the natural ability to notice moving objects in one’s peripheral vision cannot be exercised.) The quality of HMD depends on the precision of head tracking and its latency. The display requires a fast update rate to keep up with fast image changes due to abrupt head movement. HMDs typically require individual ﬁtting. The size and weight of an HMD is a burden on the civilian pilots. Even military pilots, used to ﬂying with a helmet, often object. Besides, the HMD precludes the use of operational helmets and viewing devices in the simulator. The eyepoints used for the HMD are generic. They represent the eye positions of a typical pilot. Static adjustment to the pilot’s seat position, torso height, and eye separation is feasible. Dynamic adjustment to body and head movement is not in the current systems. For use with an HMD, the database models the inside of the cab as a black silhouette. The HMD reﬂects its images on beam-splitters that allow the pilot to see through into the cab. Even so, there is a potential problem when two crew members sit side by side. The silhouette of the other crew member’s head cannot be predicted perfectly and will not register accurately. Bright outside scenery may ‘‘show through’’ the edges of the other crew member’s helmet. Brightness is an issue for all simulator displays. One must assess the brightness available at the source and how much of it reaches the observer’s eye through the display system optics. These estimates are too involved to be presented here. The bottom line is that there is no difﬁculty in creating what an observer will accept as a daylight scene. The brightness of this scene is far below actual daylight. Pilots do not use their sunglasses in simulators. Simulator cabs are darkened during operation unlike aircraft cockpits in daytime. By the same token, problems of observing certain dimly lit displays in sunlight do not arise in the simulator. It was not possible to describe in this section all the types of display systems in current use. Some of the ones not covered are calligraphic displays, multi-resolution displays, and area of interest displays. MOTION CUES Motion cues, by deﬁnition, are those cues that result from the pilot being moved bodily. Awareness of motion through sight

or sound is excluded. It is a basic law of nature that, without reference to external objects, uniform rectilinear motion is undetectable. It is also a basic law of nature that, without external reference, the effect of acceleration is indistinguishable from that of a gravitational ﬁeld. What is locally measurable is speciﬁc force, which is an effective acceleration of gravity given by s = g − a 씮

씮

where g is the local acceleration of gravity and a is the acceleration of the cab relative to an inertial system. Rotation relative to an inertial frame is also measurable. It is these parameters, namely, the three components of speciﬁc force and the three components of angular velocity, that serve as motion cues for the human body, as for any other physical system. The inner ear contains organs (otholiths and semicircular canals) speciﬁcally adapted to sense these parameters. The motion parameters are felt also by other parts of the body—for example, the sinking sensation in the pit of the stomach when an elevator starts its descent. So long as the six motion parameters are reproduced correctly, there is no need to investigate the mechanism of human perception. Any and all mechanisms respond as they do in ﬂight. It takes a six-degree-of-freedom motion system to create the six motion cues. With the simulator cab on a motion platform, the pilot can sense rotational rates around the three body axes (yaw, pitch, and roll) and linear acceleration forward, sideways, and up (surge, sway, and heave). The six parameters vary from one point in the moving cab to another. However, with a rigid cab representing a rigid vehicle, if the parameters are correct at one point, they are correct at every point. When the replication of the motion parameters is only approximately correct, the errors vary from point to point in the simulator cab. It is then necessary to select a sensing point where the errors are minimized. The choice of a sensing point is inﬂuenced by the theory of perception. For example, if it is the inner ear which processes the motion cues, then the sensing point should coincide with the pilot’s head. The fact that uniform motion is intrinsically undetectable allows a pilot to have the same sensations in a stationary simulator as in a fast-moving airplane. However, acceleration and rotation are sensed. It is impossible to replicate the acceleration of the ﬂight vehicle exactly while keeping the motion platform in the conﬁnes of a room. For instance, during the takeoff run, an airplane accelerates from rest to ﬂying speed. In the process, it might roll over a few thousand feet of runway. Should the motion platform be subject to a surge acceleration equal to the airplane’s, it, too, would translate a few thousand feet and out of the conﬁnes of the building that houses the simulator. The above discussion demonstrates that a conﬁned motion platform, of necessity, violates the principle of physical equivalence under some circumstances. One attempts to replicate the motion cues approximately, and, to the extent possible, deviate from the true motion parameters to a degree that is undetectable by a human subject. In the case of the takeoff roll, the speciﬁc force, in body coordinates, is inclined to the rear and is slightly larger than

AEROSPACE SIMULATION

1 g. The motion platform, conﬁned to a small space, cannot replicate this condition. The platform can be tilted to a noseup attitude so that the direction of the speciﬁc force is correct. The magnitude remains 1 g. However, the small difference may not be obvious to the pilot. There remains the problem of achieving the tilt at the onset of acceleration. This must be done slowly, at a rate below the pilot’s threshold of detection. The translation of the vehicle motion as computed by the math model to a motion command for the motion platform is accomplished by a washout ﬁlter. The functions of the washout ﬁlter are to: 1. Limit commanded linear and angular motions to platform capability 2. Slow the platform near its limits to avoid banging into the stops 3. Stealthily return the platform to mid-range 4. Tilt the platform to simulate sustained surge and/or sway acceleration Items 3 and 4 should be accomplished at rates below the pilot’s detection threshold. The tilt due to item 4 should be combined with the instantaneous orientation so as to ensure the correct direction of speciﬁc force. The equations for accomplishing this are delicate. Many practical simulators approximate this procedure by merely superimposing Euler angles. The most common conﬁguration of a motion base is the ‘‘synergistic’’ or ‘‘six-post’’ arrangement. Motion platforms of this design are used for both training and engineering simulators. As shown in Fig. 5, the platform carries the simulation cab and the visual display system. A high-end ‘‘six-poster’’ might be rated for a load of 2 tonnes or 3 tonnes. It can provide linear accelerations as high as 1 g, angular accelerations as high as 150⬚/s2. However, linear displacements are limited to under 1 m and angular displacements to 15⬚ or 25⬚. Some unique motion platforms at research facilities can do better. The Vertical Motion Simulator (VMS) at the Ames Research Center of the National Aeronautics and Space Administration (NASA) allows 2.4 m of surge, 12 m of sway, and 18 m of heave. Motion systems are also characterized in terms of response to harmonic inputs. The recognition of particular motion cues, such as the bumping of the left main tire against the runway, depend on undistorted transmission of fairly high frequencies, up to about 50 Hz. For this reason, the computation systems driving the motion platform must compute at a rate of 앒500 fps or higher. Until quite recently, analog systems were used to meet this requirement. The phase delay is a separate issue, which is pertinent for motions that the pilot manually controls and damps. A human subject cannot do this consciously above 앒1 Hz and probably a little higher for subconscious tasks. The phase lag is a direct function of the latency. A 100 ms delay translates into 90⬚ of phase lag at 2.5 Hz. Typically, the phase lag reaches 90⬚ at 1.5 Hz or less. The frequency response depends on the mechanical system driving the motion platform as well as the computer system and the washout ﬁlter. Most high-quality motion systems are driven hydraulically. However, electric systems have advanced recently and now occupy the low end of the price range.

321

The motion amplitudes of a six-post platform is sufﬁcient for simulating a transport aircraft that maneuvers gently. (However, the very low frequency heave cues in the landing ﬂare may be truncated.) The VMS has been used extensively in the study of helicopters. Neither system is capable of the sustained high speciﬁc force (‘‘high g’’) that ﬁghter aircraft develop during steep turns and other vigorous maneuvers. These can be developed by different designs of motion platforms that act as a centrifuge. However, the unwarranted high rate of rotation in a centrifuge presents a problem. Another condition that a conﬁned simulator cannot sustain is that of 0 g, or weightlessness. It can be sustained for a minute or two in an airplane ﬂying a parabolic trajectory. This was used by NASA to expose the astronauts to weightlessness in advance of space ﬂight. The ability of an aircraft to induce a wide variety of speciﬁc force conditions suggests the in-ﬂight simulator—the use of an aircraft as a ﬂight simulator. One aircraft can simulate another. The astronauts learned to control the Space Shuttle in a modiﬁed Gulfstream G2 that simulated it. However, the subject of in-ﬂight simulation is outside our scope here. Most military simulators of ﬁghter aircraft are ﬁxed base. Cues of high speciﬁc force, typical of ﬁghter aircraft, are transmitted to the pilot through the pressure suit that ﬁghter pilots wear. ‘‘High g’’ has the effect of driving the blood into the legs and lower body and away from the brain. The pressure suit counters this effect by increasing the pressure over the legs and lower body in response to the level of speciﬁc force. In a simulator, even though the speciﬁc force remains at 1 g, the suit inﬂates in response to the computed speciﬁc force and provides the pilot with a g cue. Use of the pressure suit is a blatant deviation from the principle of physical equivalence. Rather, it is an application of the psychological phenomenon of association. The pilots have become accustomed to associating the suit pressure with high g effects. When the suit inﬂates, the human subject, in the manner of Pavlov’s dogs, may imagine that high speciﬁc force prevails. There are other pseudo-motion devices in use. One is the pressure cushion that inﬂates when increased g is computed. The cushion is supposed to simulate the increased pressure on the pilot’s buttocks. Increased pressure may be experienced when the pilot’s seat belt is secure. Squeezing the tissues between the seat and the belt is not physically equivalent to the effect of increased g. But the pilot does get a cue. The subject of motion in ﬂight simulation is controversial. The FAA insists on motion. Devices without motion are classiﬁed as ‘‘training devices’’ rather than ‘‘simulators’’ and allocated reduced credits. The utility of ﬁxed base simulation is well established in the military and elsewhere. Cases where motion degrades a simulator and induces motion sickness have been observed. The probable explanation is that ‘‘bad motion is worse than no motion.’’ Motion can be ‘‘bad’’ because it is a poor emulation of the speciﬁc force and angular rates experienced in ﬂight; or because of excessive latency; or because it is poorly synchronized with the visual cue; or because it betrays the mechanics of the motion base. How good should motion be? Some idea of the answer may be conveyed by Ref. 6. This experiment used the sway-androll motion of the VMS in a sequence of side-step maneuvers. The data demonstrated the importance of the motion cue. However, scaled-down motion produced objective performance

322

AEROSPACE SIMULATION

equal to the full-scale motion and got better subjective evaluation from the pilots. Not even the VMS was capable of altogether ‘‘good’’ full-scale motion. The motion system of the VMS has been upgraded in the wake of the Ref. 6 results. When consistent motion and visual cues are available, the motion cues should be sensed by the pilot earlier. An acceleration step of magnitude a results in a displacement at2. This displacement is not sensed visually until it has grown to the visual detection threshold ⌬x, which takes a time delay t =

2x a

So long as a is above the acceleration detection threshold, the speciﬁc force due to the acceleration is felt immediately. The importance of the motion cue varies with the task being trained. In my judgment, it is most signiﬁcant with tasks performed subconsciously. The landing ﬂare of a ﬁxed-wing aircraft and the hovering of a helicopter (with no stability augmentation) may rely signiﬁcantly on motion cues.

CONTROL LOADING In the early days of aviation (and to this day in light aircraft), pilot controls were coupled mechanically to aerodynamic control surfaces. Pilots relied on the control feel, dominated by aerodynamic forces, as a major cue. The function of the control loader is to reproduce this cue in a ﬂight simulator. In the meantime, aircraft have evolved. Hydraulically actuated controls have become the norm. Electronic controls are the trend of the future. These irreversible control systems do not feed aerodynamic forces back to the pilot. Artiﬁcial feel systems (usually springs) are used to provide the pilot with a semblance of the expected feel. Increased reliance on instrument readings makes up for the deﬁciency. Control loaders are fairly expensive. A high-quality loader may cost more than a light airplane. This creates a paradoxical situation: a control loader can be economically justiﬁed only in those cases in which the most important cues that it can provide are suppressed. A very sophisticated piece of equipment simulates a generic system of two masses, with springs, dampers, and linkage. This is traditionally approximated by a near linear model. Nevertheless, control loaders are important in special situations—for instance, hydraulic failure, giving rise to signiﬁcant control forces. The techniques of control loading are similar to the ones employed in motion system. The high-end control loaders are hydraulic, with electric systems starting to catch up. Through the 1980s, control loaders were controlled by analog computers. In the 1990s, digital controllers caught up, some of them using frame rates as high as 5000 fps.

THE VIRTUAL COCKPIT Flight simulation may be viewed as a precursor and special instance of the emerging technology of virtual reality (VR). A ﬂight simulator creates a near physically equivalent virtual environment for ﬂight crews enclosed in the cab. VR tends to avoid physical props. It would dispense with the physical cab and replace it with a virtual cockpit. Historically, VR is

younger than ﬂight simulation, and the two communities are largely disjoint. VR provides visual cues through an HMD. The image generator, rather than blocking the inside of the cab, would include it. The HMD would provide the pilot with a stereoscopic image of the inside as well as the outside of the cab. Tactile cues are produced by devices, ranging from gloves to ‘‘exoskeletons,’’ attached to the subject’s person. These are tracked, and they are controlled to apply appropriate forces to the human body. The obvious beneﬁt of this plan is that it would make the simulator generic. Reconﬁguration to any type of vehicle becomes selectable by software. But there are many technical problems to be resolved. All the drawbacks of HMDs mentioned previously apply, and their effect is magniﬁed in relation to the closeby cockpit scene. The gloves and other tactile devices are yet to be proven. Progress will probably start with the instrument displays and move to switches and secondary controls. The physical major ﬂight controls and physical seat will be last to go. VR has been employed in aerospace for prototyping the interiors of vehicles ranging from airliners to the space station. VR, together with visual and motion devices borrowed from aerospace simulation, are making a splash in the entertainment industry. Lay subjects enjoy exciting sensations of presence and motion. But experienced pilots, conditioned to true visual and motion cues, are more critical.

NETWORKING OF SIMULATORS Long-haul networking came into its own in the 1990s. Air combat simulators with dual cockpits engaging one another have been in existence since the 1960s. By the 1980s, several simulation facilities had connected their simulators by a local area network. The concept was taken a step further by the Defense Advanced Research Projects Agency (DARPA). In the SIMNET project (7), large-scale networking, including remotely located facilities, was carried out successfully. The SIMNET project used low-ﬁdelity simulators with crude visual displays. Active controls and instruments were limited to the ones normally used or monitored during combat. Everything else was eliminated or represented by static props and pictures. The purpose was to recreate the feel, the pressures, and the confusion of a battleﬁeld. In a test conducted in 1989, about 400 players participated, including tank crews and helicopter crews at separate army installations. SIMNET achieved its networking in two stages. Local networking tied simulators within one facility together by use of Ethernet. The long-haul link between different facilities used commercial 56 kbaud lines. The local and long-haul protocols were different. Like the local networking that preceded it, SIMNET addressed a set of matching simulators speciﬁcally designed to interact. By 1989, there were also isolated demonstrations of long-haul communications between existing high-ﬁdelity simulators that were separately and independently designed and owned. In 1979, an F-15 simulator located at Williams Air Force Base engaged an F-4 simulator at Luke Air Force Base. Both bases are in Arizona, and the distance between them is 80 km. The network link used four telephone lines.

AEROSPACE SIMULATION

In 1989 a long-haul link between an AH-64 Apache simulator located in Mesa, Arizona and a Bell 222 simulator located in Fort Worth, Texas was demonstrated. The Arizona simulator was in the facility of the McDonnell Douglas Helicopter Company. The Texas device was in the plant of Bell Helicopter Textron. The distance between the two facilities is 1350 km. The link employed a 2400 baud modem over a standard telephone line. These experiments showed that long-haul networking of dissimilar simulators was practical. But a communications protocol was missing. Rather than reinvent the interface by mutual arrangement between each pair of facilities, an industry standard for interfacing simulators was needed. By conforming to the standard, a simulation facility could ensure compatibility with every other facility that conformed. An open industry standard for networking of simulators was ﬁrst addressed at a conference held in Orlando, Florida, in August 1989 (8). The conference adopted the local SIMNET protocol as the starting point for the new standard. The term coined for the new protocol was distributed interactive simulation (DIS). Work on DIS continued in biannual meetings in Orlando. In 1993, the DIS protocol was formalized as IEEE Standard 1278-1993 (9). Work on upgrades continues. The number of players involved in SIMNET was large enough to enforce some of the mandatory rules of large scale networking: The participating simulators must be independent. Each must be able to join the game or withdraw without interfering with the operation of the others. The failure of any single simulator must not disrupt the game. But the SIMNET protocol also involved design decisions tailored to the low processing power of the SIMNET devices. Some of these design details were not desirable in general. The lessons of the long-haul SIMNET protocol were lost and had to be relearned. The technical challenges of long-haul networking are mostly two: bandwidth and transmission delays. These issues exist in local networking, but long distances between networked simulators render both issues more critical. When a large number of simulators interact, current state information about each vehicle must be broadcast for the beneﬁt of all. Broadcasting all this information at the rate at which it is created—typically 40 to 60 times a second— creates prohibitively large information ﬂows. Methods for reducing the required bandwidth were needed. One method, introduced in SIMNET, is called dead reckoning. This term, borrowed from navigation, refers to the extrapolation of a vehicle’s motion based on its previously known state. The SIMNET dead reckoning scheme has each simulator withhold its broadcasts so long as its state information can be reproduced with acceptable accuracy by extrapolation. The originating simulator (the sender) determines whether this is the case by simulating the extrapolation process of the remote simulator (the receiver). For each simulation frame, the result of the extrapolation is compared to the state of the vehicle computed for that frame. No broadcasts are made until the difference exceeds a preselected threshold. Other methods for relieving the bandwidth bottleneck include (a) bundling of packets at each node and (b) long-haul transmission of changed information only. The second technical issue is delay. Remote information is outdated information. A delay corresponding to the speed of light is a hard minimum imposed by the laws of nature. It

323

amounts to 3.33 애s/km. Over global distances of several thousand kilometers, the delay is comparable to a simulation frame. The delay in actual communications lines is roughly double the above. With a satellite link, the round trip to geostationary altitude imposes a delay of 200 ms, and the mechanics of the equipment on the satellite increases this to half a second or more. Further delays are caused by processing packets by servers at network nodes. An aircraft traveling at 400 knots covers 1 m in about 5 ms. A rotorcraft ﬂying at, say, 100 knots takes 20 ms to cover 1 m. Position discrepancies due to communications delays are visible in close formation ﬂying. Hit-or-miss decisions for projectiles are affected. Delays in communications channels are not predictable and not repeated precisely. A constant delay will make the remotely simulated vehicle appear to lag behind, whereas a variable delay will make it appear to jump around. To compensate for the delay, remote data must be extrapolated to the current time over the delay period ⌬t. Initially, there was the misconception that, so long as sender and receiver used the same dead reckoning scheme, the receiver error would never exceed the threshold imposed by the sender. The fallacy of this view was soon exposed (10). The sender withholds its broadcasts until after the threshold has been exceeded. At that time, the sender broadcasts an update. But the update does not reach the receiver until ⌬t later. All this time, the receiver’s error continues to grow. Even when the update arrives, the receiver is not at liberty to exploit it. Immediate reversion to the more recent data would cause a visible jump in the image. This would make the image jitter and betray that it is the image of a remotely simulated entity. The receiver must implement smoothing. Depending on the particular smoothing algorithm, the receiver will maintain the state error longer or even continue to grow it for a while after the update is received. This way, the receiver’s error always exceeds the sender’s threshold, and, in long-haul networking, by a very signiﬁcant margin (11). Dead reckoning, which, for the sender, is a bandwidth saving device, becomes a mandatory accuracy maintenance procedure for the receiver. Needless to say that dead reckoning by the sender increases the delay and so does any bandwidth saving scheme that requires processing at the nodes. The receiver must extrapolate the state in each packet over the delay that the packet experienced. To make this possible, it is necessary to include a timestamp with the variables of state in each data packet. The stamp is the time for which the variables are valid as opposed to the time at which they were computed or transmitted. The receiver subtracts the timestamp from the time at which the variables are to be displayed and extrapolates over the difference. The error in the dead reckoned state depends on the accuracy of the timestamp as well as on the extrapolation algorithm (10). The DIS protocol speciﬁed a timestamp since the 1990 draft. Two versions of a timestamp were recognized: an absolute timestamp produced by a clock synchronized to universal time coordinates (UTC) and a relative timestamp produced by a free running local clock. The relative timestamp can be used to correct for the jumping around effect of variable delay, but not for the lagging behind that the delay itself causes. To produce an absolute timestamp, clocks at remotely located simulation facilities must be synchronized to within a

324

AI LANGUAGES AND PROCESSING

Table 1. Communications Requirements Issue

Normal Requirements

Acknowledgments Transmit queue protocol

Required Deliver packets in order queued

Receive queue protocol

Process packets in order received

Receive buffer full Transmit buffer full Checksum Corrupted packet Lost packet

Halt transmission Halt process Required Ask for retransmission Ask for retransmission

Simulation Requirements Useless Deliver most recent packet and discard others Process most recent packet and discard others Impossible Impossible Required Discard Forget

millisecond or a few milliseconds. This has been made easy by the Global Positioning System (GPS). GPS time, accurate to better than a microsecond, is available as a biproduct of the position calculation. (GPS time differs from UTC time by a small integral number of seconds accumulated in leap years.) It has been shown that the state error due to delay and to clock error are additive (10). With GPS, the clock error can be effectively eliminated. Networked simulation in the service of the military has mastered the bandwidth issue and has achieved reliable highvolume operation over its own dedicated Defense Simulation Internet (DSI). Synchronization of simulation clocks is still not prevalent. Facing up to the challenge of specifying, and verifying precision and consistency for long-haul simulation, at this writing, is still pending. COMPUTER COMMUNICATIONS Virtual simulation requires interprocess communications, be it the local or long-haul networking of simulators or the communications between different processes within one simulator (Fig. 1). Table 1 lists the requirements of asynchronous communications in the service of virtual simulation. These requirements are different from the ones prevailing in other ﬁelds. They offer an incentive for simulation speciﬁc communications protocols. BIBLIOGRAPHY 1. A. Katz, Computational Rigid Vehicle Dynamics, Malabar: Krieger, 1997. 2. Federal Aviation Administration, Airplane simulator qualiﬁcation, Advisory Circular, 120-40B: 4–5, July 29, 1991. 3. Federal Aviation Administration, Helicopter simulator qualiﬁcation, Advisory Circular, 120-63: 2–3, October 11, 1994. 4. J. Neider, T. Davis, and M. Woo, OpenGL Programming Guide, Reading, MA: Addison-Wesley, 1993. 5. M. E. Mortenson, Geometric Transformations, New York: Industrial Press, 1995. 6. J. Schroeder and W. Chung, Effects of roll and lateral ﬂight simulation motion gains on a side step task, in Proceedings of the 53rd Annual Forum of the American Helicopter Society, Virginia Beach, VA, June 1997, pp. 1007–1015.

7. A. R. Rope, The SIMNET Network and Protocols, Report No. 7102 by BBN Systems and Technologies prepared for the Defense Advanced Research Projects Agency (DARPA), July 1989. 8. J. Cadiz, B. Goldiez, and J. Thompson, Summary Report—The First Conference on Standards for the Interoperability of Defense Simulations, Institute for Simulation and Training of the University of Central Florida Report IST-CF-89-1 (Contract No. N61339-89-C-0043) 1989. 9. IEEE Standard for Information Technology, Protocols for Distributed Interactive Simulation Applications, IEEE Std 1278-1993, New York: IEEE, May 1993. 10. A. Katz, Event correlation for networked simulators, J. Aircraft, 32 (3): 515–519, 1995. 11. A. Katz, M. Sharma, and D. E. Wahrenberger, Advanced Dead Reckoning and Smoothing Algorithms, Prepared for US Army STRICOM, Contract No. N61339-91-D-0001, Architecture & Standards, Delivery Order No. 0035, CDRL A030, May 25, 1996. Reading List G. Burdea and P. Coiffet, Virtual Reality Technology, New York: Wiley, 1994. W. E. Larsen, R. J. Randle, and L. N. Popish (eds.), Vertical Flight Training, NASA Reference Publication 1373, 1996. D. F. McAllister, (ed.), Stereo Computer Graphics and other True 3D Technologies, Princeton, NJ: Princeton Univ. Press, 1993. J. M. Rolfe and K. J. Staples, Flight Simulation, Cambridge: Cambridge Univ. Press, 1986, 1990. R. G. Stanton, Numerical Methods for Science and Engineering, Englewood Cliffs, NJ: Prentice-Hall, 1961.

AMNON KATZ University of Alabama

AGE TESTING. See INSULATION AGING TESTING. AGGREGATE COMPUTATION. See STATISTICAL DATABASES.

AGING OF INSULATION. See INSULATION AGING MODELS.

374

AIR TRAFFIC

Air taxi 20%

Military 10%

General aviation 16%

Air carrier 54%

Figure 1. Types of air traffic operations controlled at ARTCC Centers in 1996. Total number of operations was 40.7 million.

AIR TRAFFIC In today’s world, air travel is a primary mode of transportation. During 1996, nearly 575 million passengers boarded scheduled air carrier flights in the United States. Over the

next 10 years, air carrier traffic is expected to increase by more than 50%. Air travel also includes air taxi, general aviation, and military traffic. Safety is an issue when airspace capacity becomes saturated. To guarantee a safe environment, the Federal Aviation Administration (FAA) operates the Air Traffic Control (ATC) system to coordinate and control air traffic. To prepare for the future traffic increase, the FAA is redesigning the National Airspace System (NAS). Air carrier operations are those scheduled flights that carry passengers or cargo for a fee. More than 8 million air carrier departures were recorded in 1996. Air carrier operations include the major airlines, national carriers, smaller regional (or commuter) airlines, and cargo carriers. Airlines with annual revenues of $1 billion or more in scheduled service are called majors. There were nine major U.S. passenger airlines in 1994: America West, American, Continental, Delta, Northwest, Southwest, TWA, United, and US Airways. Two cargo carriers were classified as majors: Federal Express and United Parcel Service (1). National carriers are scheduled airlines with annual revenues between $100 million and $1 billion. National carriers often serve particular regions of the nation. Like the majors, nationals operate mostly medium-sized and large jets. In the third category are regional carriers that serve a single region of the country, transporting travelers between the major cities of their region and smaller, surrounding communities. Regional carriers are one of the fastest growing and most profitable segments of the industry. Large and medium-sized regional carriers often use aircraft that seat more than 60 passengers. Small regional, or commuter, airlines represent the biggest segment of the regional airline business. Most of the aircraft used by small regionals have less than 30 seats (1). Air taxi operations include those operations involved with charters, air ambulance service, air tours, and other unscheduled air service. General aviation operations are those flights serving individuals and organizations using self-owned or leased aircraft. Pleasure, business, and flight training trips are typical general aviation activities. Military air traffic from all defense organizations also uses the civilian air traffic control system. The intermixture of civilian and military traffic needs to be controlled by a single system to ensure safe operations. The percentage of air carrier, air taxi, general aviation, and military aircraft operating in U.S. airspace in 1996 is shown in Fig. 1. The statistics in Fig. 1 reflect the operations that were controlled by the FAA’s Air Route Traffic Control Centers (ARTCC), or Centers. The total number of Center operations was 40.7 million in 1996. The largest percentage of Center traffic is from air carrier operations. Air carriers domi-

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

AIR TRAFFIC

nate Center traffic because each flight has an instrument flight plan filed with air traffic control for the possibility of encountering instrument meteorological conditions (IMC). An instrument flight plan requires interaction with the air traffic control system. The general aviation percentage may seem small considering the number of general aviation aircraft. General aviation aircraft are not required to communicate with ATC provided that they maintain visual meteorological conditions (VMC) and avoid controlled airspace. Visual flight rules (VFR) are used for flight in VMC that requires vertical separation from clouds and a minimum visibility (2). AIR TRAFFIC CONTROL To handle the volume of air traffic, the FAA has established the Air Traffic Control system. The ATC system includes air traffic control towers (ATCT), terminal radar approach control (TRACON), Air Route Traffic Control Center (ARTCC), and flight service stations (FSS). The tower controls the airspace around an airport. The airspace typically extends 5 statute miles horizontally and 3000 feet vertically above ground level (AGL) (10). Aircraft operating in this area must contact the tower even if the aircraft is passing through this airspace or is landing at a ‘‘satellite’’ airport that lies in the area. There are 352 FAA-controlled towers in the United States. In 1995, there were more than 26 million operations at the 100 busiest airports in the United States. The 10 busiest airports (1995) are shown in Table 1. A TRACON is the radar facility that handles arrivals and departures in a high-traffic area. A TRACON usually controls aircraft within a 30- to 40-mile radius of the principal airport in the area. Controllers in the TRACON help arriving aircraft transition from en route to the airport tower. Aircraft arriving from all quadrants must be funneled to the active runway’s final approach fix before the TRACON hands the aircraft to the tower controller. Minimum separation between aircraft must be maintained to ensure safety from wake vortex turbulence. The TRACON also handles the transition from Tower to Center for departing aircraft. The ARTCC or Center handles aircraft during their en route phase between departure and arrival airports. Aircraft typically fly along predefined airways, or highways in the sky. Each airway is defined as a path between navigation aids on the ground. Under the current air traffic system, aircraft are frequently restricted to ATC-preferred routes, which are not necessarily the routes preferred by the pilot or airline. Air

Table 1. Ten Busiest US Airports Based on 1995 Statistics Rank

City–Airport

1995 Enplanements

1995 Operations

1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

Chicago O’Hare Atlanta Hartsfield Dallas–Fort Worth Los Angeles San Francisco Miami Denver New York JFK Detroit Metropolitan Phoenix Sky Harbor

31,255,738 27,350,320 26,612,579 25,851,031 16,700,975 16,242,081 14,818,822 14,782,367 13,810,517 13,472,480

892,330 747,105 873,510 716,293 436,907 576,609 487,225 345,263 498,887 522,634

375

traffic controllers instruct pilots when to change their direction, speed, or altitude to avoid storms or to maintain traffic separation. Not all aircraft follow the airway system. Depending on traffic load, weather, and aircraft equipment, it is possible for the controller to clear the aircraft on a direct route. Of the 20 ARTCCs in the continental United States, the five busiest in 1995 were Chicago, Cleveland, Atlanta, Washington, and Indianapolis (3). The FAA’s Air Traffic Control System Command Center (ATCSCC) is responsible for managing traffic flow across the United States. The ATCSCC is located in Herndon, Virginia. The command center oversees the entire ATC system and provides flow information to the other ATC components. If an area is expecting delays due to weather or airport construction, the command center issues instructions to reduce traffic congestion by slowing or holding other traffic arriving at the trouble area. The FAA operates numerous navigation aids (NAVAID) to assist aircraft operations. En route navigation primarily uses the VORTAC or VOR/DME system. A VOR/DME system consists of a network of VOR/DME radio navigation stations on the ground that provide bearing and distance information. An aircraft must have the proper radio equipment to receive the signals from these systems. Civilian traffic obtains bearings from the VOR (very high frequency, or VHF, omnidirectional range) component and distance from the DME (Distance Measuring Equipment). Military traffic uses the TAC or TACAN (Tactical Airborne Navigation) signal. The VOR/DME system is the NAVAID that defines the airways. Instrument approaches to an airport runway require electronic guidance signals generated by transmitters located near the runway. Precision approaches use the Instrument Landing System (ILS). The ILS provides horizontal (localizer) and vertical guidance (glideslope). A Category I ILS approach typically allows an aircraft to descend to 200 feet AGL without seeing the runway environment. Continued descent requires that the runway environment be in view. Each airport runway with a precision approach typically requires dedicated ILS equipment installed and certified for that runway. Nonprecision approaches are commonly defined using VOR/ DMEs, nondirectional beacons (NDB), and localizers. A nonprecision approach does not provide glide slope guidance and, therefore, limits the minimum altitude allowed without visual contact with the runway.

AIRSPACE CAPACITY The number of aircraft operations, both civilian and military, continues to grow, which strains the capacity of the airspace system. Over the period 1980 to 1992, traffic in the United States grew at an average annual rate that was 0.4 percentage point faster than the increase in capacity (3). By 2005, the number of air carrier passengers is expected to grow from 550 million (1995) to 800 million. During the same period, the number of air carrier domestic departures is expected to grow from 7.6 million to 8.9 million. Today’s restricted airspace system will not be able to accommodate the rapid growth in aviation (3). Delay in air carrier operations is one method of measuring system capacity. From 1991 to 1995, the number of air carrier operations increased more than 18% while the number of air

376

AIR TRAFFIC

Minutes 7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 Gate-hold

Taxi-out

Airborne Taxi-in

Figure 2. The average delay per flight phase (in minutes) during an air carrier’s scheduled revenue flight.

carrier operations delayed 15 min or more fell from 298,000 to 237,000. The average delay per flight held steady at 7.1 min during this period (3). Figure 2 highlights taxi-out as the flight phase with the largest average delay. Taxi-out, the time from push-back at the gate until takeoff, is susceptible to delay from airport surface traffic. Aircraft that are taxiing in are expedited to make room for more arrivals and other surface traffic. During a departure push, many aircraft are departing the airport at approximately the same time. Aircraft taxiing out are coming from numerous gates scattered across the airport and chan-

neled to one or two active departure runways. The departing aircraft will often form long lines as they inch toward the runway. For airport operations using the same runway for arrivals and departures, the departing aircraft must wait for an arrival gap before entering the runway and taking off. When a runway is dedicated for departures, aircraft separation requirements slow the departure process (3). To reinforce the effects of flight delay, consider its economic impact. Heavy aircraft of 300,000 lb or more cost $4575 per hour of delay; large aircraft less than 300,000 lb and small jets cost $1607 per hour. Single-engine and twin-engine aircraft under 12,500 lb cost $42 and $124 per hour, respectively. With approximately 6.2 million air carrier flights in 1995 and an average airborne delay of 4.1 min per aircraft, 424,000 hours of airborne delay occurred that year. At the average operating cost of approximately $1600 (1987 dollars) per hour, the delay cost the airlines $678 million (3). Poor weather was attributed as the primary cause of 72% of operations delayed by 15 min or more in 1995. Weatherrelated delays are largely the result of instrument approach procedures, which are much more restrictive than the visual procedures used during better weather conditions (3). Figure 3 shows that weather followed by airport terminal congestion were the leading causes of delay from 1991 to 1995. Closed runways/taxiways and ATC equipment, the third and fourth largest causes, had smaller effects on annual delay. Delays will become worse as air traffic levels climb. The number of airports in the United States, where cumulative annual delays are expected to exceed 20,000 hours per year, is predicted to increase from 23 in 1991 to at least 33 by the year 2002 (4).

;; ; ;; ; ; ;;;

;; ;; y;y;;; ;; ;;;; y;;;; ;; y;;; ;yy;;y; ;; 300

Other NAS equipment Closed runaways/ taxiways Terminal volume Weather

250

(Thousands)

200

150

100

50

Figure 3. The number of delayed air carrier flights (in thousands) for the period 1991 to 1995. The reasons for the delay are shown.

0

1991

1992

1993

1994

1995

AIR TRAFFIC

INCREASED AIR TRAFFIC LEVELS The FAA, air carriers, and general aviation organizations are all forecasting increased air traffic for the coming decades. The FAA predicts that by 2007, operations from all air traffic, including air carriers, regionals, air taxi, general aviation, and military aircraft, are expected to increase to 74.5 million (a 19% increase over 1995). The number of passenger enplanements on international and domestic flights, both air carrier and regional/ commuter, is expected to grow to 953.6 million by 2007 (a 59% increase over 1995). The growth rate of enplanements exceeds the growth rate of operations due to the use of larger aircraft and a higher occupancy rate on each flight (3). The FAA numbers count all activity at a U.S. airport regardless of whether the air carrier is U.S flagged or international. Figure 4 shows similar numbers for U.S. air carriers as forecast by the Air Transport Association (1). The forecast for the next decade projects that the busiest airport facilities are going to become busier. The top 100 airports in 1991 had 408.8 million revenue passengers depart, which accounted for over 94% of all passengers in the United States. From 1991 to 1995, the number of air carrier and regional/commuter enplanements increased by 32.9% (from 408.8 million to 543.4 million). By 2010, passenger boardings at the top 100 airports will increase by 69.1% (to 919.1 million) and aircraft operations are projected to increase by 27.6% (to 33.7 million) (3). The 10 busiest airports in 2010 based on operations and their percentage growth from 1995 are shown in Table 2. A comparable ranking of the 10 busiest airports as a function of passenger departures is shown in Table 3. Chicago O’Hare, Dallas–Fort Worth, Atlanta Hartsfield, and Los Angeles International are forecast to be the busiest airports by 2010 in both operations and passenger enplanements. While the air transportation industry in the United States continues to grow, it is important to remember that North America traditionally represents only about 40% of the

900 800

377

Table 2. Forecast Departures at the 10 Busiest US Airports Rank

City–Airport

1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

Chicago O’Hare Dallas–Fort Worth Atlanta Hartsfield Los Angeles Miami Phoenix Sky Harbor St. Louis Lambert Las Vegas McCarran Oakland Metropolitan Detroit Metropolitan

Total for top 100 airports

1995 Operations

2010 Operations

% Growth

892,330 873,510 747,105 716,293 576,609 522,634 516,021 508,077 502,952 498,887

1,168,000 1,221,000 1,056,000 987,000 930,000 736,000 645,000 682,000 573,000 675,000

30.9 39.8 41.3 37.8 61.3 40.8 25.0 34.2 13.9 35.3

26,407,065

33,706,000

27.6

world’s total air traffic (4). In the next decade, international air travel is expected to continue its significant increase. Passenger traffic on world air carriers has shown an annual growth rate of 5.0% over the last decade. Forecasts for the coming decade are predicting that the growth rate will increase slightly to 5.5% annually. The number of passenger enplanements worldwide would grow from 1285 million in 1995 to 2010 million in 2005 (56% growth). The fastest growing international route groups for passenger traffic are forecast to be in Transpacific and Europe–Asia/Pacific route groups (5). By the year 2010, the International Air Transport Association (IATA) predicts that the number of international passengers traveling to and from the United States will reach 226 million, an increase of 187% over the 1993 figure of 78.8 million (4). The majority of these are expected to travel on U.S. carriers. AIRCRAFT FLEET To handle the swelling number of air travelers, the air carrier fleets need to be upgraded with larger aircraft. Most of the growth in fleet size of the major U.S. carriers will occur after 2000, when aging aircraft are replaced with newer, more efficient aircraft. The fleet size, with its upswing after 2000, is shown in Fig. 5 (1).

700

Millions

600 500 400 300 200 100 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

0

Figure 4. The number of revenue passengers on US air carriers grew from 382 million in 1985 to 547 million in 1995. The growth is forecast to climb to 857 million revenue passengers by 2007.

Table 3. Forecast Passenger Enplanements at the 10 Busiest US Airports Rank 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

City–Airport Chicago O’Hare Atlanta Hartsfield Dallas–Fort Worth Los Angeles San Francisco Miami Denver New York JFK Detroit Metropolitan Phoenix Sky Harbor

Total for top 100 airports

1995 2010 % Enplanements Enplanements Growth 31,255,738 27,350,320 26,612,579 25,851,031 16,700,975 16,242,081 14,818,822 14,782,367 13,810,517 13,472,480

50,133,000 46,416,000 46,553,000 45,189,000 28,791,000 34,932,000 22,751,000 21,139,000 24,220,000 25,408,000

60.4 69.7 74.9 74.8 72.4 115.1 53.5 43.0 75.4 88.6

543,439,185

919,145,000

69.1

378

AIR TRAFFIC

7000

6564

6000

5764

5972

6174

6377

5547 5306 5000

4720

4784

4920

1996

1997

1998

5036

5027

1999

2000

4000

3000

2000

1000

Figure 5. Jet aircraft forecast to be in service by US air carriers.

0

At the end of 1995, U.S. air carriers had firm orders placed for 604 new aircraft and options on an additional 799 aircraft. The price tag for the firm orders was $35.5 billion. The firm orders were distributed among aircraft from Airbus Industries, Boeing Commercial Aircraft Company, McDonnellDouglas Aircraft Company, and the Canadian Regional Jet. The most popular aircraft on order was the Boeing 737, with 218 firm orders and 260 options. FREE FLIGHT In April 1995, the FAA asked RTCA, Inc., an independent aviation advisory group, to develop a plan for air traffic management called Free Flight (6). Free Flight hopes to extend airspace capacity by providing traffic flow management to aircraft during their en route phase. By October 1995, RTCA had defined Free Flight and outlined a plan for its implementation (7). The Free Flight system requires changes in the current method of air traffic control. Today, controllers provide positive control to aircraft in controlled airspace. Free Flight will allow air carrier crews and dispatchers to choose a route of flight that is optimum in terms of time and economy. Economic savings will be beneficial both to the air carriers and to the passengers. Collaboration between flight crews and air traffic managers will be encouraged to provide flight planning that is beneficial to the aircraft and to the NAS. User flexibility may be reduced to avoid poor weather along the route, to avoid special-use airspace, or to ensure safety as aircraft enter a high-density traffic area such as airports. The new system will offer the user fewer delays from congestion and greater flexibility in route determination (3). Flights transitioning the airspace in Free Flight will have two zones surrounding the aircraft. A protected and an alert zone are used to provide safety for the flight. The size and shape of the zones depend on the size and speed of the aircraft. The goal is that the protected (or inner) zones of two

2001

2002

2003

2004

2005

2006

2007

aircraft will never touch. The aircraft may maneuver freely as long as its alert zone does not come in contact with another aircraft’s alert zone. When a conflict occurs between two aircraft alert zones, changes in speed, direction, or altitude must be made to resolve the conflict. The conflict resolution may be made by the air traffic manager or from the airborne collision avoidance system, TCAS (Traffic Alert and Collision Avoidance System). The FAA and airspace users must invest in new technology to implement Free Flight. New communication, navigation, and surveillance systems are required to maintain situational awareness for both the air traffic manager and the flight crew. The FAA and aviation community are working together to phase in Free Flight over the next 10 years (8). NATIONAL AIRSPACE SYSTEM In parallel with RTCA’s development of the Free Flight concept, the FAA began to develop a new architecture for the NAS that would support future aviation needs and Free Flight. The new NAS architecture transitions from air traffic control to air traffic management. The new NAS architecture is focused on the implementation of Free Flight to handle aircraft traffic and routing. The FAA’s Pilot/Controller Glossary defines the NAS as ‘‘the common network of U.S. airspace; air navigation facilities, equipment and services; airports or landing areas; aeronautical charts, information and services; rules, regulations and procedures; technical information; and manpower and material. Included are system components shared jointly with the military’’ (9). The new NAS architecture and Free Flight require a change from the old system of air traffic control to a new air traffic management system. The air traffic managers will now be responsible for monitoring/managing an aircraft’s flight along its route. In the new system, an aircraft’s route will primarily be the responsibility of the aircraft crew instead of ATC. The air traffic manager will need to intervene only to

AIR TRAFFIC

provide conflict resolution and route planning in high-density traffic areas. The new NAS architecture recommends new communications, navigation and landing, surveillance, and weather systems for the next 20 years. Major NAS plans in the communications, navigation, and surveillance (CNS) are as follows (9): • Use satellite-based (Global Positioning System, or GPS) landing and navigation systems while decommissioning ground-based facilities. • Use automatic GPS-derived position reports as air traffic management’s surveillance system. • Use digital air/ground communications instead of today’s analog radios. Major changes in the ATC decision support systems include the following: • A common air traffic management platform • A new conflict detection/resolution and collaborative decision-making tool • A new integrated display/ controller workstation for ATC towers A new NAS Infrastructure Management System (NIMS) provides centralized monitoring and control to NAS facilities from operational control centers (OCC) (9). The NAS architecture defines changes to aircraft avionics, ground-based ATC equipment, ground-based navigation and landing aids, and the airport environment. A summary of the changes in each area is provided. Airborne Equipment Implementation of the NAS requires several new avionics advancements. The avionics systems that are being defined for NAS are in communications, navigation, surveillance (CNS), and cockpit displays. Global Positioning System. The Global Positioning System (GPS) is proposed as the primary navigation aid in the new NAS. GPS uses Department of Defense (DoD) satellites to derive the present position and velocity of the user vehicle. A GPS receiver has a position accuracy within 100 m, 95% of the time. This accuracy is sufficient for en route and oceanic navigation. To navigate in the airport terminal area or to perform an approach in instrument weather, the aircraft needs the increased accuracy of a differential GPS (DGPS) system. A stand-alone GPS has position error due to clock error, atmospheric effects, and DoD-induced noise. Differential GPS can effectively remove these errors by adding a differential correction signal. The NAS defines two types of differential GPS systems: wide-area augmentation system (WAAS) and localarea augmentation system (LAAS) (9). The WAAS uses a network of GPS base stations to determine the GPS errors. The differential correction signal is uplinked to geostationary WAAS satellites. The aircraft receives the correction signal from the WAAS satellite and corrects its GPS position. Position accuracy with WAAS DGPS is within 7.6 m 95% of the time. The WAAS DGPS will provide

379

sufficient accuracy for an aircraft to make a Category I instrument approach (200 ft ceiling/1800 ft visibility) (9). The LAAS is dedicated to a single airport or airports in the same area. A GPS base station is located at or near the airport. The differential correction signal is broadcast to all aircraft within a 30 mile region using an RF datalink. The LAAS is more accurate than the WAAS since the aircraft are in closer proximity to the base station providing the corrections. The LAAS DGPS can be used for Category II approaches (100 ft ceiling/1200 ft visibility) and Category III approaches (0 ft ceiling). Category III has three subcategories (A, B, C) with visibility minimums of 700 ft, 150 ft, and 0 ft, respectively (10). The LAAS DGPS is also useful for ground navigation. Accurate positioning on the airport surface increases the pilot’s situational awareness during taxi operations. It also provides ATC with an accurate knowledge of all ground traffic. Automatic Dependent Surveillance-Broadcast. An ATC radar screen displays aircraft position using the airport surveillance radar (ASR) and the secondary surveillance radar (SSR). The ASR transmits a radar signal that reflects from the aircraft skin. The SSR interrogates an aircraft’s transponder, which returns the aircraft’s transponder code and its altitude. Aircraft equipped with a newer Mode-S transponder can return additional data, such as heading and velocity. The proposed NAS architecture phases out the SSR system. It will be replaced with Automatic Dependent Surveillance-Broadcast (ADS-B). At approximately twice per second, the aircraft’s on-board ADS-B broadcasts aircraft position (latitude/longitude/altitude) and status information using the Mode-S transponder. The ADS-B periodically broadcasts the flight identification and the aircraft’s ICAO (International Civil Aviation Organization) address. For air carriers, the flight identification is the flight number (for example, NW132) that passengers, pilots, and controllers use to identify a particular flight. The ICAO address is a unique number that is assigned to an aircraft when it is manufactured. ADS-B provides controllers with the accurate aircraft identification and position needed to implement Free Flight. ADSB also provides information to the ground controller during airport surface operations. Positive identification and accurate position (using LAAS DGPS) during taxi-in and taxi-out operations are especially important for safe and timely operations in low-visibility conditions (11). Traffic Alert and Collision Avoidance System (TCAS). TCAS is an airborne surveillance system that monitors nearby aircraft and detects impending collisions. The position and altitude of nearby traffic are shown on a cockpit display. TCAS transmits transponder interrogation signals similar to the SSR groundbased system. Aircraft receiving the signal respond with a normal transponder reply that includes altitude. The TCAS can determine the bearing to the aircraft using a multielement antenna. TCAS protects a safety zone around the aircraft. A track is started for every traffic target detected by TCAS. The collision avoidance logic calculates the time to a possible conflict with each of the traffic targets. If the time to a collision or nearmiss counts down to 45 s, a traffic advisory is generated informing the pilot of the situation. If the time gets to 25 s, a resolution advisory is issued. A resolution advisory commands

380

AIR TRAFFIC

the pilot to climb, descend, or maintain the current altitude to avoid a collision. When both aircraft are TCAS equipped, the TCAS systems communicate the resolution advisory to prevent both aircraft from taking the same avoidance maneuver (12). Datalink Communications. Communication frequencies at airports are congested with radio chatter. A flight crew may have to wait for an opening to make a request of ATC. There can also be confusion over which aircraft was given an ATC command. The congestion can cause delays and impact safety. Many of the voice communications are routine and could be handled by digital datalink. Digital datalink communications are routed between ATC and aircraft computer systems. The data are processed and presented to the controller or pilot as needed. Controller-pilot data link communications (CPDLC) use a two-way datalink. Controller commands and instructions are relayed to the aircraft using an addressed datalink. Only the intended aircraft receives the instruction. The ATC command is processed on the aircraft and presented to the flight crew. The flight crew performs the command. Acknowledgment back to the controller can either be pilot initiated or automatically generated when the crew complies with the instruction (13). Flight Information Service (FIS) data can also be accessed via datalink. FIS contains aeronautical information that the pilot uses in flight planning. Without the FIS datalink, the pilot must access the information by request over a voice channel. Datalinks are to be used for navigation and surveillance data as well as communications. A one-way datalink is used for ADS-B, DGPS, and Terminal Information Service (TIS). An aircraft broadcasts its ADS-B position using a one-way datalink. Other aircraft and ATC can receive the ADS-B report and track the aircraft position. A one-way broadcast uplinks the LAAS DGPS differential corrections to aircraft in the airport area. The TIS system is a one-way broadcast of airport traffic to aircraft on the ground. Weather data can be transmitted between the aircraft and ground across a datalink (9). Many of the datalink services will initially use the Mode-S transponder datalink. With the development of VHF data radios and satellite service, the NAS architecture may change the primary datalink provider for these services. Cockpit Displays. Cockpit displays are used in the NAS architecture to display air traffic management information that is transferred to the aircraft. Moving map navigation displays using GPS position will assist the pilot both in the air and on the ground. While airborne, the navigation display can display the desired route, possible weather, and other traffic. Suggested route changes from air traffic management to avoid congestion or special-use areas can be displayed on the moving map. The pilot can negotiate the route change with ATC. Terrain data can be incorporated into the moving map to ensure safe clearance of all terrain features. The moving map display is very beneficial during airport surface operations. At night or in low-visibility conditions, the airport surface is very confusing to aircrews that are not familiar with the airport. A joint NASA/FAA experiment has shown that a taximap displaying airport layout, taxi routes, and other traffic can improve safety and efficiency. The taxi-

map display can also reduce runway incursions that occur when a taxiing aircraft enters an active runway with an arrival or departure in progress (11). ATC commands transmitted via the CPDLC datalink can be shown both graphically on the moving taximap and textually for clarification. Aircrew acknowledgments can be displayed as well. Ground-Based NAS Equipment The ground-based equipment needed for the NAS architecture involves improvements and development at all NAS facilities. Traffic flow management and air traffic control tools are improved at the ARTCCs (Centers), the TRACONs (Approach Control), and the ATCT (Towers). En Route ARTCC Equipment. The new NAS architecture upgrades the existing ARTCC Center equipment. New display systems, surveillance data processors (SDP), and flight data processors (FDP) are improvements to existing systems. The SDP system will collect information from the surveillance systems such as the ADS-B reports. The SDP will provide aircraft tracking and conflict detection/resolution. The FDP will correlate aircraft tracks with flight plan information. The FDP will also communicate with other ARTCC Centers and terminals to ensure that all air traffic management units have the same flight plan information for an aircraft (9). Air Traffic Management Decision Support Services. Air traffic management (ATM) combines the ATC and traffic flow management (TFM) functions. ATM support tools are called decision support services. The TFM decision support services function includes the collaborative decision-making tool that aids the pilot/controller interaction in flight planning (9). The decision support services for the ATC function involves conflict detection/resolution and Center/TRACON Automation System (CTAS). The Center/TRACON Automation System is a tool developed by NASA to support air traffic management. CTAS computes an aircraft’s route and intentions 40 min into the future. The aircraft destination, as filed in the flight plan, and aircraft type are considered in the calculations. CTAS examines the aircraft mix that is arriving at an airport and provides the arrival sequencing and separation for efficient operation (8). Ground Controller Equipment. Sensors and conflict detection/resolution equipment dominate enhancements to the ground controller equipment. At a large, busy airport, the number of aircraft taxiing can be significant. During arrival and departure pushes, in good and bad weather, it is difficult to manage the position of each aircraft and its intentions. Three systems that will help the ground controller manage the traffic safely and efficiently are the Airport Surface Detection Equipment (ASDE), the Airport Target Identification System (ATIDS), and the Airport Movement Area Safety System (AMASS) (9). Airport Surface Detection Equipment (ASDE) is a radar system that detects aircraft and other vehicles moving on the airport surface. The ASDE antenna is a large rotodome that typically mounts on top of the control tower. The rotodome physically spins at 60 revolutions per minute. The ASDE system ‘‘paints’’ surface traffic using the radar reflection from the

AIR TRAFFIC

381

Extended runways New runways

6 5 4 3 2

target. The ASDE system is already installed at numerous airports. A large ASDE monitor is mounted in the control tower to display traffic. One drawback with the ASDE system is that traffic appears as ‘‘blips’’ on the monitor with no flight identification tags. The ATIDS solves that problem by applying tags to the ASDE targets. ATIDS is a multilateration system that listens to the Mode-S transmissions from aircraft. By timing the arrival of the transmission at multiple sites, it is possible to determine the aircraft location through triangulation. The ATIDS system uses flight plan information to correlate the aircraft’s transponder code with the flight number (14). AMASS tracks ASDE targets and performs collision detection analysis on airport traffic. AMASS alerts the ground controller of possible conflicts. AMASS also alerts the controller to possible runway incursion incidents where a taxiing aircraft is entering an active runway incorrectly. AMASS correlates position information from the ASDE and ATIDS systems and applies the ATIDS identification tag to the targets on the ASDE display. Airport Facilities and Procedures. To increase capacity, the nation’s airports have been building new runways and extending existing runways. Extending the length of the runways can help increase capacity by making general aviation runways into air-carrier-length runways. New procedures are also being defined for parallel approaches and reduced separation standards. Adding new runways and extending existing runways adds capacity without the cost of adding new airports. By 1997, 64 of the top 100 airports had recently completed, or were in the process of developing, new runways or runway extensions to increase airport capacity. Many of these are at the busiest airports that are forecast to have more than 20,000 h of annual air carrier delay in 2005 (3). Figure 6 lists the number of new runways and runway extensions that are currently planned. There are 17 new runways and 10 runway extensions not shown on the figure because they are planned but have not been assigned a firm completion date (3). The largest capacity gains result from the construction of new airports. Considering that the new Denver International Airport, which opened in 1995, cost more than $4 billion, building new airports is not always feasible. Only one new

2010+

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

0

1996

1 Figure 6. The number of new runways and runway extensions being planned for US airports.

airport was under construction in 1997. The new airport is being created from the conversion of Bergstrom Air Force Base in Austin, Texas to a civilian facility. The closed military base was to be open for passenger service by 1999. The new facility will add capacity to the system at a reduced cost from building a new airport (3). Terminal area capacity can be increased by redesigning terminal and en route airspace. Relocating arrival fixes, creating new arrival and departure routes, modifying ARTCC traffic flows, and redefining TRACON boundaries can all increase capacity. Improvements to en route airspace must be coordinated with terminal area improvements to avoid a decrease in terminal capacity. If the en route structure is improved to deliver more aircraft to the terminal area, then additional delays would decrease the terminal capacity (3). Instrument Approach Procedures. Instrument approach procedures can improve capacity by reducing the separation standards for independent (simultaneous) instrument approaches to dual and triple parallel runways. Landing and hold short operations for intersecting runways and simultaneous approaches to converging runways can also increase capacity. Simultaneous instrument approaches to dual parallel runways are authorized when the minimum spacing between runways is 4300 ft. The spacing minimum has been reduced to 3000 ft when the airport has a parallel runway monitor, one localizer is offset by 2.5⬚, and the radar has a 1.0 s update. Airport capacity is expected to increase by 15 to 17 arrivals per hour (3). Simultaneous arrivals to three parallel runways are also authorized. Spacing requirements state that two runways are a minimum of 4000 ft apart. The third runway must be separated by a minimum 5300 ft. Radar with a 1.0 s update rate must also be used (3). Land and hold short operations (LAHSO) allow simultaneous arrivals to intersecting runways. Land and hold short operations require that arriving aircraft land and then must hold short of the intersecting runway. Current regulations define land and hold short operations only for dry runways. Special criteria for wet operations are being developed and should be implemented by early 1997. During tests at Chicago O’Hare, a 25% increase in capacity was achieved during wet

382

AIR TRAFFIC CONTROL

operations using land and hold short operations on intersecting runways (3). Simultaneous approaches can be performed on runways that are not parallel provided that VFR conditions exist. VFR conditions require a minimum ceiling of 1000 ft and minimum visibility of 3 miles. The VFR requirement decreases runway capacity in IFR (Instrument Flight Rules) conditions and causes weather-related delays. Simultaneous instrument approaches to converging runways are being studied. A minimum ceiling of 650 ft is required. The largest safety issue is the occurrence of a missed approach (go-around) by both aircraft. An increase in system capacity of 30 arrivals per hour is expected (3). Reduced Separation Standards. A large factor in airport capacity is separation distance between two aircraft. The main factor in aircraft separation is generation of wake vortexes. Wake vortexes are like horizontal tornadoes created from an aircraft wing as it generates lift. Wake vortex separation standards are based on the class of the leading and trailing aircraft. Small aircraft must keep a 4 nautical mile (nm) separation when trailing behind large aircraft. If the lead aircraft is a Boeing 757, then a small aircraft must trail by 5 nm. Large aircraft only need to trail other large aircraft by 3 nm. The FAA and NASA are studying methods of reducing the wake vortex separation standards to increase capacity. Any reduction in the spacing standards must ensure that safety is preserved (3). EMERGING TECHNOLOGIES Several new technologies are being developed that are not specifically defined in the NAS. One technology that will increase system capacity is the roll-out and turn-off (ROTO) system. The ROTO system reduces runway occupancy time for arrivals by providing guidance cues to high-speed exits. The ROTO system with a heads-up display gives steering and braking cues to the pilot. The pilot is able to adjust braking and engine reversers to maintain a high roll-out speed while reaching the exit speed at the appropriate time. In low visibility, ROTO outlines the exit and displays a turn indicator. Present ROTO development uses steering cues to exit the runway; future systems could provide automatic steering capability (11). BIBLIOGRAPHY 1. The Airline Handbook, Air Transport Association, 1995. 2. Air Traffic, FAA Administrators Fact Book, April 30, 1997, http:// www.tc.faa.gov//ZDV/FAA/administrator/airtraffic.html 3. 1996 Airport Capacity Enhancement Plan, Federal Aviation Administration, Department of Transportation. (http://www.bts. gov/NTL/data/96_ace.pdf) 4. North American Traffic Forecasts 1980–2010: Executive Summary, International Air Transport Association (IATA), 1994 edition. (http://www.atag.org/NATF/Index.html) 5. Growth in Air Traffic To Continue: ICAO Releases Long-Term Forecasts, press release, International Civil Aviation Organization, Montreal, Canada, March 1997. 6. FAA and Aviation Community to Implement Free Flight, press release, FAA News, Washington, DC, March 15, 1996.

7. Free Flight Implementation, Final Report of RTCA Task Force 3 RTCA, Inc. October 31, 1995. 8. T. S. Perry, In search of the future of air traffic control, IEEE Spectrum, 34 (8): 18–35, 1997. 9. National Airspace System (NAS) Architecture, version 2.0, Federal Aviation Administration, Department of Transportation, October 21, 1996. 10. Federal Aviation Regulations/Airmen’s Information Manual 1998, Jeppesen-Sanderson, Inc., 1998. 11. S. Young and D. Jones, Flight testing of an airport surface movement guidance, navigation, and control system, Proc. Inst. Navigation’s Tech. Meet., Jan. 21–23, 1998. 12. M. Kayton and W. Fried, Avionics Navigation Systems, 2nd ed., New York: Wiley, 1997. 13. J. Rankin and P. Mattson, Controller interface for controller-pilot data link communications, Proc. 16th Dig. Avionics Syst. Conf., October 1997. 14. R. Castaldo, C. Evers, and A. Smith, Positive identification of aircraft on airport movement area—results of FAA trials, IEEE J. Aerosp. Electron. Systems, 11 (6): 35–40, 1996.

JAMES M. RANKIN Ohio University

382

AIR TRAFFIC CONTROL

AIR TRAFFIC CONTROL The United States air traffic management (ATM) system provides services to enable safe, orderly, and efficient aircraft operations within the airspace over the continental United States and over large portions of the Pacific and Atlantic oceans and the Gulf of Mexico. It consists of two components, namely, air traffic control (ATC) and traffic flow management (TFM). The ATC function ensures that the aircraft within the airspace are separated at all times, while the TFM function organizes the aircraft into a flow pattern to ensure their safe and efficient movement. The TFM function also includes flow control such as scheduling arrivals to and departures from the airports, imposing airborne holding due to airport capacity restrictions, and rerouting aircraft due to unavailable airspace. In order to accomplish the ATC and TFM functions, the ATM system uses the airway route structure, facilities, equipment, procedures, and personnel. The federal airway structure consists of lower-altitude victor airways and higher altitude jet routes (1). The low-altitude airways extend from 1200 ft (365.8 m) above ground level (AGL) up to, but not including, 18,000 ft (5486.4 m) above mean sea level (MSL). The jet routes begin at 18,000 ft (5486.4 m) and extend up to 45,000 ft (13,716 m) above MSL. A network of navigational aids mark the centerline of these airways, making it possible to fly on an airway by navigating from one navigational aid to the other. The airways are eight nautical miles wide. Figure 1 shows the location of the jet routes and navigation aids that are within the airspace controlled by the Oakland and Los Angeles Air Route Traffic Control Centers. The jet routes are designated by the letter J, such as J501. Navigation facilities are indicated by a three-letter designation such as PYE. Four types of facilities are used for managing traffic. They are the flight service stations (FSSs), air traffic control towers (ATCTs), terminal radar approach controls (TRACONs), and air route traffic control centers (ARTCCs) (1). These facilities J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

383

J50

1

AIR TRAFFIC CONTROL

J67

J189

J92

RBL

Oakland ARTCC

LLC

J65-126

ENI J1-3

J143

-9 J32

FMG

4

J32

J84

OAL J92

BTY

0

J8

2

6

AVE

EHF

J6

J1

J86

50

7

-65

6

J1-

-12

J5-

J88

Jet route

J9

J6 -6

RZS

J5

5 J50

J17

J8 F IM 812 6

3

LAX

ZLA

-10

J9

J9

PMD 46 0-1

PDZ

SLI

DAG

-11

0

46 0-1 -10 -107 J9 0 J6

J9

-10

0

0

J60

BLD

J7

2-8

6

J64

07 0-1 H E C J6 -107

J128 J6

J76

PGS J9

EED

J64-128

2

6 J23

36 J4-10 J2 -104 T N P J10-231 J4-1 P K 74-9 J74-96 04 E 6 J78 J4 -13 J5 J78-134 -1 4-1 J 0 1 6 9 69 J21 04 TRM 2 J5 JLI B L H -10-65 69 J2-18

J50J1

PSP

J9

3

OCN J1

MZB J1

Los Angeles ARTCC

J6

LAS

J10

J11

J7

J11

-1

J65

6 8-12 1 J50

J6

07

J50

J9

J6-8

CZQ

7

J1

J501

J189

J110

FRA

J110

SNS BSR

J58-80

J65

ECA

10

-94

MVA

J84

LIN

J5

J58-80

J7

OAK

4

2

J126

J88

VORTAC

J3

J9

2

SAC

PYE

J9

ZOA

TIJ

IPL

J2-18

BZA

Figure 1. Oakland and Los Angeles Air Route Traffic Control Center airspace.

provide service during different phases of flight. The flight service stations provide preflight and inflight weather briefings to the pilots. They also request the flight plan information which consists of the departure and arrival airports, airspeed, cruise altitude, and the route of flight, which they pass on to the ARTCCs. Flight plan filing is mandatory for flight operations under instrument flight rules. It is not required for flight operations under visual flight rules but it is highly recommended. The ATCTs interact with the pilots while the aircraft are on the ground or shortly into the flight. During a part of the climb, the TRACONs are responsible. TRACON airspace, known as terminal control area (TCA), is

in the shape of an upside-down wedding cake. At higher altitudes, the ARTCCs take on the responsibility for providing the ATM services to the aircraft. The process is reversed as the aircraft nears the destination airport. The main types of equipment used in ATM are the radars, displays, computers, and communications equipment. Radars provide information regarding the positions of the aircraft within the airspace. This information is processed in conjunction with the flight plans to predict future locations of the aircraft. The display of this information is used by the air traffic controllers in the facilities to determine if the established rules and procedures would be violated in the near fu-

384

AIR TRAFFIC CONTROL

ture. To prevent violations, the air traffic controllers issue clearances to the pilot to modify the flight path of the aircraft such as to speed up, slow down, climb, descend, and change heading. The procedures used by the air traffic controllers and pilots include rules and methods for operations within the particular airspace. For example, the rules define the minimum separation distance between any two aircraft, the authority of an individual facility over the airspace segment, the transfer of responsibility from one facility to the other, and the phraseology for verbal communications. For pilots, these rules specify their responsibility and authority, flight and navigation procedures, reporting requirements, and compliance with ATM instructions. The communications equipment enable both voice and computer-to-computer communications. Voice communication is used between pilots and the ATM facilities and also between ATM facilities. Information transfer from one facility computer to the next is done using the communications equipment.

HISTORICAL DEVELOPMENT OF THE ATM SYSTEM The present-day ATM system in the United States has evolved in response to the needs of the several different groups of users and providers of the ATM services (2). These groups include air carrier, air taxi, military, general aviation, business aviation, pilots association, and air traffic controllers association. The ATM system has changed with technological advancements in the areas of communication, navigation, surveillance, computer hardware, and computer software. Detailed historical accounts of ATM development are available in Refs. 1 and 3. In the history of ATM development, five periods are easily identifiable. Early aviation developments took place during the period from 1903 to 1925. This period saw the development of aircraft construction methods, use of radio as a navigation aid, nighttime navigation using ground lighting, and the development of airmail service. The important legislative action that marks this period is the Airmail Act of 1925, which enabled the Postmaster General to contract with private individuals and corporations for transporting mail. An important consequence of this Act was that companies like Boeing, Douglas, and Pratt and Whitney got into the business of supplying aircraft and engines to the budding airmail industry. With the increase in air traffic activity, a need for regulation was felt to unify the industry through common sets of rules and procedures. An advisory board made its recommendation in the Morrow Report which led to the signing of the Air Commerce Act into law in 1926. This Act marks the beginning of the second period of ATM development. The period between 1926 and 1934 saw Charles Lindbergh’s flight across the Atlantic, installation of ground-to-air radio in aircraft, development of ground-based radio navigation aids, airline aircraft equipped with two-way radio telephone, radio-equipped air traffic control tower, and the development of a new generation of faster higher-flying transport aircraft capable of being flown solely with reference to cockpit instrumentation. The third phase of the ATM development is marked by the creation of the Bureau of Air Commerce in 1934. During the third phase that lasted until 1955, numerous changes took place that shaped the ATM system to its present form. The principal airlines established interline agreements

in 1935 to coordinate traffic into the Newark, Chicago, and Cleveland airports. The center established at Newark became the first airway traffic control unit (ATCU) in the world. In 1938, the US Congress created the Civil Aeronautics Authority which in 1940 was reorganized as the Civil Aeronautics Administration (CAA). This period saw the development of visual flight rules (VFR) and instrument flight rules (IFR). The civil airways system, controlled airports, airway traffic control areas, even and odd altitude levels, and radio fixes for mandatory position reporting by IFR aircraft were established during this phase. By 1942, 23 ARTCCs (former ATCUs) provided coverage of the complete continental airways system. During the World War II years between 1941 and 1945, the CAA set up approach control facilities at the busiest airports to separate arriving and departing aircraft out to 20 miles. In 1947, the International Civil Aviation Organization (ICAO) was formed. It adopted the US navigation and communication standard as the worldwide standard and English as the common language for air traffic control. The most important development of this period was the radio detection and ranging (radar) device. The postwar era saw the development of direct controller/pilot interaction, implementation of the VHF omnidirectional range (VOR) and distance measuring equipment (DME), installation of the instrument landing system (ILS) for pilot aiding during landing, and application of radar for surveillance in the airport areas. The fourth phase of ATM development occurred during 1955 to 1965. A short-range air navigation system known as the VORTAC system was developed by colocating the civilian VOR and the US Navy developed tactical air navigation (TACAN) system in common facilities. Experience with radar use during the postwar era eventually led to the development of air route surveillance radar (ARSR). The first such system was installed at the Indianapolis Center in 1956. In the same year, the first air traffic control computer was also installed at the Indianapolis Center. Research and development efforts were begun by the CAA for a secondary radar system that would use a ground interrogator to trigger transponders onboard the aircraft and obtain replies to display the aircraft identification and altitude on the controller’s radar screen. An experimental version of this system known as the air traffic control radar beacon system (ATCRBS) was implemented in 1957. In 1958 the US Congress passed the Federal Aviation Act which created the Federal Aviation Agency as the new independent agency to succeed the CAA. Due to the acceptance of radar surveillance as the principal tool for control of air traffic, new separation standards were needed. Other significant changes during this period were the introduction of high-speed commercial jet aircraft and increase in traffic volume. To accommodate these developments and to keep the task of ATM manageable, smaller segments of airspace known as sectors were developed based on air traffic flow patterns and controller workload considerations. To reduce the workload associated with bookkeeping caused by sectorization, a computerized flight information system for updating flight information and automatically printing flight progress strips was developed. By 1963 several of the flight data processing (FDP) computers were placed into operational ATM service. The first prototype of a computerized radar system for arrival and departure control called the automated radar terminal system (ARTS) was installed in the Atlanta, Georgia, air traffic control tower in 1964. In addition to the steady

AIR TRAFFIC CONTROL

80

Millions of operations

60

40

20

Aircraft handled Instrument operations Flight services Airport operations

0 1960

1970

1980

1990

2000

385

overwhelmed the system at some airports. Flow control measures such as ground holding and airborne holding were put into practice for matching the traffic rate with airport acceptance rate. The traffic growth starting from the middle of the fourth phase of the ATM development to the present is shown in Fig. 2. The graphs in the figure are based on the data provided in the FAA Air Traffic Activity report (4), FAA Aviation Forecasts publication (5), and the FAA Administrator’s Fact Book (6). It should be noted that the number of airport operations is representative of usage by all aircraft operators including general aviation while the aircraft handled is representative of higher-altitude traffic reported by the ARTCCs. Several interesting trends can be observed from the graphs: traffic growth subsequent to the Airline Deregulation Act of 1978, traffic decline after the PATCO strike in 1981, and the eventual recovery after approximately 3 years. All the graphs except the one for flight service usage show an increasing trend. The decreasing trend in the flight service usage since 1979 is due to (a) improved cockpit equippage, with part of the service being provided by the airline operations centers (AOCs), and (b) consolidation of the FAA flight service facilities.

Fiscal year Figure 2. Air traffic activity historical data.

progress toward automation, this period of ATM development saw the air traffic controllers get organized as a union called the Professional Air Traffic Controllers Organization (PATCO). The fifth phase of ATM development spans the period from 1965 to the late 1990s. Several administrative changes have taken place during this period. The Department of Transportation (DOT) was created in 1967, and the Federal Aviation Agency was brought under its wings as the Federal Aviation Administration (FAA). The National Transportation Safety Board (NTSB) was created to investigate transportation accidents and report its findings to the Secretary of Transportation. This phase of ATM development has also seen numerous technological changes. Alongside the FDP system for flight data processing, a second system called the radar data processing (RDP) system was developed for integrating information from multiple radar sites, automatic aircraft tracking, and handoff capabilities. The RDP system was implemented in all the ARTCCs by 1974. Both the FDP and RDP systems are parts of the ARTCC host computer. Major terminal facilities upgrade has included the installation of the ARTS-IIIA systems which are capable of tracking both transponder equipped and nonequipped aircraft. ARTS-II and en route ARTS (EARTS) versions of the ARTS system were also developed for low- and medium-activity facilities. Other changes of major significance during this period are the Airline Deregulation Act of 1978, the air traffic controllers strike of 1981 that led to massive firing of air traffic controllers by President Ronald Reagan, and the formation of a new union called the National Air Traffic Controllers Association (NATCA) in 1987. The Airline Deregulation Act made it possible for the airlines to determine their own fare and route structures without government approval. The unprecedented growth that resulted put a strain on the ATM system. For operational advantages, the airlines adopted a hub-and-spoke system that

OPERATIONS WITHIN THE ATM SYSTEM Flight operation within the current ATM system is described via an example flight from the San Francisco International Airport to the Los Angeles International Airport. Some of the facilities that provide separation and flow control services to this flight are shown in Fig. 3. The dashed lines show that the aircraft is tracked by the primary radar and the secondary radar beacon system during the aircraft’s flight through the TRACON and ARTCC airspaces. The airport surveillance radars (ASRs) provide information to the TRACONs, and the air route surveillance radars (ARSRs) provide information to the ARTCCs. The surveillance data along with the filed flight plan provide the information for decision-making to enable safe and efficient flight operations within the airspace. In preparation for the flight, the pilot of the aircraft contacts the Oakland Automated Flight Service Station located in Oakland, California, and furnishes the following flight plan information: type of flight such as VFR or IFR, aircraft identification or pilot’s name, aircraft type such as LJ23 for Learjet 23, departure point such as KSFO, estimated time of departure, altitude, route-of-flight, destination such as KLAX, and the estimated time en route. Based on this information, the air traffic control specialist briefs the pilot. The standard briefing includes current or forecast conditions which may adversely impact the planned flight, a recommendation for VFR flight, a synopsis of weather systems affecting the flight, current weather conditions, en route forecast, a destination forecast, winds aloft forecast, notices to airmen, and ATC delays for IFR flights. In addition to the standard briefing, the pilot can request an abbreviated briefing for updated forecasts and outlook briefing for a planned flight more than 6 hours away. In the case of airline pilots, the weather briefing is provided by the airline dispatcher. On completion of the weather briefing, the flight service specialist enters the flight plan data into the FSS computer. The computer sends the flight plan information to the host computer at the departure ARTCC which for this flight is Oakland ARTCC located in

386

AIR TRAFFIC CONTROL

Figure 3. Air traffic control process. Fremont, California. The flight plan entered into the airline host computer by the airline dispatcher is also sent to the host computer at the ARTCC via the aeronautical data network system (ADNS). The ARTCC host computer checks if preferred routings are applicable to the proposed flight plan. If they are, the flight plan is modified. Thirty minutes prior to the proposed departure from San Francisco International, the flight plan is activated, a transponder code is assigned to the aircraft, and the flight plan data are transmitted from the ARTCC host computer to the ARTS computer at the Bay TRACON located in Oakland, California. Flight plan activation also causes a flight progress strip to be printed at the clearance delivery position in the tower cab at San Francisco International. When the pilot is ready to depart, the pilot contacts the clearance delivery controller at the assigned frequency. The clearance delivery controller confirms that the printed flight progress strip conforms with the letter of agreement between the San Francisco Tower and the Bay TRACON. If changes to the route or altitude are needed, they are entered into the flight data input output (FDIO) system. Based on the facility directives, the clearance delivery controller initially assigns an altitude that is delegated to the local controller. This area is known as the departure fan (1). The clearance delivery controller communicates the complete clearance including the restrictions and the departure frequency to the pilot. The flight progress strip is passed to the ground controller. There is also an automated clearance delivery process known as the predeparture clearance that is available to airlines. The clearance input from the FDIO system in the tower is sent to the

ARTCC host computer which reroutes it to the airline host computer via ADNS. The airline host computer then delivers the clearance to the aircraft communications, addressing, and reporting system (ACARS) in the cockpit or to the gate printer. Once clearance is received, the pilot contacts the ground controller in the tower cab for taxi instructions to the active runway. The ground controller is responsible for separation of aircraft and vehicles on airport movement areas except the active runways. Thus, the ground controller issues instructions to the pilot to safely taxi to the active runway. The ground controller coordinates with the local controller if the aircraft has to cross any active runway. The pilot then contacts the local controller, also referred to as ‘‘tower’’ controller. The local controller sequences this flight into the local flow while ensuring that the aircraft will not be in conflict with the other inbound and outbound aircraft. Next, the local controller instructs the pilot to contact the departure controller at the Bay TRACON. As soon as the ARTS computer at the Bay TRACON detects this flight’s transponder transmissions, it sends a departure message to the host computer at the Oakland ARTCC. The departure controller radar identifies the aircraft and verifies the accuracy of the readout provided by the aircraft’s transponder. Subsequently, the controller advises the pilot that radar contact has been established and authorizes the aircraft to climb to the requested altitude. The controller also vectors the aircraft to join the proper airway. During the initial climb phase, the departure controller is responsible for separating this aircraft from all other aircraft in the vicinity.

AIR TRAFFIC CONTROL

The departure controller initiates a handoff with the Oakland ARTCC before the aircraft leaves the Bay TRACON boundary. Radio contact with the Oakland ARTCC is established before the aircraft enters the ARTCC airspace. The ARSR detects the aircraft and sends the position information to the ARTCC host computer. The host computer uses the flight plan information and the aircraft position information to track the aircraft and display the aircraft’s position, identification, altitude, and groundspeed on the controller’s plan view display (PVD). Using the displayed information and the decision support tools, the controller provides separation services to the aircraft within the sector airspace. As the aircraft climbs from lower to higher altitudes, it is handed off from the low-altitude sector controller to the high-altitude sector controller and then from one high-altitude sector controller to the next until the aircraft nears the ARTCC boundary. It is then handed off to the Los Angeles ARTCC located in Palmdale, California. The Los Angeles ARTCC host computer activates the flight plan about 30 min before the aircraft is scheduled to enter the ARTCC airspace, and the flight progress strip is printed for the sector controller who would receive the handoff. Once the aircraft is automatically handed off by the Oakland Center, the sector controller verifies the accuracy of the altitude readout from the aircraft’s transponder and issues the altimeter setting from the closest airport equipped with a weather observer. The aircraft is continuously tracked by the ARTCC host computer using the ARSR. Separation services are provided by the sector controllers as the aircraft flies from one sector to the next. As the aircraft nears its destination, the Los Angeles high-altitude sector controller initiates a descent clearance to transition the aircraft to the TRACON. Control is transferred from high-altitude controllers to low-altitude controllers until it nears the boundary of the Southern California TRACON located in San Diego, California. At this point, the pilot is instructed to contact the approach control at Southern California TRACON. The flight information from the host computer in the Los Angeles ARTCC is sent to the Southern California TRACON about 30 minutes in advance to prepare for the aircraft’s arrival. As the aircraft enters the TRACON airspace, it is tracked by the ARTS computer and the information is displayed on the controller’s plan position indicator (PPI). Once the handoff is accepted by the approach controller, the aircraft is constrained to continue descent within the confines of the airspace allocated to the approach controller. The approach controller vectors the aircraft to position it for easy transition to the ILS aproach course. The pilot is then cleared for the ILS approach and advised to contact the local controller at the Los Angeles International ATCT and report crossing the final approach fix. Beyond the final aproach fix, the local controller assumes responsibility for sequencing the aircraft into the inbound traffic flow and ensuring the aircraft flight path is conflict free all the way to the touchdown point on the airport surface. The local controller instructs the pilot that the aircraft is cleared to land. After landing, the local controller instructs the pilot to contact the ground controller for taxi instructions to the parking area. The ground controller issues taxi instructions to the pilot and monitors the movement of the aircraft from the tower cab. In reduced visibility conditions, the movement

387

is monitored on a radar display driven by the airport surface detection equipment (ASDE). FUTURE ATM DEVELOPMENTS The future ATM system will be based on collaboration between the cockpit, airline operations centers, Central Flow Control Facility, ARTCCs, TRACONs, and ATCTs. This will be enabled using satellite-based navigation and surveillance systems, datalink technologies, and decision support tools on the ground and in the cockpit. Aircraft would be intelligent data collection and information processing agents and would actively participate in the flow management and separation functions of ATM. The motivations for collaborative ATM are improved safety and economy. Traditionally, the ATM system has been mainly focused on safety. Both the flow management and separation functions are geared toward safety. Flow control is applied in an anticipatory and strategic way to prevent overwhelming the system, and separation minimums are applied tactically to prevent aircraft from getting close to each other. Successful application of these methods is dependent on the predictability of traffic which is derived from knowledge about the intent of the aircraft, their current states, their capabilities, and their operational procedures. The longer-term intent information is provided by the filed flight plan which consists of segments within the fixed route structure, while the shorter-term intent information is obtained by maintaining the track data received from the primary and secondary surveillance radar systems. The airline operators and pilots are interested in the most economical flight operations under the cover of safety provided by the ATM system. Many of their aircraft are equipped with onboard navigation aids which provide the freedom from being constrained to operate on the airway routes where ground-based navigation aids are available. They want to use their navigation ability to fly routes of their choice including shortest distance or wind optimal routes. It is estimated that US airlines incur a loss of $5.5 billion annually due to operations under the current procedures (7). The users also want to be able to negotiate with each other to fly the route of their choice. The challenge for the future ATM system is to allow the user preferences while preserving the predictability of traffic so as not to compromise safety. It is envisioned that in this flexible ATM environment, restrictions would be limited in extent and duration to correct the identified problem. Intermittent positive control would be applied by the ATM system to ensure separation, to preclude exceeding airport capacity, to prevent unauthorized flight through special use airspace, and to ensure safety of flight. This concept has come to be known as the free flight concept (8). The technologies that will enable the transition from the ground-based ATM system to an air–ground-based ATM system are described next. Navigation and Surveillance Systems The global positioning system (GPS) is the emerging navigation system that provides global navigation capability to suitably equipped aircraft. This system, developed by the Department of Defense, is based on a constellation of 24 orbiting satellites that broadcast their positions and the clock time

388

AIR TRAFFIC CONTROL

(9,28). The difference between the time at which the signal is received by the GPS receiver station and the time at which the data were sent by the satellite provides the range with respect to the satellite. These relative positions are used with the broadcast positions to determine the inertial position of the GPS receiver since the broadcast positions are given with respect to an inertial frame of reference. Information from three satellites is adequate for position estimation if the GPS receiver clock is synchronized with the satellite clock or if the altitude is known. By adding information from one more satellite, the GPS receiver clock bias can be removed. Thus, information from four satellites is needed for accurate position determination. The standard positioning service that is available for civil use provides an accuracy of 100 m (328 ft) horizontally 95% of the time and 170 m (560 ft) vertically (10). Accuracies better than 10 m in all three dimensions are available for military use. The positioning accuracy can be significantly improved by using range error corrections broadcast from a ground-based GPS unit located at a surveyed location. This system is known as the differential GPS (DGPS) (9,28). It is also known as the local area augmentation system (LAAS) since it only provides local calibration corrections. An extension of this system is the wide area augmentation system (WAAS), which uses a network of ground-based monitor stations, communications networks, and master stations. The GPS measurements taken by the monitor stations are sent to the master stations where error corrections are computed using the known locations of the monitor stations. The error corrections are uplinked to the airborne system using a satellite, radio, or telephone datalink. The GPS-based technologies will enable precise en route/ terminal area navigation, approach, and landing (11). Accurate area navigation may lead to a more efficient structuring of the airspace and reduction of the separation minimums. In the future it will be possible to transmit the information derived from the airborne GPS or other navigation systems to the ground via a satellite datalink. These data will provide an additional source of surveillance information for the ATM system. The concept for transmitting the aircraft position information to the ATM system is known as automatic dependent surveillance (ADS) (7,28). ADS will significantly impact oceanic ATM since radar coverage is unavailable over the oceans. Accurate surveillance information will allow a reduction of oceanic ATM separation minimums and bring the service standards in line with what is available over the continental United States. Such a system would also improve the safety of domestic flight operations by providing backup surveillance information during radar system outages. A broadcast version of the ADS known as ADS-broadcast (ADS-B) is also under development. In addition to providing the basic ADS capabilities, this system is intended for broadcasting the aircraft’s position so that it can be read and displayed in the cockpits of nearby aircraft (7). This system is also expected to aid the air-to-air and air–ground cooperative decision-making process. ADS is also envisioned to be the surveillance system of choice for other countries that do not yet have the kind of radar coverage as in the United States. Communications The communications system of the future is expected to shift from being largely voice-based to datalink-based (7,28). One

of the drivers for this change is frequency congestion at busy facilities where controllers are constantly in voice communication with several aircraft with little time left for standard readback of clearance information. The promise of datalink is that standard information such as speed, altitude, and heading assignments along with changes entered by the controller can be quickly sent to the cockpit. Several different design options are being considered (7). The first option is to send the information from the ARTCC host computer to the airline host computer via the aeronautical data network system which would route the information to the cockpit using the already available ACARS service provided by Aeronautical Radio Inc. (ARINC). The second option for uplinking data to the cockpit is to communicate with the onboard mode-S transponders. Although the mode-S transponders are capable of selective interrogation and have a built-in datalink capability to support data communication with the ground and similarly equipped aircraft operating in the neighborhood, they are not designed to support large amounts of data transfer. The bandwidth of the ACARS and mode-S systems can be increased to overcome the transfer rate limitations. In addition to these two options, other satellite-based high bandwidth communication systems are also being considered. The improved datalink capability will permit clearance delivery, data exchange, and even negotiation of complete flight segments between ATC and cockpit. As data communication increases, voice will be used predominantly for checks and confirmations. For example, the pilot would verbally confirm to the controller that the clearance has been received rather than reading back the clearance. The complete clearance could be digitally transmitted back to ground for record-keeping and verification purposes. Increased use of datalink and the reduced role of voice communications is not without human factor concerns. Pilots are aware of the traffic situation by listening to the communication between other pilots and controllers. The voice system therefore provides yet another safety net for flight operations. Other concerns are related to cockpit workload increase caused by the need to interpret large amounts of displayed data sent using the datalinks and boredom caused by the lack of aural stimulus. Boredom added to the natural tendency to sleep during the nighttime hours has safety implications for nighttime flight operations. Weather Prediction Systems Approximately 40% of aviation accidents are attributed to adverse weather conditions (12). Weather is the largest single contributor to traffic flow problems. It is also the least predictable. Although advances have been made in weather processing, adequate sensor coverage has not been available to provide the spatial and temporal scale of weather observations needed for accurate short-term predictions. Since most flights are completed within 2 h, the focus is on events that occur on a 0 to 2 h time scale and within a 50 mi (80 km) space scale. This spatiotemporal scale is known as mesoscale (12). To enable mesoscale predictions, a national network of Doppler weather radars is being developed. This network, known as the next generation radar (NEXRAD) network, is designed for wide-area surveillance and detection of weather phenomena in the en route areas. A special-use Doppler radar

AIR TRAFFIC CONTROL

termed terminal doppler weather radar (TDWR) has been developed to provide windshear data within the terminal areas. This system will be integrated with the low-level windshear alert system (LLWAS) to enhance the weather prediction accuracy (12,13). LLWAS uses direct anemometer measurements. Plans have been made to field automated surface weather observing systems at small and medium-sized airports. This system, known as the automated weather observing system (AWOS), is designed to provide data to the national observation network. Traditionally, vertical wind profiling data consisting of windspeed, temperature, pressure, and humidity aloft have been measured by launching balloon systems from widely distant locations. In the future vertical wind profiling will be done using a microwave Doppler system. An important resource for aviation weather is the wind and temperature data observed by thousands of aircraft for navigation and performance monitoring. Some airlines already have their flights provide wind and temperature data periodically via ACARS downlink. As datalink technologies mature, it will be possible to collect the airborne observation data in large databases to augment the data collected by the ground-based observation systems. Access to airborne observation data will enable identification of turbulence regions which are usually much smaller than what can be predicted using the ground-based systems (12). Finally, improved weather observations will also be available from weather satellite systems using radar and radiometer measurements of winds, temperature, humidity, and precipitation. In addition to the enhancements in the weather sensor systems, the computational and information processing algorithms are also expected to improve. Computational algorithms will make short-term forecasts (nowcasts) possible within 10 min of thunderstorm formation by detecting temperature and moisture boundaries in the observation data. The currently available weather systems that generate large amounts of data which the aviation user has to sort through to obtain the needed facts will be replaced by rule-based weather information systems (12). These systems will provide precise weather messages in contrast with the often lengthy and ambiguous weather briefings provided by the presently available systems. Decision Support Systems As progress is made toward a more cooperative and flexible air traffic environment, the biggest challenge for ATM is to improve or at least retain the current levels of safety. Currently, safety is defined in terms of separation requirements. Lateral separation is maintained largely by constraining the traffic to fly on fixed airways. Vertical separation is achieved by constraining the aircraft to fly at assigned altitudes. Longitudinal separation is maintained by ensuring that the aircraft on the same airway are separated by a physical distance as a function of the relative speed of the aircraft, their location with respect to the surveillance radar, and their weight class. The path constraints make the traffic movement predictable, which in turn makes it possible to identify separation violations that are likely to occur in the future. In a flexible air traffic environment with few constraints on traffic movement, decision support systems will be needed for achieving the same or better levels of predictability. These systems will predict the future positions of the aircraft, check if they would

389

violate the separation minimums in advance, and provide conflict resolution options to the controller. Advanced automation systems such as Automated En Route Air Traffic Control (AERA) system and the CenterTRACON Automation System (CTAS) that are under development use trajectory prediction methods for providing the data needed for conflict detection, conflict resolution and traffic management (14,15). The trajectory prediction process involves using the knowledge of the present states and performance characteristics of the aircraft along with the intent information to determine how the states would evolve along the intended path. Factors that influence trajectory prediction are atmospheric conditions such as ambient temperature and wind velocity, the capabilities of the onboard navigation equipment, and the piloting strategies (16). The type and accuracy of the navigation equipment directly translates into how precisely the aircraft is able to maintain track with reference to its desired course. Piloting strategies such as flying a constant airspeed, an average groundspeed, or attempting to reach a particular location at a fixed time directly influence the along-track position of the aircraft. In the future with advances in datalink technologies, shorter-term intent information consisting of waypoints provided periodically by the aircraft operators may be acceptable in lieu of the long-term flight plan. The data-linked information consisting of the state of the aircraft, measured wind velocity, and ambient temperature is expected to improve prediction accuracy. Along with the advancement of longer-term prediction techniques that are needed for strategic planning and conflict resolution, improvement of shorter-term trajectory prediction methods will support tactical conflict detection and resolution needed to support free flight. Short-term trajectory prediction is based solely on the knowledge of the present state of the aircraft. Knowledge of the flight plan and weather are not needed. The prediction method consists of propagating the states of the aircraft from the present to a short time into the future by assuming that aircraft controls are fixed for the duration of prediction. For example, a constant turn rate is assumed for the aircraft in a turn. Currently, short-term trajectory prediction is done by the ARTCC host computer software and can be graphically displayed as a trend vector on the air traffic controller’s plan view display (PVD) (17). Controllers can use the trend vectors to detect conflicts. The host computer program also detects conflicts that are likely to occur within 3 min based on the predicted trajectories. A feature of the decision support systems of the future will be the ability to detect conflicts with high reliability. The conflict detection process consists of checking if two or more aircraft will be within a small region at the same time. Several different algorithms are available for this task. Conflict detection can be done by using the brute-force method of comparing every pair of aircraft trajectories at each time instant along their entire length. This method is computationally intensive, and therefore the brute-force method has been combined with heuristics. Several heuristics are used for eliminating most trajectory combinations, and the brute-force method is applied on the remaining trajectories (18). Very efficient sorting-based methods have also been developed for conflict detection (19). Efficiency of conflict detection methods is important because they are used often for examination of the planning and conflict resolution alternatives.

390

AIR TRAFFIC CONTROL

In addition to decision-aiding for the air traffic control function, advanced automation systems will aid the traffic planning process. These systems will use the predicted trajectories to identify regions of high traffic density, forecast center/sector workload, and assist controllers in scheduling traffic into an airport to optimally utilize the available capacity. Some of these capabilities are already available within the Enhanced Traffic Management System (ETMS) that uses strategic prediction of traffic volume for its monitor/alert function (8). It has been suggested that this function should be extended to include measures of sector complexity and controller workload. As the traffic demand continues to grow and nonairway direct routes or wind optimal routes are flown, methods for predicting sector and center workload will be crucial for managing the air traffic operations. Since center/sector complexity is related to the level of difficulty experienced by the controllers, automation systems will utilize structural and flow complexity mesures for aiding the traffic management staff in resource planning, rerouting, and examining alternative airspace configurations. Human Factors ATM is a complex interaction between sensors, computers, and communications with humans as decision makers. Controllers and supervisors in all air traffic control facilities share each others’ work, supervise it, and provide assistance for safety and efficiency (20). The workspace and the computer interface are designed so that other controllers can easily assess the traffic situation and take over the control position. With the evolution of automation systems, the trends are toward the individual controller interacting more with the automation system and less with other controllers (20). The preferences and choices of individual controllers may make the system less understandable to others, thus making it difficult for other controllers to provide assistance or assume control. Automation will need to provide for easy access to the individual controllers preferences so that other controllers are able to analyze the traffic situation and make a smooth transition into the control position. The development of automation systems will have to be guided by correct assumptions about controller’s knowledge and ability and the air traffic control procedures. The current trends have been to automate mundane and routine tasks such as data entry and updating of information while leaving tasks that require human ingenuity to the humans in the control loop. In the future, advanced decision aids will generate choices for the controller and also assist the controller in evaluating the outcome of a particular control choice. For example, if the controller wishes to investigate whether a change of heading of the aircraft will resolve a predicted separation violation, the automation system will build the proposed trajectory and compare it against all other trajectories to determine if the proposed resolution would resolve the conflict. Both providing choices and testing choices have human factors implications. In the first case, if the controller makes decisions based solely on the choices presented, the controller may eventually lose the skills needed for generating alternative solutions. In the second case when the controller examines the alternatives using automation, the controller may lose the analytical skills needed for assessing possible situations that may result as a consequence of a particular choice.

Preventing loss of crucial traffic control skills will have to be a part of the design criteria for future automation systems. As the tools move from a monitoring role to a decision-aiding role, they have to be designed with additional safety features. In addition, the tools should be designed to degrade gracefully such that the controller is able to smoothly transition in the event of a failure. New human factors issues will need to be addressed as ATM transitions from a centrally controlled and managed system to a distributed system where the cockpit and the airline operations centers participate in the control and management functions. The automation systems will have to keep all the participants involved in the control loop so that they are knowledgeable about the traffic situation. Traffic situation displays and automation systems will have to be provided in the cockpit to inform the crew of the traffic in the neighborhood and to enable them to plan and coordinate flight path changes with crews of other aircraft. The traffic monitoring, management, and separation responsibilities in the cockpit may increase the crew workload. This is especially significant because the number of crew members is expected to decrease and assisting each other to solve traffic problems may detract them from their piloting and flight management responsibilities. Shared traffic control and management responsibilities with the cockpit has the potential for increased controller workload caused by the communications needed for cooperative resolution of traffic situations. One of the reasons for increasing automation in the ATM system has been to maintain the controller workload with traffic growth. Airspace sectorization and procedure development have also been guided by workload considerations. Additionally, traffic management decisions are influenced by controller workload assessments. For example, the monitor/alert function of the Enhanced Traffic Management System is based on a traffic volume threshold that is acceptable with regard to controller workload. Research on controller workload has been motivated by a desire to understand occupational stress, reduce operational errors, enhance safety, task performance, and efficiency, and improve controller training. Three distinct approaches have been employed for workload research. The first technique attempts to measure the physiological state of the air traffic controller. Measurements of this type have included galvanic skin response (GSR), heart rate, electrocardiogram (ECG), blood pressure, biochemical analysis of body fluids, and behavioral symptoms (21). The second method attempts to measure the controller workload in terms of the physical interactions with the human–computer interface system. Measurements of this type include number of keystrokes, slew ball entries, and communications per unit of time (22). Since the job of air traffic control is primarily cognitive and information-intensive, rather than physical and labor-intensive, the third method attempts to measure the psychological state of the air traffic controller. Workload is measured in terms of the cognitive demand of the task and the available mental capacity (23). Each of the three methods of workload research has its limitations. The first method based on physiological measurements has had limited success as an indicator of stress related to workload (21). The main difficulty with the second approach of assessing workload in terms of physical interactions with the human–computer interface system is that it ignores the fact that cognitive workload can be significant.

AIR TRAFFIC CONTROL

Reference 24 suggests that the task of maintaining vigilance for critical events such as loss of separation, altitude deviations, VFR pop-ups, incorrect pilot readbacks, and other infrequent events imposes considerable mental workload. The third approach is limited in that the task demand and mental capacity are not related in a straightforward way. The research in Ref. 23 suggests that the relationship between mental capacity and task demand depends on the strategies employed to meet the demand and on the skill in choosing the most efficient strategy in cases where multiple options are available. Inadequacies of these methods have led to attempts at understanding the cognitive structures employed by the controller. The testing methods have included (a) memory tasks such as traffic drawing and flight strip recall and (b) the assessment task of potential conflicts between aircraft (25). Subjective ratings of how operationally meaningful concepts such as weather, traffic volume, and projected proximity between aircraft are related has been used to determine the conceptual structures for decision-making (26). Research into the cognitive structures employed in air traffic control suggests that controllers use the spatial and temporal traffic patterns rather than the instantaneous position of the aircraft displayed on the controller’s workstation (25,26). It is believed that the five cognitive stages that form the bridge between the events and actions are selective attention, perception, situation awareness, planning and decision-making, and action execution (24). The research in Ref. 26 has found that weather is a central concept in the controller’s cognitive process because it impacts aircraft routing and imposes flow restrictions. Factors that reduce available airspace such as weather phenomena and special use airspace (SUA) within the sector increase workload. Traffic involving aircraft with vastly different performance characteristics increases controller workload. The establishment procedure for en route sectors calls for sectors to be designed to reduce/prevent the mix of such aircraft (27). Mixed traffic could be an issue as the ATM transitions into a more flexible environment. Research indicates that situation awareness is more difficult in crowded, complex, and heterogeneous airspace (24). Further research is expected to result in traffic pattern recognition algorithms that will use traffic data to predict controller workload. Once these algorithms are calibrated against the ratings provided by the controllers, it will be possible to use them for ATM functions.

GLOBAL ATM Although ATM has been discussed in terms of the domestic air traffic operations within the United States, it is recognized that civic aviation is an international activity. There are 183 International Civil Aviation Organization (ICAO) member nations that are interested in the development of airborne systems, ground systems, standards, and procedures for enabling seamless operations worldwide. For achieving this goal, the ICAO develops standards which are collectively known as the International Standards and Recommended Practices (1). Except for a few minor differences, the ATM system in the United States conforms to the ICAO standards. The airspace in Europe is shared by several nations, and the 36 member states of the European Civil Aviation Confer-

391

ence (ECAC) have been working toward harmonizing the ATC systems. Eurocontrol, the management organization of the ECAC, has the goal of integrating the ATC systems of various nations toward a uniform European air traffic management system (EATMS). The development of EATMS has to address the diverse needs of all the nations in Europe. For guiding the development of the future global ATM system, the ICAO has developed a future air navigation system (FANS) concept for communications navigation and surveillance combined with air traffic management (CNS/ATM). ICAO recommends use of VHF radio for voice communications and aeronautical mobile satellite service (AMSS) for both voice and data communications. In high-density areas, mode-S is the datalink system of choice. It calls for the development of a multimode receiver standard for supporting the global navigation satellite system (GNSS), which includes the GPS developed by the United States and the global navigation satellite system (GLONASS) developed by the Russian Federation, instrument landing systems (ILS), and microwave landing system (MLS). In addition to GNSS, the international standard allows the aircraft operator to use navigation equipment that meets the required navigation performance (RNP) requirements in the particular class of airspace. Automatic dependent surveillance (ADS) is slated to be the surveillance system of the future for both domestic and oceanic airspaces. Surveillance information for operations within the terminal area will be provided by the mode-S transponder system. In the future, primary radar will be used for weather only. For collision avoidance, the traffic-alert and collision avoidance system (TCAS) has been in use in the United States, but the ICAO standard which is being developed calls for the aircraft collision avoidance system (ACAS). This system is required to display the locations of the surrounding traffic for situational awareness and for enabling cooperative air–ground decision-making. The future ATM developments in the United States will both influence and be influenced by the ICAO standards.

BIBLIOGRAPHY 1. M. S. Nolan, Fundamentals of Air Traffic Control, Belmont, CA: Wadsworth, 1994. 2. S. Kahne and I. Frolow, Air traffic management: Evolution with technology, IEEE Control Syst. Magazine, 16 (4): 12–21, August 1996. 3. G. A. Gilbert, Historical development of the air traffic control system, IEEE Trans. Commun., 21: 364–375, 1973. 4. N. Trembley, FAA Air Traffic Activity, Washington, DC: Federal Aviation Administration, US Department of Transportation, 1994. 5. Office of Aviation Policy and Plans, FAA Aviation Forecasts— Fiscal Year 1992–2003, Washington, DC: Federal Aviation Administration, US Department of Transportation, 1992. 6. Office of Business Information and Consultation, Administrator’s Fact Book, Washington, DC: Federal Aviation Administration, US Department of Transportation, 1996. 7. T. S. Perry, In search of the future of air traffic control, IEEE Spectrum, 34 (8): 19–35, August 1997. 8. Final Report of the RTCA Task Force 3 Free Flight Implementation, RTCA, Inc., Washington, DC, October 26, 1995. 9. B. W. Parkinson and J. J. Spilker, Jr. (eds.), Global Positioning

392

ALARM SYSTEMS System: Theory and Applications, Vols. I and II, Washington, DC: American Institute of Aeronautics and Astronautics, 1996.

10. Federal Aviation Administration, Airworthiness approval of global positioning system (GPS) navigation equipment for use as a VFR and IFR supplemental navigation system, Advisory Circular, AC No. 20-138, Washington, DC, May 25, 1994. 11. L. Schuchman, B. D. Elrod, and A. J. Van Dierendonck, Applicability of an augmented GPS for navigation in the national airspace system, Proc. IEEE, 77: 1709–1727, 1989. 12. J. McCarthy, Advances in weather technology for the aviation system, Proc. IEEE, 77: 1728–1734, 1989. 13. J. Evans and D. Turnbull, Development of an automated windshear detection system using Doppler weather radar, Proc. IEEE, 77: 1661–1673, 1989. 14. D. J. Brudnicki and D. B. Kirk, Trajectory modeling for automated en route air traffic control (AERA), Proc. Amer. Control Conf., Seattle, Washington, June 21–23, 1995, 5: pp. 3425– 3429. 15. H. Erzberger, T. J. Davis, and S. Green, Design of center-TRACON automation system, AGARD Guidance and Control Symp. Mach. Intell. Air Traffic Manage., Berlin, Germany, 1993. 16. G. B. Chatterji, B. Sridhar, and K. Bilimoria, En-route flight trajectory prediction for conflict avoidance and traffic management, AIAA Guidance, Navigation Control Conf., AIAA 96-3766, San Diego, CA, 1996. 17. MIT Lincoln Laboratory, Air Traffic Control Overview: Kansas City ARTCC, MIT Lincoln Laboratory, Group 41, Lexington, MA, 1997. 18. D. R. Isaacson and H. Erzberger, Design of a conflict detection algorithm for the center/TRACON automation system, 16th Digital Avionics Syst. Conf., Irvine, California, 1997. 19. B. Sridhar and G. B. Chatterji, Computationally efficient conflict detection methods for air traffic management, Proc. Amer. Control Conf., Albuquerque, NM, June 4–6, 1997, 2: pp. 1126–1130. 20. V. D. Hopkin, Man–machine interface problems in designing air traffic control systems, Proc. IEEE, 77: 1634–1642, 1989. 21. J. H. Crump, Review of stress in air traffic control: Its measurement and effects, Aviation Space Environ. Med., 50 (3): 243– 248, 1979. 22. M. D. Rodgers, C. A. Manning, and C. S. Kerr, Demonstration of POWER: Performance and objective workload evaluation research, Proc. Hum. Factors Soc. 38th Annu. Meet., Nashville, TN, 1994, p. 941. 23. A. T. Welford, Mental work-load as a function of demand, capacity, strategy and skill, Ergonomics, 21 (3): 151–167, 1978. 24. C. D. Wickens, A. S. Mavor, and J. P. McGee (eds.), Flight to the Future; Human Factors in Air Traffic Control, Washington, DC: National Academy Press, 1997. 25. M. S. Schlager, B. Means, and C. Roth, Cognitive task analysis for the real(-time) world, Proc. Hum. Factors Soc. 34th Annu. Meet., Orlando, FL, 1990, pp. 1309–1313. 26. K. Harwood, R. Roske-Hofstrand, and E. Murphy, Exploring conceptual structures in air traffic control (ATC), Proc. 6th Int. Symp. Aviation Psychol., Columbus, OH, 1991, pp. 466–473. 27. Air Traffic Service, Air Traffic Control FAA Order 7210.46, Federal Aviation Administration, US Department of Transportation, Washington, DC, March 16, 1984. 28. M. Kayton and W. R. Fried, Avionics Navigation Systems, New York: Wiley, 1997.

B. SRIDHAR G. B. CHATTERJI NASA Ames Research Center

AIRCRAFT COMPUTERS. See AIR TRAFFIC. AIRCRAFT. See AIR TRAFFIC.

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering c 1999 John Wiley & Sons, Inc. Copyright

AIRCRAFT COMPUTERS The aircraft industry and the computer industry are relative newcomers in two centuries of technical innovation. It is only natural that these powerful industries have merged to provide continuous improvements in capabilities and services for aircraft customers. Landau (1) defines an aircraft as any structure or machine designed to travel through the air. He then defines a computer as a person who computes or a device used for computing. From these definitions, an aircraft computer is a device used on (or in association with) any air-traveling machine or structures used to make computations. Computers can be found in every aspect of the aircraft industry. On the aircraft, there are computers for flight control and display, computers monitoring and regulating flight functions, computers recording and processing flight activities, computers providing passenger entertainment, and computers providing communication and navigation. Equally important are the ground-based computers at airports, maintenance depots, and air traffic control stations that provide services for all aspects of flight. Figure 1 shows a typical aircraft central computer (CC) used in modern fighter aircraft. This particular computer is also referred to as a fire-control computer (FCC), because it directs the delivery of weapons in conjunction with the aircraft’s sensor systems. Aircraft Analog Computers. Early aircraft computers were used to take continuous streams of inputs to provide flight assistance. Examples of aircraft analog inputs are fuel gauge readings, throttle settings, and altitude indicators. Landau (1) defines an analog computer as a computer for processing data represented by a continuous physical variable, such as electric current. Analog computers monitor these inputs and implement a predetermined service when some set of inputs calls for a flight control adjustment. For example, when fuel levels are below a certain point, the analog computer would read a low fuel level in the aircraft’s main fuel tanks and would initiate the pumping of fuel from reserve tanks, or balancing fuel between wing fuel tanks. Some of the first applications of analog computers to aircraft applications were for automatic pilot applications, where these analog machines took flight control inputs to hold altitude and course. The analog computers use operational amplifiers to build the functionality of summers, adders, subtracters, and integrators on the electric signals. Aircraft Digital Computers. As the technologies used to build digital computers evolved, digital computers became smaller, lighter, and less power-hungry, and produced less heat. This made them increasingly acceptable for aircraft applications. Digital computers are synonymous with stored-program computers. A stored-program computer has the flexibility of being able to accomplish multiple different tasks simply by changing the stored program. Analog computers are hard-wired to perform one and only one function. Analog computers’ data, as defined earlier, are continuous physical variables. Analog computers may be able to recognize and process numerous physical variables, but each variable has its unique characteristics that must be handled during processing by the analog computer. The range of output values for the analog computer is bounded as a given voltage range; if they exceed this, they saturate. Digital computers are not constrained by physical variables. All the inputs and outputs of the digital computer are in a digital representation. The processing logic and algorithms performed by the computer work in a single representation of the cumulative data. It is not uncommon to see aircraft applications that have analog-to-digital and digital-to-analog signal 1

2

AIRCRAFT COMPUTERS

Fig. 1. Typical aircraft central computer.

converters. This is more efficient than having the conversions done within the computers. Analog signals to the digital computer are converted to digital format, where they are quickly processed digitally, and returned to the analog device through an digital-to-analog converter as an analog output for that device to act upon. These digital computers are smaller, more powerful, and easier to integrate into multiple areas of aircraft applications. Landau (1) defines a digital computer as a computer for processing data represented by discrete, localized physical signals, such as the presence or absence of an electric current. These signals are represented as a series of bits with word lengths of 16, 32, and 64 bits. See microcomputers for further discussion. Wakerly (2) shows number systems and codes used to process binary digits in digital computers. Some important number systems used in digital computers are binary, octal, and hexadecimal numbers. He also shows conversion between these and base-10 numbers, as well as simple mathematical operations such as addition, subtraction, division, and multiplication. The American Standard Code for Information Interchange (ASCII) of the American National Standard Institute is also presented, which is Standard No. X3.4-1968 for numerals, symbols, characters, and control codes used in automatic data-processing machines, including computers. Microcomputers. The improvements in size, speed, and cost through computer technologies continually implement new computer consumer products. Many of these products were unavailable to the average consumer until recently. These same breakthroughs provide enormous functional improvements in aircraft computing. Landau (1) defines microcomputers as very small, relatively inexpensive computers whose central processing unit is a microprocessor. A microprocessor (also called MPU or central processing unit [CPU]) communicates with other devices in the system through wires (or fiber optics) called lines. Each device has a unique

AIRCRAFT COMPUTERS

3

address, represented in binary format, that the MPU recognizes. The number of lines is also the address size in bits. Early MPU machines had 8-bit addresses. Machines of 1970–1980 typically had 16-bit addresses; modern MPU machines have 256 bits. Common terminology for an MPU is random-access memory (RAM), read-only memory (ROM), input– output, clock, and interrupts. RAM is volatile storage. It holds both data and instructions for the MPU. ROM may hold both instructions and data. The key point of ROM is that it is nonvolatile. Typically, in an MPU, there is no operational difference between RAM and ROM other than its volatility. Input–output is how data are gotten to and from the microcomputer. Output may be from the MPU, ROM, or RAM. Input may be from the MPU or the RAM. The clock of an MPU synchronizes the execution of the MPU instructions. Interrupts are inputs to the MPU that cause it to (temporarily) suspend one activity in order to perform a more important activity. An important family of MPUs that greatly improved the performance of aircraft computers is the Motorola M6800 family of microcomputers. This family offered a series of improvements in memory size, clock speeds, functionality, and overall computer performance. Personal Computers. Landau (1) defines personal computers as electronic machines that can be owned and operated by individuals for home and business applications such as word processing, games, finance, and electronic communications. Hamacher et al. (3) explain that rapidly advancing very large-scale integrated circuit (VLSI) technology has resulted in dramatic reductions in the cost of computer hardware. The greatest impact has been in the area of small computing machines, where it has led to an expanding market for personal computers. The idea of a personally owned computer is fairly new. The computational power available in hand-held toys today was only available through large, costly computers in the late 1950s and early 1960s. Vendors such as Atari, Commodore, and Compaq made simple computer games household items. Performance improvements in memory, throughput, and processing power by companies such as IBM, Intel, and Apple made facilities such as spreadsheets for home budgets, automated tax programs, word processing, and three-dimensional virtual games common household items. The introduction of Microsoft’s Disk Operating System (DOS) and Windows has also added to the acceptance of the personal computers through access to software applications. Improvements in computer technology offer continual improvements, often multiple times a year. The durability and portability of these computers is beginning to allow them to replace specialized aircraft computers that had strict weight, size, power, and functionality requirements.

Avionics In the early years of aircraft flight, technological innovation was directed at improving flight performance through rapid design improvements in aircraft propulsion and airframes. Secondary development energies went to areas such as navigation, communication, munitions delivery, and target detection. The secondary functionality of aircraft evolved into the field of avionics. Avionics now provides greater overall performance and accounts for a greater share of aircraft life-cycle costs than either propulsion or airframe components. Landau (1) defines avionics [avi(ation) + (electr)onics] as the branch of electronics dealing with the development and use of electronic equipment in aviation and astronautics. The field of avionics has evolved rapidly as electronics has improved all aspects of aircraft flight. New advances in these disciplines require avionics to control flight stability, which was traditionally the pilot’s role. Aircraft Antennas. An important aspect of avionics is receiving and transmitting electromagnetic signals. Antennas are devices for transmitting and receiving radio frequency (RF) energy from other aircraft, space applications, or ground applications. Perry and Geppert (4) illustrates the aircraft electromagnetic spectrum, influenced by the placement and usage of numerous antennas on a commercial aircraft. Golden (5)

4

AIRCRAFT COMPUTERS

illustrates simple antenna characteristics of dipole, horn, cavity-backed spiral, parabola, parabolic cylinder, and Cassegrain antennas. Radiation pattern characteristics include elevation and azimuth. The typical antenna specifications are polarization, beam width, gain, bandwidth, and frequency limit. Computers are becoming increasingly important for the new generation of antennas, which include phased array antennas and smart-skin antennas. For phased array antennas, computers are needed to configure the array elements to provide direction and range requirements between the radar pulses. Smart-skin antennas comprise the entire aircraft’s exterior fuselage surface and wings. Computers are used to configure the portion of the aircraft surface needed for some sensor function. The computer also handles sensor function prioritization and deinterleaving of conflicting transmissions. Aircraft Sensors. Sensors, (the eyes and ears) of aircraft, are electronic devices for measuring external and internal environmental conditions. Sensors on aircraft include devices for sending and receiving RF energy. These types of sensors include radar, radio, and warning receivers. Another group of sensors are the infrared (IR) sensors, which include lasers and heat-sensitive sensors. Sensors are also used to measure direct analog inputs; altimeters and airspeed indicators are examples. Many of the sensors used on aircraft have their own built-in computers for serving their own functional requirements such as data preprocessing, filtering, and analysis. Sensors can also be part of a computer interface suite that provides key aircraft computers with the direct environmental inputs they need to function. Aircraft Radar. Radar (radio detection and ranging) is a sensor that transmits RF energy to detect air and ground objects and determines parameters such as the range, velocity, and direction of these objects. The aircraft radar serves as its primary sensor. Several services are provided by modern aircraft radar. These include tracking, mapping, scanning, and identification. Golden (5) states that radar is tasked either to detect the presence of a target or to determine its location. Depending on the function emphasized, a radar system might be classified as a search or a tracking radar. Stimson (6) describes the decibel (named after Alexander Graham Bell) as one of the most widely used terms in the design and description of radar systems. The decibel (dB) is a logarithmic unit originally devised to express power ratios, but also used to express a variety of other ratios. The Power ratio in dB is expressed as 10 log10 P2 /P1 , where P2 and P1 are the power levels being compared. Expressed in terms of voltage the gain is (V 2 /V 1 )2 dB provided the input voltage V 1 and output voltage V 2 are across equal resistances. Stimson (6) also explains the concept of the pulse repetition frequency (PRF), which is the rate at which a radar system’s pulses are transmitted: the number of pulses per second. The interpulse period T of a radar is given by T = 1/PRF. For a PRF of 100 Hz, the interpulse period would be 0.01 s. The Doppler effect, as described by Stimson (6), is a shift in the frequency of a radiated wave, reflected or received by an object in motion. By sensing Doppler frequencies, radar not only can measure range rates, but can also separate target echoes from clutter, or can produce high-resolution ground maps. Computers are required by an aircraft radar to make numerous and timely calculations with the received radar data, and to configure the radar to meet the aircrew’s needs. Aircraft Data Fusion. Data fusion is a method for integrating data from multiple sources in order to give a comprehensive solution to a problem (multiple input, single output). For aircraft computers, data fusion specifically deals with integrating data from multiple sensors such as radar and infrared sensors. For example, in ground mapping, radar gives good surface parameters, while the infrared sensor provides the height and size of items in the surface area being investigated. The aircraft computer takes the best inputs from each sensor, provides a common reference frame to integrate these inputs, and returns a more comprehensive solution than either single sensor could have given. Aircraft Navigation. Navigation is the science of determining present location, desired location, obstacles between these locations, and best courses to take to reach these locations. An interesting pioneer of aircraft navigation was James Harold Doolittle (1886–1993). Best known for his aircraft-carrier-based bomber raid on Tokyo in World War II. General Doolittle received his master’s and doctor of science degrees in aeronautics

AIRCRAFT COMPUTERS

5

from Massachusetts Institute of Technology, where he developed instrumental blind flying in 1929. He made navigation history by taking off, flying a set course, and landing without seeing the ground. For a modern aircraft, with continuous changes in altitude, airspeed, and course, navigation is a challenge. Aircraft computers help meet this challenge by processing the multiple inputs and suggesting aircrew actions to maintain course, avoid collision and weather, conserve fuel, and suggest alternative flight solutions. An important development in aircraft navigation is the Kalman filter. Welch and Bishop (7) state that in 1960, R.E. Kalman published his famous paper describing a recursive solution to the discrete-data linear filtering problem. Since that time, due in large part to advances in digital computing, the Kalman filter has been the subject of extensive research and application, particularly in the area of autonomous or assisted navigation. The Kalman filter is a set of mathematical equations that provides an efficient computational (recursive) implementation of the least-squares method. The filter is very powerful in several aspects: it supports estimation of past, present, and even future states, and it can do so even when the precise nature of the modeled system is unknown. The Global Positioning System (GPS) is a satellite reference system that uses multiple satellite inputs to determine location. Many modern systems, including aircraft, are equipped with GPS receivers, which allow the system access to the network of GPS satellites and the GPS services. Depending on the quality and privileges of the GPS receiver, the system can have an instantaneous input of its current location, course, and speed within centimeters of accuracy. GPS receivers, another type of aircraft computer, can also be programmed to inform aircrews of services related to their flight plan. Before the GPS receiver, the inertial navigation systems (INS) was the primary navigation system on aircraft. Fink and Christiansen (8) describe inertial navigation as the most widely used “self-contained” technology. In the case of an aircraft, the INS is contained within the aircraft, and is not dependent on outside inputs. Accelerometers constantly sense the vehicle’s movements and convert them, by double integration, into distance traveled. To reduce errors caused by vehicle attitude, the accelerometers are mounted on a gyroscopically controlled stable platform. Aircraft Communications. Communication technologies on aircraft are predominately radio communication. This technology allows aircrews to communicate with ground controllers and other aircraft. Aircraft computers help establish, secure, and amplify these important communication channels. Aircraft Displays. Displays are visual monitors in aircraft that present desired data to aircrews and passengers. Adam and Gibson (9) illustrate F-15E displays used in the Gulf War. These illustrations show heads-up displays (HUDs), vertical situation displays, radar warning receivers, and low-altitude navigation and targeting system (Lantirn) displays typical of modern fighter aircraft. Sweet (10) illustrates the displays of a Boeing 777, showing the digital bus interface to the flight-deck panels and an optical-fiber data distribution interface that meets industry standards. Aircraft Instrumentation. Instrumentation of an aircraft means installing data collection and analysis equipment to collect information about the aircraft’s performance. Instrumentation equipment includes various recorders for collecting real-time flight parameters such as position and airspeed. Instruments also capture flight control inputs, environmental parameters, and any anomalies encountered in flight test or in routine flight. One method of overcoming this limitation is to link flight instruments to ground recording systems, which are not limited in their data recording capacities. A key issue here is the bandwidth between the aircraft being tested and its ground (recording) station. This bandwidth is limited and places important limitations on what can be recorded. This type of data link is also limited to the range of the link, limiting the aircraft’s range and altitude during this type of flight test. Aircraft computers are used both in processing the data as they are being collected on the aircraft and in analyzing the data after they have been collected. Aircraft Embedded Information Systems. Embedded information system is the latest terminology for an embedded computer system. The software of the embedded computer system is now referred to as embedded information. The purpose of the aircraft embedded information system is to process flight inputs (such as sensor and flight control) into usable flight information for further flight-system or aircrew utilization. The

6

AIRCRAFT COMPUTERS

embedded information system is a good example of the merging of two camps of computer science applications. The first, and larger, camp is the management of information systems (MIS). The MIS dealt primarily with large volumes of information, with primary applications in business and banking. The timing requirements of processing these large information records are measured in minutes or hours. The second camp is the real-time embedded computer camp, which was concerned with processing a much smaller set of data, but in a very timely fashion. The real-time camp’s timing requirement is in microseconds. These camps are now merging, because their requirements are converging. MIS increasingly needs real-time performance, while real-time systems are required to handle increased data-processing workloads. The embedded information system addresses both needs. Aircraft and the Year 2000. The year 2000 (Y2K) has been a major concern for the aircraft computer industry. Many of the embedded computers on aircraft and aircraft support functions are vulnerable to Y2K faults, because of their age. The basic problem with these computers has been that a year is represented by its low-order two digits. Instead of the year having four digits, these computers saved processing power by using the last two digits of the calendar year. For example, 1999 is represented as 99. This is not a problem until you reach the year 2000, represented as 00. Even with this representation, problems are limited to those algorithms sensitive to calendar dates. An obvious problem is when an algorithm divides by the calendar date, which is division by 0. Division by 0 is an illegal computer operation, causing problems such as infinite loops, execution termination, and system failure. The most commonly mentioned issue is the subtraction of dates to determine time durations and to compare dates. There problem is not that the computer programs fail in a very obvious way (e.g., divide-by-zero check) but, rather that the program computes an incorrect result without any warning or indication of error. Lefkon and Payne (11) discuss Y2K and how to make embedded computers compliant. Aircraft Application Program Interfaces. An application programming interface (API) is conventionally defined as an interface used by one program to make use of the services of another program. The human interface to a system is usually referred to as the user interface, or, less commonly, the human–computer interface. Application programs are software written to solve specific problems. For example, the embedded computer software that paints the artificial horizon on a heads-up display is an application program. A switch that turns the artificial horizon on or off is an API. Gal-Oz and Isaacs (12) discuss APIs and how to relieve bottlenecks of software debugging. Aircraft Control. Landau (1) defines, a control as an instrument or apparatus used to regulate a mechanism, or a device used to adjust or control a system. There are two concepts with control. One is the act of control. The other is the type of device used to enact control. An example of an act of control is when a pilot initiates changes to throttle and stick settings to alter flight path. The devices of control, in this case, are the throttle and stick. Control can be active or passive. Active control is force-sensitive. Passive control is displacement-sensitive. Mechanical control is the use of mechanical devices, such as levers or cams, to regulate a system. The earliest form of mechanical flight control was wires or cables, used to activate ailerons and stabilizers through pilot stick and foot pedal movements. Today, hydraulic control, the use of fluids for activation, is usual. Aircraft control surfaces are connected to stick and foot pedals through hydraulic lines. Pistons in the control surfaces are pushed or pulled by associated similar pistons in the stick or foot pedal. The control surfaces move accordingly. Electronic control is the use of electronic devices, such as motors or relays, to regulate a system. A motor is turned on by a switch, and quickly changes control surfaces by pulling or pushing a lever on the surface. Automatic control is a system-initiated control, which is a system-initiated response to a known set of environmental conditions. Automatic control was used for early versions of automatic pilot systems, which tied flight-control feedback systems to altitude and direction indicators. The pilot sets his desired course and altitude, which is maintained through the flight control’s automatic feedback system. To understand the need for computers in these control techniques, it is important to note the progression of the complexity of the techniques. The earliest techniques connected the pilot directly to his control surfaces.

AIRCRAFT COMPUTERS

7

As the aircraft functionality increased, the pilot’s workload also increased, requiring his (or his aircrew’s) being free to perform other duties. Additionally, flight characteristics became more complex, requiring more frequent and instantaneous control adjustments. The use of computers helped offset and balance the increased workload in aircraft. The application of computers to flight control provides a means for processing and responding to multiple complex flight control requirements. Aircraft Computer Hardware. For aircraft computers, hardware includes the processors, buses, and peripheral devices inputting to and outputting from the computers. Landau (1) defines hardware as apparatus used for controlling spacecraft; the mechanical, magnetic, and electronic design, structure, and devices of a computer; and the electronic or mechanical equipment that uses cassettes, disks, etc. The computers used on an aircraft are called processors. The processor takes inputs from peripheral devices and provides specific computational services for the aircraft. There are many types and functions of processors on aircraft. The most obvious processor is the central computer, also called the mission computer. The central computer provides direct control and display to the aircrew. The federated architecture (discussed in more detail later) is based on the central computer directing the scheduling and tasking of all the aircraft subsystems. Other noteworthy computers are the data-processing and signal-processing computers of the radar subsystem and the computer of the inertial navigation system. Processors are in almost every component of the aircraft. Through the use of an embedded processor, isolated components can perform independent functions as well as self-diagnostics. Distributed processors offer improved aircraft performance and, in some cases, redundant processing capability. Parallel processors are two or more processors configured to increase processing power by sharing tasks. The workload of the shared processing activity is distributed amongst the pooled processors to decrease the time it takes to form solutions. Usually, one of the processors acts as the lead processor, or master, while the other processor(s) act as slave(s). The master processor schedules the tasking and integrates the final results. On aircraft, this is particularly useful in that processors are distributed throughout the aircraft. Some of these computers can be configured to be parallel processors, offering improved performance and redundancy. Aircraft system redundancy is important, because it allows distributed parallel processors to be reconfigured when there is a system failure. Reconfigurable computers are processors that can be reprogrammed to perform different functions and activities. Before computers, it was very difficult to modify systems to adapt to their changing requirements. A reconfigurable computer can be dynamically reprogrammed to handle a critical situation, and than returned to its original configuration. Aircraft Buses. Buses are links between computers (processors), sensors, and related subsystems, for transferring data inputs and outputs. Fink and Christiansen (8) describe two primary buses as data buses and address buses. To complete the function of an MPU, a microprocessor must access memory and peripheral devices. This is accomplished by placing data on a bus, either an address bus or a data bus, depending upon the function of the operation. The standard 16-bit microprocessor requires a 16-line parallel bus for each function. An alternative is to multiplex the address or data bus to reduce the number of pin connections. Common buses in aircraft are the Military Standard 1553 Bus (Mil-Std-1553) and the General-Purpose Interface Bus (GPIB), which is the IEEE Standard 488 Bus. Aircraft Software. Landau (1) defines software as the programs, routines, etc. for a computer. The advent of software has provided great flexibility and adaptability to almost every aspect of life. This is especially true in all areas of aerospace sciences, where flight control, flight safety, in-flight entertainment, navigation, and communications are continuously being improved by software upgrades. Operation Flight Programs. An operational flight program (OFP) is the software of an aircraft embedded computer system. An OFP is associated with an aircraft’s primary flight processors, including the central computer, vertical and multiple display processors, data processors, signal processors, and warning receivers. Many OFPs in use today require dedicated software integrated support environments to upgrade and maintain them as the mission requirements of their parent aircraft are modified. The software integrated support environment [also called avionics integrated support environment (AISE), centralized software support activity

8

AIRCRAFT COMPUTERS

(CSSA), and software integration laboratory (SIL)] not only allows an OFP to be updated and maintained, but also provides capabilities to perform unit testing, subsystem testing, and some of the integrated system testing. Assembly Language. Assembly language is a machine (processor) language that represents inputs and outputs as digital data and that enables the machine to perform operations with those data. For a good understanding of the Motorola 6800 Assembler Language, refer to Bishop (13). According to Seidman and Flores (14) the lowest-level (closest to machine) language available to most computers is assembly language. When one writes a program in assembly code, alphanumeric characters are used instead of binary code. A special program called an assembler (provided with the machine) is designed to take the assembly statements and convert them to machine code. Assembly language is unique among programming languages in its one-to-one correspondence between the machine code statements produced by the assembler and the original assembly statements. In general, each line of assembly code assembles into one machine statement. Higher-Order Languages. Higher-order languages (HOLs) are computer languages that facilitate human language structures to perform machine-level functions. Seidman and Flores (14) discuss the level of discourse of a programming language as its distance from the underlying properties of the machine on which it is implemented. A low-level language is close to the machine, and hence provides access to its facilities almost directly; a high-level language is far from the machine, and hence insulated from the machine’s peculiarities. A language may provide both high-level and low-level constructs. Weakly typed languages are usually high-level, but often provide some way of calling low-level subroutines. Strongly typed languages are always high-level, and they provide means for defining entities that more closely match the real-world objects being modeled. Fortran is a low-level language that can be made to function as high-level by use of subroutines designed for the application. APL, Sobol, and SETL (a set-theoretic language) are high-level languages with fundamental data types that pervade their language. Pascal, Cobol, C, and PL/I are all relatively low-level languages, in which the correspondence between a program and the computations it causes to be executed is fairly obvious. Ada is an interesting example of a language with both low-level and high-level properties. Ada provides quite explicit mechanisms for specifying the layout of data structures in storage, for accessing particular machine locations, and even for communicating with machine interrupt routines, thus facilitating low-level requirements. Ada’s strong typing qualities, however, also qualify it as a high-level language. High-level languages have far more expressive power than low-level languages, and the modes of expression are well integrated into the language. One can write quite short programs that accomplish very complex operations. Gonzalez (15) developed an Ada Programmer’s Handbook that presents the terminology of the HOL Ada and examples of its use. He also highlights some of the common programmer errors and examples of those errors. Sodhi (16) discusses the advantages of using Ada. Important discussions of software life-cycle engineering and maintenance are presented, and the concept of configuration management is presented. The package concept is one of the most important developments to be found in modern programming languages, such as Ada, Modula-2, Turbo Pascal, C++, and Eiffel. The designers of the different languages have not agreed on what terms to use for this concept: package, module, unit, and class are commonly used. But it is generally agreed that the package (as in Ada) is the essential programming tool to be used for going beyond the programming of very simple class exercises to what is generally called software engineering, or building production systems. Packages and package like mechanisms are important tools used in software engineering to produce production systems. Feldman (17) illustrates the use of Ada packages to solve problems. Databases. Database are essential adjuncts to computer programming. Databases allow aircraft computer applications the ability to carry pertinent information (such as flight plans or navigation waypoints) into their missions, rather than generating them in route. Databases also allow the aircrew to collect performance information about the aircraft’s various subsystems, providing a capability to adjust the aircraft in flight and avoid system failures. Elmasri and Navathe (18) define a database as a collection of related data. Data are described as known facts that can be recorded and have implicit meaning. (A simple example consists of the names, telephone numbers, and addresses of an indexed address book. A database management system (DBMS) is a collection

AIRCRAFT COMPUTERS

9

Fig. 2. An aircraft avionics support bench.

of programs that enable users to create and maintain a database. The DBMS is hence a general-purpose software system that facilitates the processes of defining, constructing, and manipulating databases for various applications. Veriﬁcation and Validation. A significant portion of the aircraft computer’s life-cycle cost is system and software testing, performed in various combinations of unit-level, subsystem-level, integrated-system-level, developmental, and operational testing. These types of tests occur frequently throughout the life of an aircraft system because there are frequent upgrades and modifications to the aircraft and its various subsystems. It is possible to isolate acceptance testing to particular subsystems when minor changes are made, but this is the exception. Usually, any change made to a subsystem affects other multiple parts of the system. As aircraft become increasingly dependent on computers (which add complexity by the nature of their interdependences), and as their subsystems become increasingly integrated, the impact of change also increases drastically. Cook (19) shows that a promising technology to help understand the impact of aircraft computer change is the Advanced Avionics Verification and Validation (AAV&V) program developed by the Air Force Research Laboratory. Sommerville (20) develops the concepts of program verification and validation. Verification involves checking that the program conforms to its specification. Validation involves checking that the program as implemented meets the expectations of the user. Figure 2 shows an aircraft avionics support bench, which includes real components from the aircraft such as the FCC line replaceable unit (LRU) sitting on top of the pictured equipment. Additional equipment includes the buses, cooling, and power connection interfaces, along with monitoring and displays. On these types of benches, it is common to emulate system and subsystem responses with testing computers such as the single-board computers illustrated. Figure 3 shows another verification and validation asset called the workstation-based support environment. This environment allows an integrated view of the aircraft’s performance by providing simulations of the aircraft’s controls and displays on computer workstations. The simulation is interfaced with stick and throttle controls, vertical situation displays, and touch-screen avionics switch panels.

10

AIRCRAFT COMPUTERS

Fig. 3. A workstation-based aircraft avionics support environment.

Object-Oriented Technology. Object-oriented (OO) technology is one of the most popular computer topics of the 1990s. OO languages such as C++ and Ada 95 offer tremendous opportunities to capture complex representations of data and then save these representations in reusable objects. Instead of using several variables and interactions to describe some item or event, this same item or event is described as an object. The object contains its variables, control-flow representations, and data-flow representations. The object is a separable program unit, which can be reused, reengineered, and archived as a program unit. The power of this type of programming is that when large libraries of OO programming units are created, they can be called upon to greatly reduce the workload of computer software programming. Gabel (21) says that object-oriented technology lets an object (a software entity consisting of the data for an action and the associated action) be reused in different parts of the application, much as an engineered hardware product can use a standard type of resistor or microprocessor. Elmasri and Navathe (18) describe an object-oriented database as an approach with the flexibility to handle complex requirements without being limited by the data types and query languages available in traditional database systems. Open System Architecture. Open system architecture is a design methodology that keeps options for updating systems open by providing liberal interfacing standards. Ralston and Reilly (22) state that open architectures pertain primarily to personal computers. An open architecture is one that allows the installation of additional logic cards in the computer chassis beyond those used with the most primitive configuration of the system. The cards are inserted into slots in the computer’s motherboard—the main logic board that holds its CPU and memory chips. A computer vendor who adopts such a design knows that, since the characteristics of the motherboard will be public knowledge, other vendors who wish to do so can design and market customized logic cards. Open system architectures are increasingly important in modern aircraft applications, because of the constant need to upgrade these systems and utilize the latest technical innovations. It is extremely difficult to predict interconnection and growth requirements for next-generation aircraft. This is exactly what an open architecture attempts to avoid the need for. Client–Server Systems. A client–server system is one in which one computer provides services to another computer on a network. Ralston and Reilly (22) describe the file-server approached as an example of client-server interaction. Clients executing on the local machine forward all file requests (e.g. open, close,

AIRCRAFT COMPUTERS

11

read, write, and seek) to the remote file server. The server accepts a client’s requests, performs its associated operation, and returns a response to the client. Indeed, if the client software is structured transparently, the client need not even be aware that files being accessed physically reside on machines located elsewhere on the network. Client–server systems are being applied on modern aircraft, where highly distributed resources and their aircrew and passenger services are networked to application computers. Subsystems. The major subsystems of an aircraft are its airframe, power plant, avionics, landing gear, and controls. Landau (1) defines a subsystem as any system that is part of a larger system. Many of the subsystems on an aircraft have one or more processors associated with them. It is a complex task to isolate and test the assorted subsystems. Another layer of testing below subsystem testing is unit testing. A unit of a subsystem performs a function for it. For example, in the radar subsystem, the units include its signal processor and its data processor. In order to test a system adequately, each of its lowest-level items (units) must be tested. As the units affect and depend upon each other, another layer of testing addresses that layer of dependences. In the same fashion, subsystem testing is performed and integrated with associated subsystems. It is important to test not only at the unit and the subsystem level, but at the system and operational level. The system level is where the subsystems are brought together to offer the system functionality. System integration is the process of connecting subsystem components into greater levels of system functionality until the complete system is realized. The operational level of testing is where the subsystem is exercised in its actual use. Line Replaceable Units. LRUs are subsystems or subsystem components that are self-contained in durable boxes containing interface connections for data, control, and power. Many LRUs also contain built-in test (BIT) capabilities that notify air and maintenance crew when there is a failure. A powerful feature of LRUs is that functionality can be compartmentalized. When a failure is detected, the LRU can easily be pulled and replaced, restoring the aircraft to service within moments of detection. Graceful Degradation. All systems must have plans to address partial or catastrophic failure. System failure in flight controls is often catastrophic, while system failure in avionics can be recovered from. For this reason, most flight-critical systems have built-in redundant capabilities (sometimes multiple layers of redundancy), which are automatically activated when the main system or subsystem fails. Degraded system behavior occurs when the main system fails and backup systems are activated. The critical nature of system failure requires immediate activation of backup systems and recognition by all related subsystem of the new state of operation. Graceful degradation is the capability of aircraft computers to continue operating after incurring system failure. Graceful degradation is less than optimal performance, and may activate several layers of decreasing performance before the system fails. The value of graceful degradation is that the aircrew has time to respond to the system failure before there is a catastrophic failure.

Aerospace Computer technologies have helped provide a continuum of improvements in aircraft performance that has allowed the airspace where aircraft operate to increase in range and altitude. Landau (1) defines aerospace as the earth’s atmosphere and the space outside it, considered as one continuous field. Because of its rapidly increasing domain of air and space travel, the United States Air Force is beginning to refer to itself as the United Sates Aerospace Force. Modern air–space vehicles are becoming increasingly dependent on information gleaned from ground stations, satellites, other air–space vehicles, and onboard sensors to perform their mission. These vehicles use signals across the electromagnetic spectrum. Antennas can be found in multiple locations on wings, the fuselage, tails, and draglines. If antennas are located too close together, their signals can interfere with each other; this is called crossed frequency transmission. This interference reduces the efficiency of each affected antenna. Placement of multiple antennas requires minimizing the effects of crossed frequency transmissions. Techniques for this include antenna placement, filtering, and timing. This presents another

12

AIRCRAFT COMPUTERS

challenge for aircraft computers to sort and process these multiple signals. Perry and Geppert (4) show how the aircraft electromagnetic spectrum is becoming busy, and thus, dangerous for aerospace communications. Legacy Systems. Legacy systems are fielded aircraft, or aircraft that are in active use. Probably the only nonlegacy aircraft are experimental or prototype versions. Legacy aircraft are often associated with aging issues, more commonly known as parts obsolescence. A growing problem in these systems is the obsolescence of entire components, including the many computers used on them. Aircraft, like many other systems, are designed with expected lifetimes of 10 to 15 years. Because of the high replacement costs, lifetimes are often doubled and tripled by rebuilding and updating the aircraft. To reduce costs as many as possible of the original aircraft components are kept. Problems arise when these components are no longer produced or stockpiled. Sometimes subsystems and their interfaces have to be completely redesigned and produced at great cost in order to keep an aircraft in service. System architectures and standard interfaces are constantly being modified to address these issues. Aircraft evolve during their lifetimes to a more open architecture. This open architecture, in turn, allows the aircraft components to be more easily replaced, thus making further evolution less expensive. Unmanned Air Vehicles. Unmanned air vehicles (UAVs) are aircraft that are flown without aircrews. Their use is becoming increasingly popular for military applications. Many of the new capabilities of UAVs come from the improved computers. These computers allow the vehicles to have increased levels of autonomy and to perform missions that once required piloted aircraft. Some of these missions include reconnaissance and surveillance. These same types of missions are finding increasing commercial importance. UAVs offer tremendous advantages in life-cycle cost reductions because of their small size, ease of operation, and ability to be adapted to missions.

Man–Machine Systems An aircraft is an example of a man–machine system. Other examples are automobiles and boats. These machines have the common attribute of being driven by a human. Landau (1) defines man–machine systems as sets of manually performed and machine-performed functions, operated in conjunction to perform an operation. The aircraft computer is constantly changing the role of the human in the aircraft machine. The earliest aircraft required the constant attention of the pilot. Improved flight control devices allowed the pilot freedom for leisure or for other tasks. Modern aircraft computers have continued the trend of making the aircraft more the machine, and less the man system. Human Factors of Aircraft Computers. Human factors is the science of optimal conditions for human comfort and health in the human environment. The human factors of aircraft computers include the positioning of the controls and displays associated with the aircrew’s workloads. They also provide monitoring and adjustment of the aircraft human environment, including temperature, oxygen level, and cabin pressure. Man–Machine Interface. The man–machine interface is the place where man’s interactions with the aircraft coordinate with the machine functionality of the aircraft. An example of a man–machine interface is the API, which is where a person provides inputs to and receives outputs from computers. These types of interfaces include keyboards (with standard ASCII character representation), mouse pads, dials, switches, and many varieties of monitors. A significant interface in aircraft comprises their associated controls and displays, which provide access to the flight controls, the sensor suite, the environmental conditions, and the aircraft diagnostics through the aircraft’s central computer. Control sticks, buttons, switches, and displays are designed based on human standards and requirements such as seat height, lighting, accessibility, and ease of use. Voice–Activated Systems. Voice–activated systems are interfaces to aircraft controls that recognize and respond to aircrew’s verbal instructions. A voice-activated input provides multiple input possibilities beyond the limited capabilities of hands and feet. Voice–activated systems have specifed sets of word commands, and are trained to recognize a specific operator’s voice.

AIRCRAFT COMPUTERS

13

Aircraft Computer Visual Veriﬁcation. Visual verification is the process of physically verifying (through sight) the correct aircraft response to environmental stimuli. This visual verification is often a testing requirement. It is usually done through the acceptance test procedure (ATP) and visual inspections of displays through a checklist of system and subsystem inputs. Until recently, visual verification has been a requirement for pilots, who have desired the capability to see every possibility that their aircraft might encounter. This requirement is becoming increasingly difficult to implement, because of the growing complexity and workload of the aircraft’s computers and their associated controls and displays. In the late 1980s to early 1990s, it used to take about 2 weeks to visually verify the suite of an advanced fighter system’s avionics. This can no longer be accomplished at all with current verification and validation techniques. Several months would be required to achieve some level of confidence that today’s modern fighters are flight-safe. Air Trafﬁc Control. Air traffic control is the profession of monitoring and controlling aircraft traffic through an interconnected ground–based communication and radar system. Perry (23) describes the present capabilities and problems in air traffic control. He also discusses the future requirements for this very necessary public service. Air traffic controllers view sophisticated displays, which track multiple aircraft variables such as position, altitude, velocity, and heading. Air traffic control computers review these variables and give the controllers continuous knowledge of the status of each aircraft. These computers continuously update and display the aircraft in the ground–based radar range. When potential emergency situations, such as collision, arise, the computer highlights the involved aircraft on the displays, with plenty of lead time for the controller to correct each aircraft’s position.

Aircraft Control And Computers D’ Azzo and Houpis (24) give a good explanation of the complexity of what is needed for an aircraft control system. The feedback control system used to keep an airplane on a predetermined course or heading is necessary for the navigation of commercial airliners. Despite poor weather conditions and lack of visibility, the airplane must maintain a specified heading and altitude in order to reach its destination safely. In addition, in spite of rough air, the trip must be made as smooth and comfortable as possible for the passengers and crew. The problem is considerably complicated by the fact that the airplane has six degrees of freedom. This makes control more difficult than control of a ship, whose motion is limited to the surface of the water. A flight controller is used to control aircraft motion. Two typical signals to the system are the correct flight path, which is set by the pilot, and the level position of the airplane. The ultimately controlled variable is the actual course and position of the airplane. The output of the control system, the controlled variable, is the aircraft heading. In conventional aircraft there are three primary control surfaces used to control the physical threedimensional attitude of the airplane: the elevators, rudder, and ailerons. A directional gyroscope is used as the error-measuring device. Two gyros must be used to provide control of both heading and attitude of the airplane. The error that appears in the gyro as an angular displacement between the rotor and case is translated into a voltage by various methods, including the use of transducers such as potentiometers, synchros, transformers, or microsyns. Selection of the method used depends on the preference of the gyro manufacturer and the sensitivity required. Additional stabilization for the aircraft can be provided in the control system by rate feedback. In other words, in addition to the primary feedback, which is the position of the airplane, another signal proportional to the angular rate of rotation of the airplane around the vertical axis is fed back in order to achieve a stable response. A rate gyro is used to supply this signal. This additional stabilization may be absolutely necessary for some of the newer high-speed aircraft. In reading through this example, it should be obvious that as the complexity of the control feedback system of the aircraft increases, there is a need for computer processing to evaluate the feedback and to adjust

14

AIRCRAFT COMPUTERS

or recommend flight control adjustments. Additional feedback may come from global positioning, from groundbased navigation systems through radio inputs, and from other aircraft. The computer is able to integrate these inputs into the onboard flight control inputs, and provide improved recommendations for stable flight.

Real-Time Systems The computers on aircraft are required to perform their functions within short times. Flight control systems must make fine adjustments quickly, in order to maintain stable flight. Sensor suites must detect and analyze potential threats before it is too late. Cabin pressure and oxygen must be regulated as altitude changes. All these activities, plus many others on aircraft, must happen in real time. Nielsen (25) defines a real-time system as a controlled (by software or firmware) system that performs all of its process functions within specified time constraints. A real-time system usually includes a set of independent hardware devices that operate at widely differing speeds. These devices must be controlled so that the system as a whole is not dependent upon the speed of the slowest device. Hatley and Pirbhai (26) describe timing as one of the most critical aspects of modern real-time systems. Often, the system’s response must occur within milliseconds of a given input event, and every second it must respond to many such events in many different ways. Flight-Critical Systems. Flight-critical systems are those activities of an aircraft that must be completed without error in order to maintain life and flight. The aircraft flight controls, engines, landing gear, and cabin environment are examples of flight-critical systems. Failures in any of these systems can have catastrophic results. Flight-critical systems are held to tight levels of performance expectations, and often have redundant backups in case of failure. Federated Systems. Federated systems are loosely coupled distributed systems frequently used in aircraft system architectures to tie multiple processors in multiple subsystems together. The loose coupling allows the multiple subsystems to operate somewhat autonomously, but have the advantage of the shared resources of the other subsystems. A typical aircraft federated system might include its central computer, its INS, its radar system, and its air-vehicle management system. The INS provides the radar with the aircraft’s present position, which is reported to the pilot through displays put forth by the central computer. The pilot adjusts his course through the air-vehicle management system, which is updated by the INS, and the cycle is repeated. These subsystems perform their individual functionality while providing services to each other. Cyclic Executive. A cyclic executive on an aircraft computer provides a means to schedule and prioritize all the functions of the computer. The executive routine assigns the functions and operations to be performed by the computer. These assignments are given a specific amount of clock time to be performed. If the assignment does not complete its task in its allocated time, it is held in a wait state until its next clock period. From the beginning of the clock period to its end is one clock cycle. High-priority functions are assigned faster clock cycles, while low-priority functions are assigned slower cycles. For example, the high-priority executive function might be assigned a speed of 100 cycles per second, while some lower-priority function might have 5 cycles per second to complete its tasks. Sometimes the latter might take several clock cycles to perform a task. An additional feature of cyclic executives is that they are equipped with interrupts, which allow higher-priority systems to break into the executive assignments for system-level assigned tasking. There are several types of scheduling methodologies that provide performance improvements in cyclic executives. One of the more prominent is rate monotonic analysis (RMA), which determines the time requirement for each function and the spare time slots, and then makes time assignments.

AIRCRAFT COMPUTERS

15

BIBLIOGRAPHY 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26.

S. Landou Webster Illustrated Contemporary Dictionary, Encyclopedic Edition, Chicago: J. G. Ferguson, 1992. J. F. Wakerly, Digital Design Principles and Practices, Englewood Cliffs, NJ: Prentice-Hall, 1985, pp. 1–48, 53–138. V. C. Hamacher Z. G. Vranesic S. G. Zaky Computer Organization, 2nd ed., New York: McGraw-Hill, 1984. T. Perry L. Geppert Do portable electronics endanger flight, IEEE Spectrum, 33 (9): 26–33, 1996. A. Golden Radar Electronic Warfare, Washington: AIAA Education Series, 1987. G. W. Stimson Introduction to Airborne Radar, El Segundo, CA: Hughes Aircraft, 1983, pp. 107, 151–231. G. Welch G. Bishop An introduction to the Kalman filter, Department of Computer Science, University of North Carolina at Chapel Hill, Chapel Hill, NC, http://www.cs.unc.edu/˜welch/media/pdf/kalman.pdf, 1997. D. Fink D. Christiansen Electronics Engineers’ Handbook, 3rd ed., New York: McGraw-Hill, 1989. J. Adam T. Gibson Warfare in the information age, IEEE Spectrum, 28 (9): 26–42. W. Sweet The glass cockpit, IEEE Spectrum, 32 (9): 30–38, 1995. D. Lefkon B. Payne Making embedded systems year 2000 compliant, IEEE Spectrum, 35 (6): 74–79, 1998. S. Gal-Oz M. Isaacs Automate the bottleneck in embedded system design, IEEE Spectrum, 35 (8): 62–67, 1998. R. Bishop Basic Microprocessors and the 6800, Hasbrouck Heights, NJ: Hayden, 1979. A. Seidman I. Flores The Handbook of Computers and Computing, New York: Van Norstrand Reinhold, 1984, pp. 327–502. D. W. Gonzalez Ada Programmer’s Handbook, Redwood City, CA: Benjamin/Cummings, 1991. J. Sodhi Managing Ada Projects, Blue Ridge Summit, PA: TAB Books, 1990. M. B. Feldman E. B. Koffman Ada Problem Solving and Program Design, Reading, MA: Addison-Wesley, 1992. R. Elmasri S. B. Navathe Fundamentals of Database Design, 2nd ed., Redwood City, CA: Benjamin/Cummings, 1994. R. Cook The advanced avionics verification and validation II final report, Air Force Research Laboratory Technical Report ASC-99-2078, Wright-Patterson AFB. I. Sommerville Software Engineering, 3rd ed., Reading, MA: Addison-Wesley, 1989. D. Gabel Software engineering, IEEE Spectrum, Vol. 31 (1): 38–41, 1994. A. Ralston E. Reilly Encyclopedia of Computer Science, New York: Van Nostrand Reinhold, 1993. T. Perry In search of the future of air traffic control, IEEE Spectrum, 34 (8): 18–35, 1997. J. J. D’ Azzo C. H. Houpis Linear Control System Analysis and Design, 2nd ed., New York: McGraw-Hill, 1981, pp. 143–146. K. Nielsen Ada in Distributed Real-Time Systems, New York: Intertext, 1990. D. J. Hatley I. A. Pirbhai Strategies for Real-Time System Specification, New York: Dorset House, 1988.

READING LIST G. Buttazo Hard Real-Time Computing Systems, Norwell, MA: Kluwer, 1997. R. Comerford PCs and workstations, IEEE Spectrum, 30, (1): 26–29, 1993. D. Dooling Aerospace and military, IEEE Spectrum, 35 (1): 90–94, 1998. J. Juliussen D. Dooling Small computers, aerospace & military, IEEE Spectrum, 32 (1): 44–47, 76–79, 1995. K. Kavi Real-Time Systems, Abstractions, Languages, and Design Methodologies, Los Alamitos, CA: IEEE Computer Society Press, 1992. P. Laplante Real-Time Systems Design and Analysis, an Engineer’s Handbook, Piscataway, NJ: IEEE Press, 1997. M. S. Roden Analog and Digital Communication Systems, 2nd ed., Englewood Cliffs, NJ: Prentice-Hall, 1985. H. Taub Digital Circuits and Microprocessors, New York: McGraw-Hill, 1982. C. Weitzman Distributed Micro/Minicomputer, Englewood Cliffs, NJ: Prentice-Hall, 1980.

CHARLES P. SATTERTHWAITE Air Force Research Laboratory Embedded Information System Engineering Branch (AFRL IFTA)

26

JET TRANSPORT MAINTENANCE

JET TRANSPORT MAINTENANCE As certain items of an aircraft and its systems deteriorate, it is necessary to ensure that the design remains airworthy. Maintenance is the action necessary to sustain or restore the integrity and performance of the aircraft. It includes inspection, overhaul, repair, preservation, and replacement of parts such as to ensure conformity to the basis of certification. This is called continuing airworthiness. The preceding is an admirable definition, but hardly a complete one. It merely says that maintenance is needed to mend aircraft and to keep them airworthy. Maintenance organizations do more. The maintenance department at any airline must also sustain equipment availability, that is, the ability of the aircraft to fly the published revenue schedule. The definition of maintenance must also include ‘‘the management of failure.’’ The performance of maintenance on an aircraft restores safety and reliability levels when deterioration or damage has occurred. It does not include servicing, which is the replenishment of consumables needed to keep an item or aircraft in operating condition (e.g., cleaning, fueling, catering, fluid replenishment, and lavatory service, etc.). J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

JET TRANSPORT MAINTENANCE

Maintenance is performed upon aircraft by the airlines to protect their investment, maximize safety, minimize schedule disruptions, and comply with governmental regulations. In the United States, the Federal Aviation Regulations (FARs), arising from federal law require that: • ‘‘The owner or operator, is primarily responsible for maintaining (the) aircraft in an airworthy condition. . . .’’ • ‘‘No person may operate an aircraft . . . unless the mandatory replacement times, inspection intervals and related procedures . . . have been complied with.’’ • ‘‘Each certificate holder (airline) is primarily responsible for . . . the airworthiness of its aircraft . . . the performance of maintenance . . . in accordance with its manual and the regulations . . .’’ Maintenance is not free; it accounts for approximately 10% of an airline’s employees and 10% to 15% of its operating expenses. It ensures the aircraft is safe and available to fly the revenue schedule, thus enhancing schedule reliability by reducing both the frequency and duration of flight delays and cancellations. Admittedly, technical delays and cancellations do not constitute a dominant expense compared with most elements of direct operating cost (i.e., those costs associated with flight crew, fuel, airframe and engine maintenance, insurance, depreciation, and interest). But loss of service is unacceptable. It is poor business practice. The direct economic consequences are major to both the airlines and their customers. According to Federal Aviation Administration (FAA) data available for 1994, the calculated annual total operating delay cost to the airlines was $2.5 billion and $7.0 billion attributable to passengers. The average duration of a delay was 14.23 minutes. Reported in The Handbook of Airline Economics, 1995, ‘‘An average of 0.1% to 0.2% of a typical airlines flights will be interrupted due to maintenance problems. As many as 5% to 10% of all flights could experience cancellations or delays due to maintenance problems.’’ The cost to the airline includes: • • • •

Loss of ticket revenue Poor customer relationships Increased spares inventories Increased numbers of maintenance stations requiring skills and labor • Costs arising from reroutes, equipment substitution, passenger handling (hotels, buses, meal vouchers, and so on), substitute service, disruption to the aircraft movement plan. The list is endless.

The cost to the customer takes the form of disrupted plans, missed business appointments, lost time, late shipments, and other delays. It is expected that people and cargo will go to their destination on time. FAILURE The primary consideration of all maintenance decisions is neither the failure of a given item nor the frequency of its occur-

27

rence, but rather the consequences of that failure upon the aircraft and its operation. Consequences of failure are twofold: (1) those affecting safety and (2) those affecting availability of equipment (economic). Safety Related. Failure that jeopardizes the safety of the aircraft or its occupants must be prevented. Aircraft cannot be of such design that any single failure of the device will have catastrophic results. This is aeronautical dogma. Today’s aircraft are subject to very few single critical failure modes. This safety related reliability is attributed to the design requirements of the relevant governmental regulations as well as the specifications of operating organizations and manufacturers. Current design practice ensures that vital functions are protected by redundancy, fault tolerance, fail tolerance, and fail safe features. This assures that, if there is failure, a given function will remain available from other sources to insure a safe completion of flight. It should be noted there are some safety considerations which create an exception to this single failure concept and which require, at least in practice, the accountability for a second failure (i.e., a second failure or combination of failures may necessitate even further redundencies). Economic. If the loss or deterioration of a particular function neither endangers the equipment nor its occupants, then the consequences of that failure are economic. Examples include systems, components, or features in a design that are not specifically required to demonstrate conformity to the basis of certification (i.e., aircraft is in a configuration that has been shown to meet all design certification requirements). For example, a fuel system might require two pumps to meet design certification requirements. An extra (third) fuel pump might be added to the design of the fuel system, solely for economic reasons (e.g., to decrease the frequency and risk of flight delays or cancellations caused by pump failures). FAILURE MANAGEMENT Safety related failure can be managed. Consider that if the design only addresses the avoidance of single catastrophic failures, the aircraft and its occupants will not be placed in peril. But single failures of components or systems will cause the loss of availability of the equipment once the aircraft lands. For once a single failure occurs, a ‘‘no-go’’ condition arises until repair or replacement is accomplished. There are three failure management strategies: 1. The components and systems are designed to an exceptional degree of reliability. This is an inordinately costly strategy. Cost effective design trades must be made between the loss of availability arising from ‘‘no-go’’ situations and the cost of exceptionally reliable components. 2. If a high degree of reliability is not cost effective, then the design could instead include a high degree of deferability, that is, a good minimum equipment list (MEL). The MEL is the list specifying the minimum equipment required for flight dispatch. The concept of a MEL arose out of the excess capability in the design that just ‘‘happened.’’ Traditionally, all installed equipment specified by the airworthiness and operating regulations must be operative. However, experience indicates that, with varying levels of redundancy designed into aircraft, opera-

28

JET TRANSPORT MAINTENANCE

tion of every system or installed component is not necessary when the remaining operative equipment provides an acceptable level of safety. This was recognized in the mid 1950s. Consequently, regulatory agencies granted permission to operate with certain items of equipment inoperative; the intent being to permit revenue operations to a location where repairs or replacements could be made. This action permits economic aircraft utilization as well as offering a reliable flight schedule to the flying public without compromising flight safety. Contemporary practice demands that consideration be given to deferability in the design as a conscious activity when defining system architecture and functionality. It should be noted that even with a MEL, ‘‘no go’’ conditions will not be totally eliminated. 3. A third strategy assures that ‘‘no-go’’ conditions can be minimized. It involves both a design and a maintenance management technique. This design approach embraces the incorporation of features that are extra to those required for certification. The predominant strategy for this is the same as that used to avoid safety related failures; that is the inclusion of redundancy, fault tolerance and fail safe, fail passive features but beyond that required to certify the design. This is not without its price. It increases the number of failure possibilities. It adds more items that can fail. It results in equipment that is more complex and integrated which makes fault isolation more difficult. It adds to the cost of the aircraft. But this approach, judiciously applied, greatly reduces the consequences of any single failure. Excess features in the design put initial failures of a system into the economic rather than the safety related failure category. AIR CARRIER MAINTENANCE REQUIREMENTS Maintenance requirements are dictated by numerous factors: regulatory provisions, type of equipment, fleet size, route structure, and flying schedules. The type of equipment establishes maintenance frequency cycles. The size of the fleet determines quantitative maintenance work load. Route structure and flight schedules influence the location and number of stations which must possess the capability of performing the work.

to the Airworthiness Limitations which set forth each mandatory replacement time, structural inspection intervals, and related structural inspection procedures. In addition, they must include an inspection program that includes the frequency and extent of the inspections necessary to provide for the continued airworthiness of the airplane . . . . . . Accumulated flight hours, calendar time, number of operating cycles or the number of landings are the generally accepted measurements used when specifying maintenance intervals (i.e., periods). The selection of a specific parameter is dictated by the particular operating environment encountered . . .

Scheduled Maintenance Task Definition Aircraft, engine manufacturers, representatives of airlines (both foreign and domestic), FAA and foreign regulatory agency observers, develop the maintenance recommendations for a new design. They do this as members of a maintenance steering group (MSG). The structure, methodology, and composition of this recommending group is defined in the Air Transport Association (ATA) document titled Maintenance Program Development Document MSG-3, commonly referred to as MSG-3. MSG-3 uses a process called ‘‘decision tree’’ analysis. It employs a logic which is task rather than process oriented, designed to uncover hidden failures and to separate safety related from economic failure. It also includes characterizing servicing and lubrication tasks. MSG-3 consists of: • Defining maintenance significant items (MSI) • Analyzing the MSI • Recommending inspection and check tasks arising from this analysis • Defining scheduling information for the tasks • Preparing the maintenance/inspection recommendations into a Maintenance Requirements and Review Proposal document. The proposal document is submitted to the FAA by the manufacturer, in partial fulfillment of 14 CFR 25.1329. The FAA in turn convenes an internal maintenance review board (MRB) for review and approval of the document. The resulting FAA MRB report defines the scheduled maintenance tasks for the aircraft.

Regulatory Provisions The definition of maintenance requirements, addressing safety related failure for an aircraft, begins during design and certification. The Federal Aviation Regulations (FARs) are published in the Code of the Federal Regulations (CFR). 14 CFR 25.1329 requires the preparation of instructions for continuing airworthiness. These instructions must include, among other things, the following: . . . Scheduling information (scheduled maintenance) for each part of the airplane and its engines, auxiliary power units, propellers, accessories, instruments, and equipment that provides the recommended periods at which they should be cleaned, inspected, adjusted, tested, and lubricated, and the degree of inspection, the applicable wear tolerances, and work recommended at these periods. The recommended overhaul periods and necessary references

Defining Maintenance Significant Items. The MSG-3 process begins by defining which line replaceable units (LRU), system installations, and items of aircraft structure are sufficiently significant in their maintenance requirements to justify special consideration in design to assure safety and optimum maintainability, and for establishing required maintenance inspections/checks. This results in a list called maintenance significant items to be analyzed for relevance to the following: • Safety related items. Any system or component malfunction which results in the loss of airworthiness is by definition safety related. • Potential economic impacts. These address such issues as:

JET TRANSPORT MAINTENANCE

High initial design, manufacturing and ownership cost High maintenance cost Premature removal rates Significant access problems Potential for mechanical dispatch delays Significant weight increase with consequent reduced aircraft performance or increased fuel burn System or component redundancies in excess of that required for airworthiness Components thus selected have priority in design, toward improving operational performance, reducing maintenance requirements, and enhancing their maintainability to lessen the maintenance cost and/or departure delays. Maintenance Processes. There are three recognized processes in use to define maintenance check intervals: (1) Hard time, (2) on condition, and (3) condition monitoring. Hard Time. This process applies a fixed period to the component which is the maximum period the component can be continued in service without overhaul or replacement. It is similar to fixed time between overhaul (TBO) and defines the maximum time and/or cycles that an item is permitted to operate on an aircraft between overhauls. Time typically relates to operating flight time or in some instances elapsed calendar time. Cycles relate to operating cycles (e.g., takeoff and landings or the number of takeoff thrust applications). Overhaul means restoration of an item to its service life in accordance with the instructions defined in relevant manuals. Hard time maintenance should be avoided. It is very costly. Deterioration and failure are not directly related to time. Studies have shown that approximately 90% of the components in an aircraft derive no benefit from using hard time. Items selected for hard time should be limited to: • components or assemblies which have definite life limits (e.g., metal fatigue) or, • whose failure would have a direct adverse effect upon airworthiness if they malfunctioned in flight or, • component deterioration which is not directly observable or easily measured. An example of a hard time component is the landing gear. Premature failure could have deleterious effects. Also landing gear forgings are subject to fatigue which is not directly measurable. On Condition. This is a process in which the component’s operational condition, as determined by some form of test or check, dictates its time of replacement. It consists of repetitive inspections or tests to determine the condition of units, systems, or portions of structure with regard to continued serviceability. It accepts operation of a component or system until failure (i.e., failure being the inability to perform within specified limits or failing to perform intended function). Its application is therefore limited to items whose failure during aircraft operation will not have catastrophic consequences. Items and appliances listed as on condition must be restricted to components on which a determination of continued airworthiness may be made by visual inspection, measurements, tests, or other means without a tear down inspection

29

or overhaul. These on condition checks are to be performed within the time limitations prescribed for the inspection or check. Performance tolerances and wear or deterioration limits are defined in the instructions for continuing airworthiness. On condition maintenance can involve removal or bench test and is thus not restricted to on-aircraft inspections, although on-aircraft inspection/tests are preferred. Condition Monitoring. This process is based upon reliability centered techniques. It is a refinement of on condition. The process applies to items that show deterioration over time. It consists of monitoring deterioration of a given component or system as it trends toward failure. The process is rooted upon: • An effective data collection system • A system for effectively assessing the need for changes in maintenance interval or design and for taking appropriate action. Action consists of appropriate reviews of the following: Actuarial or engineering studies employed to determine need for maintenance program changes Actual maintenance program changes involving inspection frequency and content, functional checks, overhaul limits, and times Aircraft, aircraft system or component modification, or repair Other actions peculiar to the operating conditions that prevail. Preparation of appropriate reports The airlines use appropriate items from the following performance data as the basic data collection elements of the program: • Unscheduled removals • Confirmed failures • Deficiencies observed and corrected but not otherwise reportable • Pilot reports • Sampling inspection • Functional checks • Shop findings • Bench checks • Mechanical reliability reports • Mechanical interruption summary reports Condition monitoring has significant advantages: • Maximum economic, safe utilization of equipment to its airworthiness limits is assured • Accurate identification of incipient failures is possible, thereby allowing economical repair before the occurrence of extensive costly damage; is most beneficial with high cost items such as engine components • Better spares inventory control results The principal disadvantage of condition monitoring is the sizable data collection and analysis requirement imposed upon the airlines.

30

JET TRANSPORT MAINTENANCE

SCHEDULED MAINTENANCE Scheduled maintenance (sometimes referred to as routine or recurrent maintenance) includes: (1) the mandatory tasks defined by the FAA Maintenance Review Board (MRB) Report, (2) the accomplishment of recurring airworthiness directives (ADs), and (3) discretionary (economic) checks, inspections or modifications. The FAA issues ADs when an unsafe condition has been found to exist in particular aircraft, engine, propellers, or appliances installed on aircraft, and that condition is likely to exist or develop in other aircraft, engines, propellers, or appliances of the same type design. Once an AD is issued, no person may operate an aircraft to which the AD applies except in accordance with the requirements of that AD. Discretionary maintenance tasks are those not required by the MRB report. They include for example: • Repair of items not related to airworthiness that is, economic failures • Modifications to cabin interiors such as installing passenger entertainment or refurbishing seats • Exterior painting or refurbishment • Manufacturer’s service bulletins not related to an airworthiness directive Packaging Scheduled Maintenance Scheduled maintenance requirements are grouped into work packages known as blocks. The principle of blocks is to accomplish all of the mandatory tasks in small packages. This allows greater utilization of the aircraft since the aircraft is removed from service for short periods rather than for a single extended overhaul period. The principle is shown in Fig. 1. Regardless of the means in which the tasks are packaged, all of the required tasks defined by the MRB will be accomplished when all of the defined blocks have been accomplished. The complete package of defined blocks is sometimes referred to as a ‘‘complete overhaul cycle.’’ Blocks have numerous names within the maintenance community. The exact nomenclature, composition and number of blocks varies between airlines. The following typical groupings illustrate the concept. Daily Check. This exists under several common names; post flight, maintenance pre-flight, service check, overnight to name a few. It is the lowest scheduled check. It is a cursory inspection of the aircraft to look for obvious damage and deterioration. It checks for ‘‘general condition and security’’ and reviews the aircraft log for discrepancies and corrective action. The accomplishment of the daily check requires little specific equipment, tools, or facilities. It is a basic requirement that the aircraft remains airworthy. Usually this check will be accomplished every 24–60 hours of accumulated flight time. Examples of daily check items include: • • • •

Visually inspect tail skid shock strut pop-up indicator Check fluid levels Check general security and cleanliness of the flight deck Check emergency equipment is installed

A Check. This is the next higher level of scheduled maintenance. It is normally accomplished at a designated mainte-

nance station in the route structure. It includes the opening of access panels to check and service certain items. Some limited special tooling, servicing and test equipment is required. The A check includes the lower check, the daily check. Examples of A check items include: • General external visual inspection of aircraft structure for evidence of damage, deformation, corrosion, missing parts • Check crew oxygen system pressure • Operationally check emergency lights • Lubricate nose gear retract actuator • Check parking brake accumulator pressure • Perform tests of certain systems using the built-in-testequipment (BITE) (See BITE later). B Check. This is a slightly more detailed check of components/systems. It does not involve, however, detailed disassembly or removal of components. Contemporary maintenance programs do not use the B check interval. For a number of reasons, the tasks formerly defined for this interval have, for many aircraft, been distributed between the A and C check. C and D Checks. The following two checks are traditionally known as heavy checks. They are normally accomplished at a main maintenance base of the airline where specialized manpower, materials, tooling, and hangar facilities are available. The aircraft will usually be removed from the revenue schedule for several days (3 to 20 days) while these checks are performed. See phase checks later in this article for a description of exceptions. C Check. This is an extensive check of individual systems and components for serviceability and function. It requires a thorough visual inspection of specified areas, components, and systems as well as operational or functional checks. It is a high level check which involves extensive tooling, test equipment, and special skill levels. The C check includes the lower checks, that is, daily, A, and B checks. Examples of C check items include: • Visually check flight compartment escape ropes for condition and security • Check operation of dc bus tie control unit • Visually check the condition of entry door seals • Operationally check flap asymmetry system • Pressure decay check Auxiliary Power Unit (APU) fuel line shroud • Inspect engine inlet thermal anti-ice (TAI) ducting for cracks • Operationally check Ram Air Turbine (RAT) deployment and system D Check. This can also be known as the structural check. It includes detailed visual and other nondestructive test inspections of the aircraft structure. It is an intense inspection of the structure for evidence of corrosion, structural deformation, cracking, and other signs of deterioration or distress. Structural checks involve extensive disassembly to gain access for inspection. Structural checks are man-hour and cal-

JET TRANSPORT MAINTENANCE

31

Flight time (hours)

10

80

400

1,600

16,000

Check level

Preflight

A check + Preflight

B check + A check + Preflight

C check + B check + A check + Preflight

D check + C check + B check + A check + Preflight

Check type

Number in cycle

Man-hours

A/C daily utilization

Flying days per year

Preflight

1,600

2

8

240

A

200

B

Approximate Out of check service time occurrence per check

Daily

1 hour

8

2/month

1 shift

40

36

3/year

1–1.5 shifts

C

10

450

1/year

10–12 shifts

D

1

1,500

8 years

15–18 shifts

Remarks: 1. The higher check always includes the lower check. 2. Block maintenance addresses inspections of the airframe and installed systems. 3. Individual component maintenance is not included. 4. Repair or replacement arising from inspections is not included. 5. A – Quick opening doors, servicing, detail walkaround. 6. B – Cowl, access panels, compartment doors opened lubrication, filter changes, operational checks. 7. C – Major access panels and fairings removed, system test, corrosion control, lubrication. 8. C – Major structural inspections, NDT work, internal structure.

Figure 1. Block maintenance.

endar time intensive. The D check includes the lower checks, A, B, C, and daily checks. Examples of D check items include:

Example • The size of the B and C checks has become too large • Divide the check into parts • Allocate the resultant parts or “segments” • Append them to the A and B checks

• Inspect stabilizer attach bolts • Inspect floor beams • Detailed inspection of wing box structure

C

Variations of Scheduled Maintenance The number of scheduled maintenance tasks for a large aircraft like the 747 are extensive. This is particularly true for the higher C and D checks. Their accomplishment removes the aircraft from service for several weeks. This is considered unacceptable and defeats the concept of removing the aircraft from service in small manageable blocks. A solution is to divide these higher checks into segmented blocks or phases. Such phasing levels the workload as well. This is shown conceptually in Fig. 2. A typical phase check provides for a thorough visual inspection of specified areas, components and systems as well as operational or functional checks of specified components

B

A

A Existing

C/3

2C/3

3C/3

C/3

2B/3

3B/3

A

A

A

B

A

Segmented

Figure 2. ‘‘Segmented’’ block maintenance.

32

JET TRANSPORT MAINTENANCE

and systems. Each check includes the requirements of traditional lower check work items and portions of C and D checks at the required task intervals. Phased checks may occur at 200 to 800 flight hour intervals, depending upon the work packaging plan and other airline operating variables. Changing Scheduled Maintenance Frequencies Individual airlines, when first placing a given aircraft model into service, use the aircraft MRB document for defining maintenance tasks and intervals. However, as experience is gained on the equipment, and advanced techniques are developed for flight and maintenance operations, the FAA allows for escalation of task intervals. Actuarial techniques, using condition monitoring data, are employed by the airlines to petition the FAA for a change in specified intervals. UNSCHEDULED MAINTENANCE Unscheduled maintenance (nonroutine, nonrecurrent) is ad hoc. It is maintenance performed to restore an item to airworthiness by correction of known or suspected malfunction and/ or defect. The resolution of aircraft malfunctions and/or defects is not always straightforward and often requires troubleshooting. Figure 3 shows a typical process that an airline might follow to troubleshoot an aircraft problem. Examples of unscheduled maintenance include: • Resolution of aircraft log discrepancies (both pilot generated and those discovered by the mechanic) • Special inspections initiated by the airline engineering group • Special inspections, repairs, or replacements arising from airworthiness directives (ADs) • Structural repairs arising from damage incurred during operations The nature of unscheduled maintenance dictates that it may be performed anywhere within the maintenance environment, that is, during scheduled maintenance or on the flight line while the aircraft is assigned to scheduled revenue service. THE MAINTENANCE ENVIRONMENT For clarity the maintenance environment is divided into three distinct categories of activity. However, in day to day operations this separation is blurred. Work normally accomplished while the aircraft is removed from the revenue schedule may occasionally be accomplished while the aircraft is flying the schedule. Line Maintenance Line maintenance is that maintenance activity performed while the aircraft is committed to the revenue schedule. It may be subdivided into Gate or Turnaround. Gate Maintenance. This maintenance is performed prior to the aircraft departure. It is incidental to flight operations. The flight line (gate) environment is the most demanding. It

is flight schedule driven. Time available is normally limited, usually 40 to 60 min, but may be as low as 20 min. Equipment and manpower are also limited. It consists of a visual check of the aircraft exterior with particular attention to indications of fluid leaks, obvious discrepancies such as worn or flat tires, low shock struts, fuselage or wing damage, and a review of aircraft log discrepancies. As a minimum, malfunctions affecting airworthiness are either repaired or deferred under the minimum equipment list (MEL). Turnaround Maintenance. This is performed at terminating locations in the flight schedule. Time available may be 8 to 16 h or more. It can also be known as overnight maintenance. The work usually consists of a daily check. Log book discrepancies including outstanding MEL deferrals are corrected. Passenger service equipment discrepancies are corrected. Servicing is accomplished. Additionally scheduled maintenance (e.g., portions of a phased check) may be performed at stations having a long turnaround. Depending upon time and manpower availability, discretionary tasks may be included. Hangar Maintenance Hangar maintenance is that activity normally affiliated with work on the aircraft when it is removed from the revenue schedule. It is predominately, though not exclusively, associated with the heavy checks (C and D) of scheduled maintenance, the incorporation of aircraft alterations, or structural repairs. Shop Maintenance Sometimes referred to as bench maintenance, it consists of repair, overhaul or refurbishment of LRUs or major assemblies (e.g., powerplants) which have been removed from the aircraft.

AIRPLANE SYSTEM DESIGN FOR MAINTENANCE Jet transport designs incorporate many features and design considerations to support the maintenance and troubleshooting of the aircraft and its many systems. The jet transport aircraft contains approximately 80 different systems providing a wide range of functions such as air conditioning, communications, electrical power, flight controls, hydraulics, navigation, pneumatics, and propulsion. In general, a system consists of a number of sensors, one or more computers that use signals from the sensors, and, as applicable, one or more actuators, pumps, valves, relays, or other devices that are controlled by the computer(s). For example, Fig. 4 shows the basic elements that make up the air data inertial reference system on the Boeing 737-700. The air data sensors are the pitot probes, the total air temperature (TAT) probe, the angle of attack (AOA) sensors, and the static ports. The air data inertial reference units (ADIRU) receive and monitor signals from these sensors, process this data, and transmit resulting signals (such as barometric altitude, airspeed, and mach number) to other systems on the aircraft. On jet transport aircraft, most system controllers are located in the aircraft’s equipment racks, which are typically located throughout the aircraft. Some system controllers, such as an

33

Research the problem on ground

Yes

Is advance information available ?

B

Yes

A

No

Yes

Fix the problem

Yes

Is a simple fix apparent ?

No

No

Figure 3. Typical airline troubleshooting process.

•Troubleshoot the problem •Determine corrective action

Off the gate troubleshooting (delay/cancellation)

•Sign-off logbook •Enter report in maintenance history, or •Action per airline policy

No

Does the problem exist ?

Obtain parts, equipment, and necessary materials

A

Mechanic attempts to verify or duplicate the problem

•Do the corrective action •Verify problem correction

Mechanic •Meets airplane •Checks logbooks •Evaluates problem

Does the problem continue ?

No

Yes

Can a fix be found with detailed troubleshooting on the gate ?

Perform initial assesment of the problem

Can problem be deferred per the MEL ? No

B

Yes

Yes

Can a fix be found with detailed troubleshooting on the gate ?

Perform initial assesment of the problem

Yes

Is there time to work on the problem ?

No

No

Defer maintenance per MEL

A

34

JET TRANSPORT MAINTENANCE

AOA sensor FO pitot probe

TAT probe

ADM

ADIRU-R

ADM

ADIRU-R

ADM

To user systems Figure 4. Basic elements of the Boeing 737-700 air data inertial reference system. Note: FO - first officer, ADM - Air Data Module, TAT - Total Air Temperature, Alt - alternate, AOA - angle of attack, stby - standby, inst - instrument, press - pressurization.

ADM To stby airspeed

Capt pitot probe Alt pitot probe

electronic engine controller, are often located with the equipment they control (e.g., on the engines). Figure 5 shows the equipment racks and their locations on a Boeing 777. Figure 6 shows the system controllers that are located on the E1 and E2 racks, located in the main equipment center, on a Boeing 777. Aircraft System Communication Aircraft systems use a variety of means for communication. Early designs relied almost entirely on analog signals. More recent designs make extensive use of digital data buses of increasing sophistication. Many of the digital data buses used on jet transport aircraft are specified in documents developed by Aeronautical Radio Incorporated (ARINC). ARINC is a corporation initiated and supported by airlines to provide technical services to the airlines. ARINC has developed a wide variety of standards and guidance documents for aircraft systems. Two principal standards for communication between aircraft systems are: ARINC 429. Until the mid 1970s, communication between aircraft systems was almost entirely accomplished using analog signals, for which a separate wire, or pair of wires, was

E7 rack E17 rack E11 rack E12 rack E10 rack E16 rack

E15 rack E16 rack

E5 rack E2/E3 Main equipment center Forward equipment center Figure 5. Boeing 777 equipment rack locations.

AOA sensor

Capt FO static static port port

Alt static port

To stby inst and cabin press

Capt FO static static port port

Alt static port

needed for each parameter. With the growing complexity of aircraft, and the resulting need for communication of increasing amounts of data between systems, use of analog means for communication was becoming very inefficient, in terms of the cost and weight of installation of the large quantity of wiring needed. In order to improve this situation, ARINC 429 was developed in the late 1970s to provide a standard for digital communication between LRUs. In ARINC 429, each bus contains a single transmitter, and multiple receivers. This is illustrated in Fig. 7. Each transmitter can broadcast a number of 32-bit words on the ARINC 429 bus; in most cases, each 32-bit word is broadcast at a specified periodic rate, and each contains a label for identification, and one or more data parameters. Example data parameters include airspeed, altitude, hydraulic pressure, landing gear position, and cabin pressure. In order to receive any parameter on a given ARINC 429 bus, an LRU must contain a receiver for that bus, and be wired to that bus. The LRU then uses the labels as identifiers to select the parameters it needs to receive. ARINC 429 was first used on the Boeing 757 and 767 aircraft. ARINC 629. In the 1980s, aircraft complexity continued to increase. The resulting increase in communications between LRUs was making the use of ARINC 429 costly in many areas. As an LRU requires a separate receiver for each bus it listens to, some LRUs required over 100 receivers. Since each receiver requires a pair of wires, the cost and weight of this wiring was becoming a significant burden. To address these and other issues, ARINC 629 was developed in the late 1980s. In ARINC 629 multiple transmitters can share the same bus, which also means that LRUs can receive signals from multiple transmitters with a single receiver. This is also illustrated in Fig. 7. This greatly reduced the number of receivers and amount of wiring from that required for ARINC 429. In addition, many design changes can be accommodated without additional hardware changes, as receiving LRUs can just be reprogrammed to receive parameters from other LRUs on the bus. In ARINC 629, each transmitting LRU broadcasts a number of word strings, each containing an identification label and a number of 16-bit words. The 16-bit words contain the various parameters to be transmitted. A specific parameter may be a single bit (e.g., for use as a Boolean discrete), a series of bits (e.g., for variables with enumerated states), a 16-bit word (e.g., to represent a variable parameter such as

JET TRANSPORT MAINTENANCE

Right Airplane information Management System (AIMS) Cabinet E2-1 Window Heat Cntrl Unit left fwd and right

Quick Access Recorder (QAR)

Air Cabin Supply Temperature Cabin Controller Press Clt (CTC) (ASCPC) right left

Traffic Alert and Collision Avoidance Sys Computer (TCAS)

Distance Measuring Equipment Interrogator (DME) left

Air Cabin VHF Supply Temperature Comm Cabin Controller Xcvr Press Clt (CTC) (VHF) (ASCPC) right center left

Instr Land Sys Rcvr (ILS) left

VOR Rcvr Marker Beacon (VOR) left

Autopilot Flight Director Computer (AFDC) left

Audio Manag Unit

Arbrn Vibration Mon Unit left

E1-2 VOR VHF Rcvr Comm Mkr Xcvr BCn (VHF) (VOR) right right

Air Traffic Control Trans (ATC) right

Distance Measuring Equipment Interrogator (DME) right

APU Generator Control Unit (APU-GCU)

VHF Comm Xcvr (VHF) left

Instr Land Sys Rcvr (ILS) center

Air Traffic Control (ATC) left

E1-3

Pre-Rec Announcement

Passenger Inflight Information Computer

Audio Entertainment Player 2

Audio Entertainment Player 1

Audio Entertainment Multiplex 1

Entertainment Multiplexer Controller (EMC)

E2-4 FCDC Battery center

Audio Entertainment Player 3

Cabin System Management Unit (CSMU)

Pass Address Cabin Interphone Cont (PACI)

E1-4 Actuator Control Elec (ACE) center

Autopilot Flight Director Computer (ADC) right

Weight Bal Comp A

Calib Module A Calib Module B

Weight Bal Comp B

E2-5

Primary Flight Computer (PFC) center

Audio Entertainment Multiplex 2

Grnd prox Warm Comp

ADF left

Transformer Rectifier Unit (TRU) left

ADF left

Generator Control Unit (GCU) left

E2-3 Arbrn Vibration Mon Unit right

Generator Control Unit (GCU) left

E1-1

E2-2 Instr Land Sys Rcvr (ILS) left

Bus Power Control Unit

Sel Cal Dec Unit

Transformer Rectifier Unit (TRU) left

Window Heat Cntrl Unit left fwd and right

Actuator Control Elec (ACE) left 2

35

Proximity Sensor Electronics Unit (PSEU) 1

Engine Data Interface Unit (EDIU) left

Actuator Control Elec (ACE) left 1

Warning Electronics Unit (WEU) left

E1-5

Flight Control Power Supply Assembly (PSA) center

E2-6

Standby Attitude Air Data Ref Unit (SAARU) E2-7

E2 rack (looking aft)

Flap/Slat Electronics Unit (FSEU) 1

Flight Control Power Supply Assembly (PSA) left

FCDC Battery left

Primary Flight Computer (PFC) left

E1-6

E1 rack (looking aft)

Figure 6. System controllers located on the E1 and E2 racks on a Boeing 777.

airspeed), or multiple 16-bit words (e.g., where finer than 16 bit resolution is required). The receiving LRU uses the label to identify the word string it needs to receive, and then selects the applicable 16-bit word(s) and, if necessary, bit(s) containing the needed parameter(s). ARINC 629 was first used on the Boeing 777 aircraft. In addition to ARINC 429 and 629, other types of digital buses are being used on commercial aircraft, including: • ARINC 636—an application of the fiber distributed data interface (FDDI), a fiber optic local area network • ARINC 717—flight data recorder • ARINC 591—quick access recorder • Ethernet—maintenance terminals, cabin systems • RS-422—quick access recorder • Telephone Consultative Committee for International Telegraphy and Telephony, Conference of European Postal Telecommunications Administrations • ARINC 453—weather radar

The digital data buses have become much more reliable than the analog wires that they replaced. However, when problems do occur in the systems that use digital data buses, troubleshooting requires more sophisticated tools than the voltmeters that were sufficient for troubleshooting most analog systems. Fortunately, aircraft design has evolved over the years to include these more sophisticated tools. Maintenance Design Evolution Just as aircraft system design has evolved, electronic support of maintenance has evolved over the years, based on the need and available technology. Since jet transport aircraft can be in service for over 30 years, there are systems in service in each of the categories identified next. As a result, mechanics need to be able to support equipment encompassing a wide range of maintenance capabilities. Manual Detection and Isolation. Early aircraft systems were relatively simple, and most importantly, were relatively iso-

36

JET TRANSPORT MAINTENANCE

Arinc 429

Arinc 629

LRU A

Connectivity:

LRU A

LRU B

LRU B

LRU C

LRU C

Transmission rate:

100 Kbits/sec (high speed)

2 Mbits/sec

Medium:

Voltage mode

Current mode

Format:

PS AS RM 29 Bit 32

Data

S D I 11 9

Label Ext 1

1

4

H-L Sync

Label word Label 8

P A R

20 1

P A R

Data word 4

20

L-H Sync

L-H Sync

Figure 7. ARINC 429 and 629 data bus characteristics.

lated from each other. A system could contain an actuator, a sensor, and an analog device to control the actuator based on the sensor input. All the interfaces were analog, meaning that the mechanic could generally troubleshoot the system with a schematic and a voltmeter. In this era, thorough aircraft testing relied on extensive use of ground support equipment. Analog Built In Test. As time passed, the systems added functionality to meet the needs of the airlines. Some of the functionality being provided was becoming more critical to safe operation of the aircraft. To support compliance with safety requirements, fault detection monitors were added to warn the flight crew of improper operation, and often to shut down the operation of the associated portion of the system. This monitoring was known as built in test (BIT). Little additional information was given to mechanics in these designs. They largely relied on flight crew reports, schematics, and voltmeters.

is shown in Fig. 8. Front panel BITE began to decrease the need for some of the ground support equipment previously used to test aircraft equipment. Depending on the system, the fault balls or LEDs could effectively point the mechanic in the right direction, but schematics and voltmeters were needed for most conditions. However, the BITE of this era was often confusing, unreliable, and difficult to use. Mechanics often distrusted it. Many systems on aircraft such as the Boeing 707, 727, early 737/747, McDonnell Douglas DC-8, DC-9, and DC-10s, employed this type of maintenance design. Digital Built In Test Equipment. In the 1970s and early 1980s, some of the increasingly complex systems began to use computers to perform their calculations. With these computers came the ability to display fault detection and isolation information in digital form, normally via numeric codes, on

Collins

Analog Built In Test Equipment. In time, aircraft design engineers realized that the output of the fault detection monitors could be made available to support mechanic troubleshooting. With these, the concept of ‘‘fault balls’’ was born, and was incorporated on some systems as early as the 1940s. Fault balls are indications, normally on the front of an LRU (i.e., system controller), that a fault has been detected. They were originally mechanical, but later were replaced with small light emitting diodes (LEDs). In many cases, the LRU front panel contained a test switch to command the LRU to test itself, in a manner similar to how ground support equipment could test the LRU. This capability became known as built-in test equipment (BITE). A typical LRU with front panel BITE

PASS FAIL

TPR

TEST

XPNDR UPPER ANT LOWER ANT RAD ALT HDNG R/A T/A TTR-920

Figure 8. LRU with front panel BITE.

JET TRANSPORT MAINTENANCE

the front panel of the LRU. The digital logic could produce codes that could better isolate the cause of the fault. The digital display, as shown in Fig. 9, offered the capability to display many different codes or even text to identify each type of fault that was detected. Some of the later LRUs had the capability to initiate ground tests and display the results in codes of text. The codes often pointed to some description in a manual that could be used to isolate and correct a fault. Many systems on the Boeing 757/767, Airbus A300/310, McDonnell Douglas DC-10, and Lockheed L-1011 employ this approach. Common Fault Display System—ARINC 604. As the number of systems grew, use of separate front panel displays to maintain the systems became less effective, particularly since each LRU often used a different technique to display its fault data. In addition, some of the systems had become increasingly integrated with each other. Digital data buses, such as ARINC 429, began to be used during this time period. Autopilot systems, as they were among the first to use these digital data buses and depend on sensor data provided by many other systems, have been a driving force in definition of more sophisticated maintenance systems. The more sophisticated monitoring was necessary to meet the integrity and certification requirements of its automatic landing function. For example, the Boeing 767 Maintenance Control and Display Panel (MCDP) brought together the maintenance functions of many related systems (i.e., flight control computers, flight management computers, and thrust management computers). As the next step, ARINC 604 defined, in 1986, a central fault display system (CFDS) which brings to one display the maintenance indications for potentially all of the systems on the aircraft. This approach enabled more consistent access to maintenance data across systems, a larger display than each of the systems could contain individually, and saved the cost of implementing front panel displays on many of the associated system controllers. In this approach, the CFDS is used to select the system for which maintenance data is desired, and then it sends the maintenance text from that system to the display. This approach was used on some of the systems on later Boeing 737s, and most systems on the Airbus A320/330/340, and

Figure 9. Digital BITE control panel.

37

McDonnell Douglas MD11. Figure 10 shows several typical CFDS displays for the Airbus A320. Onboard Maintenance System—ARINC 624. Systems continued to become more complex and integrated. A single fault on the aircraft could cause fault indications for many systems, even when displayed using the CFDS. The mechanic had little help in determining which indication identified the source fault, and which were merely effects. To solve this and related issues, ARINC 624 was developed in the early 1990s. It defines a more integrated maintenance system that can consolidate the fault indications from multiple systems, and provide additional functionality to support maintenance. Minimal ground support equipment is needed to test aircraft systems, as most of this capability is included in the maintenance system. For example, most factory functional tests of aircraft systems on the Boeing 747-400 and 777 aircraft consist of little more than execution of selected tests, monitoring fault displays, and monitoring certain bus data using the integrated maintenance system. Onboard Maintenance System Architecture ARINC 624 defines an onboard maintenance system (OMS) as: (1) built in test equipment (BITE) in each member system, (2) central maintenance computer system (CMCS), (3) airplane condition monitoring system (ACMS), and (4) onboard maintenance data (OMD). Figure 11 shows graphically an OMS architecture. Built In Test Equipment In Each Member System. In ARINC 624, BITE refers to all of the maintenance capabilities of an LRU (the term BIT is not used). Member system BITE detects, isolates, and reports faults. It also runs tests, transmits configuration data, and performs data load when requested by the CMCS. It must perform all of these functions accurately, or the central maintenance computer system (CMCS) becomes little more than an efficient garbage collector. Central Maintenance Computer System. A CMCS consists of the central maintenance computer (CMC), a maintenance access terminal (MAT), an airplane condition monitoring system (ACMS), and onboard maintenance data. The CMC consolidates fault reports from member systems into maintenance messages and correlates them to flight crew indications (flight deck effects). It can transmit these messages to ground stations, printers, and disk drives. It also requests member systems to run tests, transmit configuration data, and perform data load. The MAT provides the display for maintenance data processed through the CMC, and contains storage devices (e.g., disk drives) to support data loading and recording of reports. Figure 12 shows the MAT on the Boeing 777. Airplane Condition Monitoring System. The ACMS provides a single point for aircraft system operational data monitoring, to support analysis of trends and prediction of future maintenance needs. The MAT provides the display capability for ACMS, just as it does for CMCS. Onboard Maintenance Data. Onboard maintenance data (OMD) is the electronic availability of maintenance documentation on the aircraft. This is intended to reduce maintenance time and effort by making this data more readily available to the mechanic.

38

JET TRANSPORT MAINTENANCE

Figure 10. Airbus A320 CFDS menus showing aircraft systems displaying information on the multi-purpose control and display unit (MCDU), which is located in the flight deck.

Subordinate LRU #1

Subordinate LRU #2

Subordinate LRU #3

...

Member system #2

Member system #n

System LRU

Central maintenance computer system

Electronic library system

Central maintenance computer

Onboard maintenance data

Airplane condition monitoring

Other data and functions

Control : and display Printer

Figure 11. Onboard maintenance system architecture.

Airplane to ground station data link

JET TRANSPORT MAINTENANCE

39

Perform Ground Tests. Many systems require that tests be run to verify that faults have been corrected, and/or the system has been installed correctly. Perform Data Loading of Systems. The functionality of many systems evolves over time. Much of this functionality is controlled only by software. Loading a new software version is an efficient means to provide updated functionality. Display and Report System Configuration. For systems that data load, the system must provide a means for the airline to determine what version of software is loaded. This function also displays and reports hardware part number and serial number to support airline tracking of parts. Monitor Aircraft System Conditions. Degradation in performance of a number of aircraft systems, particularly engines and environmental controls, is more gradual in nature. Fault detection is not always the most effective way to keep these systems performing optimally. For these types of systems, the maintenance system provides a means to record values of key parameters over time, or in response to certain trigger conditions. This data can be analyzed to identify trends in performance, which can then be used to identify recommended maintenance actions to maintain good performance. Fault Reporting and Correlation Figure 12. Boeing 777 maintenance access terminal.

Onboard Maintenance System Functions. An onboard maintenance system provides the following primary functions: Detect And Isolate Faults. When equipment fails, the mechanic needs help in determining what has failed. Systems contain monitors to determine whether and where failures have occurred. Generate Maintenance Messages. A maintenance message is the data (identification number and text) displayed to the mechanic identifying what has failed, and what action should be taken to correct the fault. A maintenance message identifies a specific procedure in a fault isolation manual. The objective is that only one maintenance message is produced when a single fault exists. Note: Multiple maintenance messages (which could be produced by several LRUs monitoring faults and simultaneously detecting one) tend to confuse the mechanic. Correlate Maintenance Messages to Flight Deck Effects. Flight deck effects (FDEs) are messages to the flight crew identifying loss of function and actions that may need to be taken during the flight due to an aircraft malfunction. The FDEs are not intended to identify how to correct the fault. The flight crew will report FDEs that have occurred, and will expect the mechanic to disposition (i.e., correct or defer) them. The maintenance system relates which maintenance message identifies the fault that caused the flight deck effect. Store, Display and Report Messages and Related Flight Deck Effects. The maintenance message and related flight deck effects are stored in CMCS memory, displayed to the mechanic and/or transmitted electronically to ground stations. Transmission to ground stations prior to aircraft arrival allows ground mechanics to be prepared to fix or properly disposition the reported faults.

Each system must detect fault conditions to prevent the system from using failed components. Systems contain monitors sufficient to detect faults as necessary to meet safety requirements and other economic objectives. Figure 13 illustrates the fault detection and processing concept used on the Boeing 777. When a member system detects a fault, it: 1. Reports to the flight crew display system that the condition should be annunciated (if necessary) to the level necessary to identify the specific required flight crew awareness/actions, and/or aircraft dispatch limitations. This indication is known as a flight deck effect (FDE). Flight deck effects are normally displayed as a part of the basic flight crew display system. They provide information at the level that will best support flight crew determination of their response to this condition. In general, this means that a function is lost or degraded. For example, a pilot need not know which component caused a function to be lost, as his actions only change based on which function has been lost. 2. Reports this fault to the CMCS (to the level necessary to indicate to the mechanic what needs to be done to correct the fault—sometimes this may require additional monitors to provide better isolation than those used to identify that a fault has occurred). This indication is known as a fault report. 3. The flight crew display system reports to the CMCS that the flight deck effect is being displayed. Based one or more received fault report(s), the CMCS generates a message for the maintenance crew, and correlates it with the flight deck effect. This message is known as a maintenance message. The maintenance message contains an identification number, which points to a procedure in fault isolation manuals, and text to indicate what has been detected, and, optionally, the LRUs that could contain the fault. In a federated BITE system (where there is no CMCS consolidation function, e.g.,

40

JET TRANSPORT MAINTENANCE

Fault detection/ isolation by subsystem BITE

Maintenance message data FDE activity

FDE data

Yes Choose to fit it ?

(Alert/status)

Log book

Flight deck effect display

Fault processing/ correlation

No Defer and dispatch

Fault isolation manual

Dispatch deviation guide Airplane maintenance manual and supporting data Figure 13. Boeing 777 CMCS fault detection and processing concept.

where BITE data is displayed on LRU front panels), there is effectively a one-to-one relationship between the fault reports and maintenance messages; that is, an LRU will record a fault report when a fault is detected, and display the associated maintenance message when requested by the operator. On aircraft with a CMCS, fault reports are transmitted by systems to the CMCS. Although in many cases there is a one-to-one relationship between fault reports and maintenance messages, the CMCS may consolidate multiple fault reports (usually from multiple LRUs) to produce a single maintenance message. 4. The CMCS stores maintenance messages in fault history memory for later interrogation. As ground maintenance often results in temporary display of maintenance messages, messages are normally not stored when the aircraft is on the ground. 5. The CMCS displays maintenance messages and correlated flight deck effects to the operator or reports them to ground stations via air to ground data links. Figure 14 shows a typical CMCS single maintenance message display for the Boeing 777. Figure 15 shows a flow diagram of the steps that an airline might follow during turnaround of an aircraft. When the aircraft first arrives at the gate, the mechanic checks for any flight deck effects or log book entries. If there are none, the

aircraft can be returned to service. If there are flight deck effects or logbook entries, the mechanic will normally check the minimum equipment list and defer repair of the condition if allowed. If deferral is not allowed or desired, the mechanic will fault isolate using the CMC and the fault isolation manual (FIM), make the repair, confirm the repair, and then return the aircraft to service. The time is very limited in this condition, so the mechanic may choose to quickly replace multiple LRUs in an attempt to allow the aircraft to leave with minimal delay. This causes removal of some nonfaulty equipment, but may be in the best economic interest of the airline (as the costs of delay or cancellation can be large). Member System Fault Reporting. The member systems contain monitors that detect fault conditions. Most of these monitors have been incorporated to detect faults that threaten the integrity of the system. Others are incorporated to detect economic fault conditions (e.g., a fault causing increased use of fuel by the engine), or to provide improved fault isolation for maintenance. A wide variety of monitors are used in the various systems on the aircraft. Some example types of monitors are: • Memory monitors, that detect if the data in memory has been corrupted, using techniques such as checksum or cyclic redundancy checks (CRC)

JET TRANSPORT MAINTENANCE

EXTENDED LINE MAINTENANCE MAINTENANCE

OTHER

HELP

41

REPORT

Existing Flight Deck Effects - Maintenance Message Data Correlated to

EICAS Status

AFDC L

Autopilot Flight Director Computer (left) has an internal fault detected by: Autopilot Flight Director Computer (left)

Maintenance Message: 22–13391 ACTIVE Occured at 1320z 01APR91 during Approach Hard This message occurred in a previous leg

Recommended Maintenance Action: Possible Causes: 1) Autopilot Flight Director Computer (left), P22101 (p11)

CORRELATED FLIGHT DECK EFFECTS:

(2)

NO LAND #3 Fault Code: 221 033 00

EICAS Status ACTIVE 31MAR91 2351z

NO LAND # Fault Code: 221 032 00

EICAS Advisory ACTIVE 1APR91 12581z

GO BACK

SHOW NOTES

• Wrap-around monitors, that detect whether output circuitry may be faulty, by feedback and interrogation of the output signal • Activity monitors, that detect if input signals are being received (normally used for monitoring signals received on data buses) • Out of range monitors, that detect if an input signal is within the expected range of values for that signal • Comparison monitors, that detect whether two or more signals agree. A comparison between two inputs often cannot be used to determine which input is incorrect; a third input can be used to determine which input is incorrect • Command response monitors, that detect if components such as actuators, valves, and relays are properly following the command signal. Care must be taken in the use of monitor results for maintenance. If their characteristics are not clearly identified, the resulting maintenance indications may confuse the mechanic or cause unnecessary maintenance actions (such as the removal of equip-

Figure 14. Boeing 777 CMCS single maintenance message display.

ment that has not failed). Key criteria include: (1) No nuisance indications, (2) line level fault reporting, and (3) ‘‘Tell me what you know.’’ No Nuisance Indications. A key problem with previous BITE designs is that messages were often displayed when no fault existed. This characteristic caused mechanics to quickly lose faith in the BITE indications. Line Level Fault Reporting. In the past, BITE often reported the occurrence of faults whether or not they may have an effect on airline operation. This caused mechanics to unnecessarily remove equipment from the aircraft, increasing airline maintenance costs. To better support airline operations, faults should only be annunciated on the aircraft if they may have an effect on airline operations. These faults are called line relevant faults. In addition, even if a fault is line relevant, the information provided on the aircraft should be at a level appropriate for how the information is to be used. The purpose of a maintenance message is to inform the mechanic how a fault is to be

42

JET TRANSPORT MAINTENANCE

Aircraft blocks in at gate

Flight deck effects ?

Aircraft blocks out of gate

No

No

Logbook entries ?

Yes

MEL defer ?

Yes

No Fault isolate

CMCS

FIM

Deferral procedure Correct

Confirm (test) Return to service Figure 15. Airplane turnaround flowchart.

corrected on the aircraft. It does little good, and, in general just adds confusion, to provide information more detailed than is needed to perform the tasks. As a result, separate indications normally should not be given to the aircraft mechanic when the maintenance action is the same. For example, there are many faults that can occur and be separately detectable within a single LRU (memory, processor, and others). However, at the aircraft, the mechanic’s action is simply to replace the LRU. Therefore, the indication on the aircraft should indicate only that the LRU has failed. Identification of the components within the LRU that have failed are useful when the LRU is to be repaired in the shop. In order to support this shop repair, this type of information is stored in nonvolatile memory (NVM) within the LRU so that it will be available for interrogation in the shop. This information should be used by the shop to aid repair of the unit, and by the LRU supplier to verify proper operation of BITE and shop fault detection/isolation equipment. ‘‘Tell Me What You Know’’. Another problem with previous BITE designs, and the resulting indications, is that they often only identified the most probable LRU that has failed. In cases where the fault is in another (unidentified) LRU, the mechanic has nowhere to go; as a result, confidence in the system is lost. BITE cannot always practically identify the cause of a fault to a single LRU. The most important characteristic is to be truthful on what has been detected, including identification of all possible causes.

Central Maintenance Computer System Fault Consolidation and Flight Deck Effect Correlation. Fault consolidation is the process of combining multiple fault reports that result from a single fault into a single maintenance message that identifies the fault and the action to remove the fault from the aircraft. Flight deck effect correlation is the process of relating this maintenance message with the flight deck effect(s) that the fault causes. In general, fault consolidation can be conceptually divided into two categories: Cascaded Effect Removal. The effects of certain faults may propagate through multiple systems on the aircraft. Each of these systems may report the fault condition it detects using a fault report, and also may cause a flight deck effect to be displayed. The CMCS responsibility in these cases is to display a message for the source fault, and relate that message to all flight deck effects that were caused by that fault. An example of this is a failure of the air data module (ADM) that receives the pitot probe output. The ADM reports this internal fault to the CMCS. Systems using this data, such as the air data/inertial reference unit and the autopilot, will report that they are not receiving valid data, and potentially display a flight deck effect to indicate this condition. This is shown in Fig. 16. The CMCS must then display the maintenance message based on the report received from the ADM, and correlate this message to the autopilot flight deck effect. The maintenance messages for the air data/inertial reference unit and autopilot computer fault reports are suppressed, so that the mechanic can quickly identify the source fault. Note: most aircraft have sufficient redundancy that it takes more than one ADM fault to cause the indicated flight deck effects. Fault Isolation. Certain faults may be directly observed by multiple systems. Each system will identify, through transmission of fault reports, what condition it has detected. Based on the combination of fault reports received, the CMCS determines the maintenance message that identifies the source fault. (Different combinations of fault reports cause the CMCS to identify different maintenance messages.) Once the maintenance message is determined, the CMCS correlates this to any flight deck effects that result from this fault. An example of this is failure of a radio altimeter to transmit on a data bus, as shown in Fig. 17. As this LRU cannot transmit, it cannot report the fault condition to the CMCS. Instead, the CMCS relies on other LRUs (in this case, the autopilot, the warning system, and flight management system) to determine that the original LRU cannot transmit. Multiple inputs are required in order to determine that a failed receiver is not the cause of the report. Isolation Versus Cost/Reliability. The goal in fault isolation on the aircraft has always been to identify the single LRU that is the source of the fault. This allows the mechanic to confidently remove the failed component and correct the fault condition. Although in many cases this is possible, there are many others where it is not possible without the addition of sensors or wiring. Addition of these sensors increases the number of components that can fail, and thus sometimes can worsen the maintenance effort. In addition, they add cost and weight to the aircraft. There are clearly cases where the addition of such hardware can be beneficial, but the benefits of improved fault isolation must be weighed against the potential reduced reliability, and increased cost and weight of the additional components.

JET TRANSPORT MAINTENANCE

Flight deck effects

Pitot probe

Fault reports:

No autoland

Air data Dynamic pressure

43

Airspeed

Air data module

Air data inertial reference unit

Autopilot computer

Air data module internal fault

Pressure is invalid

Airspeed is invalid

Maintenance access terminal Central maintenance computer

Flight deck effects: AIR DATA NO AUTOLAND Correlated maintenance message: “Air data module pressure is out of range”

As a result, the CMCS cannot practically produce the perfect answer (the single faulty LRU) in all cases. It can point the mechanic to a small group of LRUs in almost all cases. Even in this case, if it is reliable in doing this, it is still a very necessary and effective tool to aid in mechanic correction of aircraft problems. Central Maintenance Computer System Fault Storage. Once maintenance messages and correlated FDEs are identified, they may be stored for later retrieval by maintenance personnel. This is particularly critical where the fault is intermittent or can only be detected in certain conditions, since in these cases the monitors may not be detecting the fault by the time that the aircraft returns to the ground. This storage of maintenance messages and correlated FDEs is called fault history. In order to be effective, the system must be designed so that maintenance messages are stored in fault history only for fault conditions. In particular, ground maintenance activity often induces perceived fault conditions which are detected by the various system monitors. For example, an LRU is expected to transmit on a data bus when the aircraft is flying; if it stops transmitting in flight, a real fault condition exists, and a maintenance message should be recorded. During maintenance, if a circuit breaker for this LRU is opened, other LRUs will report that this given LRU is no longer transmitting. This is not a real fault in the LRU, and thus, maintenance messages should not be recorded. Therefore, maintenance messages for these conditions are normally not stored in fault history when the aircraft may be undergoing maintenance. The CMCS uses flight phases to determine when a message should be stored. Flight phases identify specific regions of the aircraft flight cycle (including engine start, taxi out, takeoff, climb, cruise, descent, approach, roll-out, taxi in, engine shutdown, and maintenance).

Figure 16. CMCS cascaded effect removal.

Ground Tests Ground tests are designed to allow the mechanic to verify proper installation and operation of all or part of the system. They are initiated based on user request. Ground tests are often used to verify whether a fault has been corrected. For some faults, ground tests are designed to re-create conditions under which a fault can be detected, and then determine if the fault exists. One very important issue regarding use of these tests is to make sure that they are not run at an inappropriate time. For example, a flight control system should not run a maintenance test while the pilot is flying the aircraft, as hazardous conditions could result. The applicable systems contain safeguards to prevent such inappropriate ground test operation. Data Load/Configuration Identification Data load is used to load new software or data into an LRU. Much of the functionality of modern systems is incorporated into software. As changes to this functionality are desired, either to correct problems or add new features, software updates are required. Data loading provides the means to efficiently install the new software onto the aircraft. Data loading shares one common issue with ground tests. Each system must provide safeguards to make sure that software can only be loaded when it is safe to do so. Otherwise, loading of software into a flight control system while the aircraft is in flight, for example, could have hazardous consequences. Another important issue with data loading is that the airline must make sure that the resulting software configuration is legal for flight. To support this determination, the system must provide a configuration identification function, in which it can request and display software and hardware configuration for any of the applicable systems. This tool can also be used by the airlines to track what LRUs are installed on each aircraft.

44

JET TRANSPORT MAINTENANCE

ful in situations where an LRU has failed, and the airline needs to know the configuration of the remaining LRUs in the system, so that a compatible replacement LRU can be ready when the aircraft arrives.

Transmitter fault

Radio altimeter

Autopilot

Warning system

Flight management

No input from radio altimeter

No input from radio altimeter

No input from radio altimeter

Maintenance access terminal Central maintenance computer

Receiver fault

Autopilot

Maintenance message: Radio altimeter has no output.

Radio altimeter

Warning system

Flight management

No input from radio altimeter

Maintenance access terminal Central maintenance computer

Maintenance message: Radio altimeter has no output.

Figure 17. CMCS fault isolation.

Reporting Reporting consists of the capability to transmit the results of the various CMCS functions to output devices such as a printer, a disk drive, or to ground stations via an aircraft to ground data link. The latter is gaining increasing use, as airlines realize the benefits of knowing what faults have occurred on an aircraft prior to the aircraft arrival. With this information, they can be prepared for any maintenance action that may be required when the aircraft lands. This reporting also consolidates information in the hands of maintenance planning personnel so that they can plan for maintenance activities during overnight or longer maintenance periods. The CMCS can be programmed to transmit fault information automatically in a variety of ways as desired by the airlines. Reports can be transmitted as faults are detected, or a summary of the faults detected during the flight can be transmitted toward the end of a flight. In addition to this, ground stations can request transmission of fault information or system configuration information at any time. The latter is use-

Airplane Condition Monitoring The airplane condition monitoring system (ACMS) enables the airline to monitor engine and aircraft systems performance. The data collected and reported by the ACMS also allows the airline to conduct trend analysis of maintenance data. The ACMS capability includes engine and aircraft performance diagnostic tools, which are normally provided by engine and airframe manufacturers, respectively. The reports and event triggers may be customized by the airline to suit their specific needs. Airlines can create specific reports, determine data gathering criteria, select the output device for the reports and create custom screens for display on the maintenance access terminal (MAT). The ACMS software provides the ability to record data based on predefined triggers, or on real-time manual inputs from the airline. Triggers are logic equations that detect conditions such as engine limit exceedances, failures, stable frames, or other airline defined criteria. Data can be recorded for predetermined periods of time following the activation of a trigger, a manual input, or an input via ground stations. Alternatively, an airline may choose to record certain parameters continuously for post flight analysis on a routine basis. The data reports generated by ACMS may be downloaded, as specified by the airline, to any of the following devices: maintenance access terminal (MAT), data loader diskette, flight deck printer, or optional quick access recorder (QAR). In addition the ACMS generated reports can be downlinked to a ground station via digital communication management function (DCMF). Onboard Maintenance System User Interface To be most effective, access to maintenance displays should be provided where the mechanic is performing related tasks. To support this objective on the Boeing 777, the CMCS and ACMS may be accessed from: • A maintenance access terminal (MAT) in the cockpit • Side displays in the cockpit (optional equipment) • A portable maintenance access terminal (PMAT) plugged into remote terminals located in these areas: Flight deck Electronics/equipment bay Nose gear Right main gear Auxiliary power unit (APU) access door Menus provide access to data for all of the OMS functions. The menus on the Boeing 777 CMCS, as shown in Fig. 18, are structured to efficiently support particular user needs. For example, the functions that would most likely be used by a line mechanic are grouped under a line maintenance menu item. Maintenance messages listed under line maintenance menu items are limited to those that correlate to flight deck effects, as those are the only messages that a line mechanic would normally have reason to correct. The line maintenance menu

JET TRANSPORT MAINTENANCE LINE EXTENDED MAINTENANCE MAINTENANCE INBOUND FLIGHT DECK EFFECTS

OTHER FUNCTIONS

HELP

LINE EXTENDED MAINTENANCE MAINTENANCE

EXISTING FLIGHT DECK EFFECTS

REPORT

OTHER FUNCTIONS

PRESENT LEG FAULTS

GROUND TESTS

45

HELP

REPORT

LINE EXTENDED MAINTENANCE MAINTENANCE

OTHER FUNCTIONS

HELP

REPORT

EXISTING FAULTS INPUT MONITORING

SYSTEM CONFIGURATION

FAULT HISTORY

EXIT MAINTENANCE

CENTRAL MAINTENANCE OPTIONS

DATA LOAD ENGINE BALANCING MAINTENANCE PLANNING SHOP FAULTS MAINTENANCE ENABLE/DISABLE

PSEU AND AIR/GROUND RIGGING

EXIT MAINTENANCE

CENTRAL MAINTENANCE COMPUTER SWITCH CONTROL

1

SPECIAL FUNCTIONS

2

EXIT MAINTENANCE

3

Figure 18. Boeing 777 CMCS menus.

item also contains capabilities to run ground tests (to verify whether a fault has been corrected, or an LRU has been installed correctly) and display configuration data (to verify that the correct LRU and/or software has been installed). Extended maintenance and other functions menu items provide functions more likely to be used in an overnight or more extended maintenance periods. For example, under extended maintenance are menu items that can display all maintenance messages that are being or have been detected, whether or not they identify a cause for a flight deck effect. Those messages not correlated to flight deck effects are economic faults. These economic faults do not affect safety of the aircraft in any way, but could have economic consequences such as future aircraft availability or increased fuel usage. Certain messages are identified in a maintenance planning menu as maintenance memos. Maintenance memos highlight faults in fault tolerant parts of systems, and indicate that another similar fault will cause some impact on aircraft dispatch. AIRPLANE MAINTENANCE DESIGN TRENDS Greater Airplane System Integration Airplane systems are becoming more and more interdependent. This is due to the increasing use of data buses, which has made data from each system much more available to the rest of the systems on the aircraft. This data availability in turn has enabled increased functionality, which in many cases can result in greater efficiency, weight reduction, and other improvements. This also causes the effects of faults to

propagate more widely between systems. As a result, mechanics are more dependent on systems such as the CMCS to help them determine how to correct a given problem. Devices such as the CMCS will need to grow in complexity to allow accurate identification of the faulty components, and the effects of those faults. Use of aircraft system models in CMCS design is expected to increase in order to support this growing complexity. Greater Use of Downlinked Information With the limited amount of time a typical commercial aircraft may have between flights, advance (prior to arrival) information of faults that have occurred can facilitate more timely disposition of these conditions. If the condition is deferrable, this advance information can give maintenance personnel time to consider the options and decide on a course of action. If the condition is to be fixed before the next flight, the information can allow maintenance personnel to be prepared with replacement equipment when the aircraft lands. Transmission of this data can also aid in planning future maintenance activity—faults reported in these transmissions can more readily be scheduled for repair when the equipment, time, and personnel are available. Airlines are making increasing use of this capability as more aircraft support it. Greater Use of Prognostics The airplane condition monitoring system provides capabilities to identify trends in performance, in part to determine if and when equipment may benefit from maintenance. Increas-

46

JET TRANSPORT MAINTENANCE

ing use of these and other prognostic capabilities are expected as soon as sufficient economic benefits can be identified. Electronic Maintenance Manuals Traditionally, maintenance manuals have been printed and located away from the aircraft, causing mechanics time and effort to retrieve them. Maintenance manuals are increasingly being distributed electronically, and are often accessed via portable computers that the mechanic may bring onto the aircraft. Other means for making this data available on the aircraft (e.g., installation of computers containing this data on the aircraft) are expected to become more widely available. (See section on ‘‘Electronic Performance Support Tools’’ later in this article.) MAINTENANCE SUPPORT All of the previous discussion about design for maintenance and the ability of the aircraft to identify its faulty components and systems is to no avail unless there is accurate and up-todate technical documentation available and the work force to perform the maintenance is skilled and properly trained. Both of these areas are critical to the success of an airline’s maintenance program. Technical Documentation The amount of information and documentation required to support the maintenance of a modern jet transport aircraft is huge. For example, approximately fifty manuals, totaling over 40,000 pages, are required to support the maintenance of a typical jet transport, such as the Boeing 777. Maintenance technicians, depending on their experience and maintenance role, estimate that they spend as much as forty percent of their workday accessing technical information contained in these documents. The support documentation, mostly written by the airframe and engine manufacturers, ranges from aircraft maintenance manuals, training manuals, wiring diagram manuals, schematics manuals, fault reporting and isolation manuals, BITE manuals and ramp maintenance manuals, to such support documentation as service bulletins. There are also thousands of aircraft drawings that are kept on file because they may be needed to support a modification or repair of an aircraft. Similarly, the component manufacturers, who are contracted by the airframe and engine manufacturers to design and build many of the components for the aircraft and engines, develop and produce component maintenance manuals for each of the components they manufacture. All of the documentation used to support the maintenance of aircraft must be accurate and kept up-to-date. Much of the documentation is specific to individual aircraft (as designated by tail numbers or registration numbers, etc.) because the equipment or components installed in a given aircraft may not be of the same revision level as that installed in an aircraft produced at a later date. Most of the documentation is continuously revised throughout the life of the aircraft. For example, nearly all of the documentation supplied by the airframe manufacturer is revised on a quarterly basis. This is expensive for the manufacturers, who must produce and send multiple copies to the

airlines, totaling hundreds of thousands of pages for one model alone. It is also costly for the airlines who must continuously revise their documentation and keep records of the revision. Because of this cost to maintain all of the documentation, there was a great need to digitize it. Beginning in the early 1990s, efforts were made to digitize these documents and allow access through desktop computers and using electronic performance support tools. (See section entitled ‘‘Electronic Performance Support Tools’’ later in this article.) Often a wide variety of aircraft types are operated by an airline. They may be Boeing models, McDonnell Douglas models, Airbus models, or a combination of them all. With all of the different models operated by the airlines and the resulting wide variety of support documentation, it became necessary to standardize. Almost all of the support documents used today by the airlines conform to ATA standards that are contained in two specifications. ATA Specification Number 100 contains the standards for paper, microfiche, and microfilm documentation and ATA Specification Number 2100 contains the standards for digital data. Air Transport Association Specification Number 100. When aircraft, engine, and component manufacturers develop manuals to support their respective products, they adhere to the documentation standard in ATA Specification 100. The standards describe how the documents should be organized, so that no matter what aircraft, or aircraft system, one is researching, it can be found in the same fashion. The standards in this specification are recommendatory in nature, and become mandatory to the extent they may be incorporated into the purchase agreements executed between the individual suppliers and the individual airlines. Specific documents identified by ATA 100 include: • • • • • • • • • • • • • • •

Aircraft Maintenance Manual Wiring Manual Structural Repair Manual Illustrated Parts Catalog Component Maintenance Manual Illustrated Tool and Equipment Manual Service Bulletins Weight and Balance Manual Nondestructive Testing Manual Power Plant Build-up Manual Aircraft Recovery Manual Fault Reporting and Fault Isolation Manuals Engine Manual Engine Illustrated Parts Catalog Engine Parts Configuration Management Selection Process Manual • Miscellaneous Technical Data • System Descriptions Section • Maintenance Planning Document Air Transport Association Specification 2100. As support documents have transitioned from paper/film to digital, and from closed to open systems, a different standard was developed to establish these standards for digital maintenance data. ATA Specification 2100 established these standards for the au-

JET TRANSPORT MAINTENANCE

thoring, interchange, delivery, and use of digital data produced by aircraft, engine, and the component manufacturers. ATA Specification 2100 will replace ATA Specification 100, when all support documents have transitioned to digital format. ATA Specification 2100 is not limited to particular functional areas for aircraft as the ATA Specification Number 100 is, although further development of functional requirements may be added during ATA Specification 2100’s lifetime. Air Transport Association Chapter-Section-Subject Numbering System. Whether in paper or digital form, a standard numbering system is used throughout most jet transport technical documentation. It follows ATA Specification Number 100 which specifies all technical data be organized by this number system. The numbering system specified in ATA Specification 100 is known as the ATA chapter-section-subject numbering system. The numbering system consists of three elements. The first element assigns an ATA chapter number to each aircraft system. For example, ATA Chapter 28 is for the fuel system, ATA Chapter 34 is for navigation systems, and so on. The second element assigns an ATA section number for each subsystem. For example, a subsystem for the fuel system might be ‘‘Indicating,’’ and has a section number 30 assigned. Therefore, any document referencing a fuel indicating system component would start with the ATA Chapter section number 28-30. The third element is a unique number assigned by the aircraft manufacturer for a specific component. For example, a fuel system temperature sensor, which is used to provide a temperature indication in the flight deck, might have a ATA subject (or sometimes referred to as unit) number 06 assigned. All references to this component in the technical manuals would use the number (or portions of this number) 28– 30–06. A list of the ATA chapter-section-subject numbering system contained in ATA Specification Number 100 is as follows: ATA Chapter 5: Time limits/maintenance checks. Manufacturers’ recommended time limits, maintenance checks, and inspections. ATA Chapter 6: Dimensions and areas. The area, dimensions, stations, and physical locations of the major structural members of the aircraft. Also includes zone locations. ATA Chapter 7: Lifting and shoring. Charts showing lifting and jacking points for maintenance, overhaul and repair. Standard jacking procedures and lifting and shoring for abnormal conditions. ATA Chapter 8: Leveling and weighing. ATA Chapter 9: Towing and taxing. ATA Chapter 10: Parking and mooring. ATA Chapter 11: Required placards. The location and pictorial illustrations of placards, stencils, and markings. ATA Chapter 12: Servicing. Replenishment of all aircraft system reservoirs (fluid and gaseous), oil changes, lubrication, and toilet draining and flushing. Filter types and locations. Also cold weather maintenance and exterior cleaning. ATA Chapter 20: Standard practices. Airframe stan-

47

dard maintenance procedures applicable to multiple aircraft systems. ATA Chapter 21: Air conditioning. Airplane heating and cooling including pressurization and ventilation. ATA Chapter 22: Autoflight. Autopilot/flight director system, yaw damper, speed trim, and auto throttle. ATA Chapter 23: Communications. High frequency (HF), very high frequency (VHF), satellite communication (Satcom), ACARS, select call (Selcal), passenger address and entertainment, audio integrating and interphone systems, voice recorder, and static discharger. ATA Chapter 24: Electrical power. Electrical generation and distribution, 115/200 volts ac, 28 volts ac, 28 volts dc, and battery system. ATA Chapter 25: Equipment/furnishings. Equipment installed for crew members and passengers, including galley and lavatory, seats, insulation, storage areas, escape and life saving equipment. Includes procedures for cleaning and repair of furnishings. Also includes cargo compartments, and cargo handling equipment. ATA Chapter 26: Fire protection. Automatic fire and overheat detection for engines and APU, automatic smoke detection for lavatories and cargo compartments. Fire extinguishing for engines, APU, lavatories, and cargo compartments. Also includes portable fire extinguishers. ATA Chapter 27: Flight controls. Ailerons, rudder, elevators, horizontal stabilizer, trailing edge flaps, spoilers, speed brakes, leading edge flaps, and indicating components of the flight control system. ATA Chapter 28: Fuel. Fuel storage, ventilation, distribution, fuel jettison, and indication. ATA Chapter 29: Hydraulic power. Main hydraulic power, auxiliary, standby, and indicating components of the systems. ATA Chapter 30: Ice and rain protection. Wing, nacelle, pitot probe, window anti-icing; windshield wipers, repellent and washers; water and toilet drain heaters. ATA Chapter 31: Indicating/recording systems. Instruments, panels, clocks, recorders, warning, flight crew displays, ACMS. ATA Chapter 32: Landing gear. Body, wing and nose gears, gear doors, hydraulic and electrical extensionretraction, wheels and brakes, antiskid, nose and body gear steering, and position and warning system. ATA Chapter 33: Lights. Warning, annunciator, anticollision, navigation, and emergency lights. Also includes area lighting and instrument lighting. ATA Chapter 34: Navigation. Air data, altitude alert, windshear alerting, inertial reference (IRS), standby instruments (air data, compass, attitude), instrument landing (ILS), marker beacon, radio altimeter, weather radar (WXR), air traffic control (ATC), traffic alert/collision avoidance (TCAS), ground proximity warning (GPWS), VHF omnidirectional ranging (VOR), distance measuring (DME), automatic direction finding (ADF), global positioning (GPS), flight management computing system (FMCS). ATA Chapter 35: Oxygen. Systems and equipment for storing, regulating, and delivering oxygen.

48

JET TRANSPORT MAINTENANCE

ATA Chapter 36: Pneumatic system. Distribution of compressed air from source to using system. ATA Chapter 38: Water and waste. Systems and equipment for storing and delivering fresh water, and removal of toilet and water wastes. ATA Chapter 45: Central maintenance system. Reports maintenance messages for a number of aircraft systems. The messages reported are existing faults, flight leg faults, fault history, BITE, ground tests, and so on. The system includes the central maintenance computer. ATA Chapter 49: Airborne auxiliary power. APU, fuel control, ignition, starting, air, APU controls, indicating, exhaust, and oil systems. ATA Chapter 51: Structures. Identification of various structural sections along with interior and exterior finishing and sealing. ATA Chapter 52: Doors. Energy and exit doors, landing gear doors, and doors for cargo access and servicing. ATA Chapter 53: Fuselage. The structural members which make up the compartments for equipment, passengers, crew and cargo including skins, bulkheads, frames, stringers, floor beams, floors, pressure dome, tail cone, fuselage to wing and empennage fairings, and others. ATA Chapter 54: Nacelles/pylons. Those structural units and associated components/members which furnish a means of housing and mounting power plant. Includes skins, longerons, frames, stringers, clamshells, doors, nacelle fairings, and others. ATA Chapter 55: Stabilizers. Structure of horizontal and vertical stabilizers including the structure of the elevator and rudder. ATA Chapter 56: Windows. Passenger windows and crew windshields. Includes windows used for observing compartments and equipment. ATA Chapter 57: Wings. Structure of the wings, flaps, ailerons, and spoilers. ATA Chapter 70: Standard practices. Engine-standard maintenance procedures applicable to multiple engine systems. ATA Chapter 71: Power plant. Power plant, cowling, mounts, and drains. ATA Chapter 72: Engine. Compressors, combustion chamber, turbines, and accessory drive gearbox. ATA Chapter 73: Engine fuel and control. Control and distribution of fuel beyond main fuel disconnect on aircraft. Includes fuel control, pump, and heater, and fuel flow, temperature, and pressure indicating systems. ATA Chapter 74: Ignition. Generation, control, and distribution of ignition current. Includes ignition exciters, igniter plugs, and ignition switches. ATA Chapter 75: Air. Accessory cooling and bleed air controls. Includes compressor bleed valves and controls, and variable compressor stator vane actuator and control. ATA Chapter 76: Engine controls. Engine controls including thrust levers and cables, start levers and switches, and engine fuel shutoff components. Also includes engine fire emergency shutdown.

ATA Chapter 77: Engine indicating. Engine pressure ration (EPR), exhaust gas temperature (EGT), and tachometer indicating system. Also includes airborne vibration monitoring system. ATA Chapter 78: Exhaust. Fan thrust reverser, turbine thrust reverser, thrust reverser controls, and position indicating system. ATA Chapter 79: Oil. Storage and distribution of engine oil external to engine. Includes oil tank, oil cooler, and quantity, pressure, and temperature indicating systems. ATA Chapter 80: Starting. Engine cranking, including starter, start valve, and valve position indicating system. Aircraft Maintenance Task Oriented Support System. In addition to the ATA chapter-section-subject numbering system described previously, a second, more detailed numbering system is often used, which is referred to as Aircraft Maintenance Task Oriented Support System (AMTOSS). AMTOSS is a numbering system designed to further improve the organization of the technical manuals and to facilitate and standardize automated data retrieval. In addition, and separate from the technical manuals, it provides for a databased approach to integrating, interfacing, isolating, and coordinating the aircraft maintenance task accomplishment, job requirements, and resources support analysis typically done by a maintenance department for the airline. AMTOSS is based on the concept of using a standard and unique numbering system that is an expansion of the ATA chapter-section-subject numbering system. The numbering system, which is an expansion of the ATA three element numbering system, uses seven elements. Each element has a specified function which is specified in ATA Specification 100. Typical Documents for a Jet Transport Aircraft Maintenance. Many different manuals are used by an airline, each used for a specific function or functions. For example, some are used to support scheduled maintenance and some to perform unscheduled maintenance. Figure 19 shows which documents are typically used by an airline for scheduled maintenance and unscheduled maintenance. Each of these documents is written for a specific aircraft type, such as a Boeing 747. Another set of manuals would exist for every other aircraft type, such as an Airbus A320, McDonnell Douglas DC-10, or a Boeing 767. Some documents are customized for a specific aircraft or series of aircraft (e.g., effectivity to a range of aircraft as designated by tail numbers and registration numbers), and some are not customized and apply to all aircraft of a given type. TECHNICAL TRAINING FOR MAINTENANCE An integral part of the maintenance process at an airline is technical training. The maintenance personnel who need to be trained are not only mechanics, but also engineers, instructors, dispatchers, maintenance planners, and management. Maintenance training courses are developed and conducted by training departments at many airlines, as well as by the airframe, engine, and many component manufacturers. In addition, many colleges, universities, and independent aviation

JET TRANSPORT MAINTENANCE Scheduled maintenance

49

Unscheduled maintenance Structural damage

Through stop turn around daily planned checks

Maintenance planning data document

Structural Structural repair repair manual manual

Flight faults

Fault reporting manual

•Flight faults •Ground faults •Service problems

BITE manual

Dispatch deviation guide Maintenance tips

Task cards and indexes

Airplane maintenance manual

Fault isolation manual

Job completion

Supporting data

System schematics manual

Wiring diagram manual

Illustrated parts catalog

Standard wiring practices manual

Figure 19. Maintenance documents.

schools (that specialize in aviation training) offer courses on aviation maintenance. Regulatory Requirements for Training. The training of the mechanics and the many maintenance personnel at an airline is not only necessary for safe and efficient airline operations, but is required and regulated by government regulatory agencies in most countries. In the United States, the FAA regulation which defines the requirement for training is FAR Part 121.375 Maintenance and Preventive Maintenance Training Program. It states that: . . . Each certificate holder or person performing maintenance or preventive maintenance functions for it shall have a training program to ensure that each person (including inspection personnel) who determines the adequacy of work done is fully informed about procedures and techniques and new equipment in use and is competent to perform his duties.

Each airline typically defines its maintenance training requirements in the airline’s overall maintenance program. This maintenance program is reviewed and approved by the government regulatory agency. Training for Mechanics and Technicians. The initial training for mechanics to get their certification and ratings is referred to as ab initio training (meaning from the beginning). Ab initio training is offered by specialized aviation schools, at colleges and universities that have aviation programs, or even by some of the airlines. Many of these schools, in addition to preparing the mechanic for his certification and rating, offer various levels of degrees, ranging from diplomas of completion to Bachelors and Masters Degrees in Aviation Maintenance and other aviation specialties. In the United States these training schools are covered under FAR Part 147, Aviation Maintenance Technician Schools. It prescribes the requirements for issuing aviation maintenance technician school certificates and associated ratings and the general operating

50

JET TRANSPORT MAINTENANCE

rules for the holders of those certificates and ratings. The following ratings are issued under FAR Part 147: (1) Airframe, (2) Powerplant, and (3) Airframe and Powerplant. The number of courses and the length of time it takes to get a mechanic’s certificate and rating varies from country to country. In the United States, to complete all of the required courses and to fulfill the practical experience requirement takes approximately 2 years. Once course work is complete, the mechanic must pass written, oral, and practical examinations before being issued a certificate and associated rating for the particular area they studied. In the United States it is either an Airframe, Powerplant, or combined Airframe and Powerplant (A&P) rating. The regulations for certification of mechanics is covered in FAR Part 65, Certification: Airmen Other Than Flight Crewmembers. It prescribes the requirements for issuing the following certificates and associated ratings and the general operating rules for the holders of those certificates and ratings: (1) Air traffic control tower operators, (2) Aircraft dispatchers, (3) Mechanics, (4) Repairmen, and (5) Parachute riggers. A proposed new FAR, Part 66, specifies new rules for aviation maintenance personnel.

creases retention of ‘‘need to know’’ data. Users of ATA Specification 104 include airline training departments, manufacturer training departments, computer based training (see CBT later in this article) vendors, and regulatory agencies. ATA Specification 104 specifies five levels of target students, their entry level requirements, and the objectives that a particular level of training is intended to achieve. The five levels are as follows:

Aviation Associations and Councils. Many aviation associations and councils have been formed by the airlines, manufacturers, and aviation specialty schools to provide guidelines to colleges and universities for aviation maintenance training and accreditation. Several key associations and councils involved in aviation maintenance training are: Aviation Technician Education Council (ATEC). This organization is made up of FAA approved Aviation Maintenance Technician schools (FAR Part 147 schools), the industry (airlines, manufacturers, etc.), and governmental agencies. It was founded in 1961 to further the standing of FAA approved schools with education and industry, and to promote mutually beneficial relations with all industry and governmental agencies. This organization is very active in FAR 147 regulations and the rewrite of FAR Part 65 and 66. Council on Aviation Accreditation (CAA). The CAA is an independent council which sets standards for all aviation programs taught in colleges and universities in America. It is responsible for hearing and ruling on accreditation applications by these institutions and to review the quality of these programs every five years. Its members include the faculty of aviation institutions and industry members such as aircraft and engine manufacturers and airlines. ATA Specification 104 Maintenance Training Subcommittee. This subcommittee of the ATA developed ATA Specification 104. It contains the guidelines for aircraft maintenance training which most airlines and aircraft and engine manufacturers follow (see the next section).

ATA Specification 104 further specifies course development guidelines, objectives for each level, course development procedures, training manual format, and issuance and revision policy. It also includes guidelines for computer based learning materials as well as computer managed instruction (CMI).

Air Transport Association Specification 104 Guidelines for Aircraft Maintenance Training. ATA Specification 104 Guidelines for Aircraft Maintenance Training was developed by the Maintenance Training Subcommittee, which was made up of representatives from the airlines and airframe/engineer manufacturers. Its purpose is to provide a better understanding of the training requirements of the various job function/skill mixes resident in airline maintenance operations. By following these guidelines, training programs’ development/packaging is more precisely oriented to the skill/job of the students. This enhances the student acceptance of the training and in-

• Level 1: General familiarization—for management and other support personnel • Level 2: Ramp and transit—for personnel associated with through flight maintenance activities • Level 3: Line and base maintenance training—for personnel associated with line and base maintenance • Level 4: Specialized training—for personnel associated with base/heavy maintenance • Level 5: Component overhaul training—for shop technicians

Technical Training at the Airlines. The training departments at most airlines develop and continually conduct a wide range of courses on maintenance and procedures. These courses typically follow ATA Specification 104, guidelines for aircraft maintenance training, as described earlier. Each airline’s technical training department typically has two primary objectives: (1) to establish and maintain an adequate program of training; and (2) to maintain adequate records of the training accomplished. Training conducted at an airline typically consists of the following four types of training: Indoctrination Training. This training is designed to familiarize the airlines maintenance personnel with the airline’s operations and procedures, and to keep that knowledge current. When maintenance personnel initially join an airline, they will receive an introduction on the airline’s policy and procedures manuals, the proper use of the technical manuals and documentation, and instructions on how to use the airlines standard forms and work sheets. Initial Training. This training is designed as the formal course on each aircraft model type they are to maintain. The training is based on the airframe and engine manufacturer training programs, and often is just a subset of the courses the airline receives when they are trained by the manufacturers. Initial training at an airline is typically customized to the actual configuration of the aircraft model being trained. Recurrent Training. All training other than initial training is considered recurrent training. Recurrent training is designed to keep maintenance personnel aware of pertinent changes to the aircraft, to the airlines organization, support equipment, policies and documentation, and airport and environmental regulations. Recurrent training is also conducted to make the maintenance personnel aware of aircraft differences as newer models are added to an existing model fleet,

JET TRANSPORT MAINTENANCE

and to inform them of problematic maintenance areas on the aircraft. Qualification Training. This training is conducted to enable individuals to be certified by the airline to accomplish specific maintenance tasks that require more training than the basic program. Tasks that require this type of training are engine run-up, taxi, engine boroscoping (i.e., where an optical tool is used for visual inspection of the internal components of the engine) and nondestructive testing. Often certain tasks that fall into this category require periodic re-qualification. Training by the Aircraft/Engine Manufacturers. When a new aircraft model is introduced into an airline’s fleet, initial training is conducted by the airframe and engine manufacturers. The students are primarily the airline’s technical training instructors and engineering personnel. This training typically takes place at the manufacturer’s training center. Different types of courses are conducted for the different maintenance personnel such as an airframe course, an avionics course, and engine run-up courses. The courses conducted at the manufacturer, like the airline’s courses, also follow ATA Specification 104, guidelines for aircraft maintenance training. Maintenance Training Media in the 1990s. Training developed for the jet transports of the 1990s uses an integrated media, each designed to increase the student’s comprehension and retention level by making the training more interesting. Courses conducted today by the manufacturers typically consist of the following types of media: Classroom Lecture. Classroom lecture consists of presentations by instructors using digitally projected line drawings. These are the same graphics that are contained in the System Description Section (SDS) of Part 1 of the Airplane Maintenance Manual. Computer Based Training (CBT). CBT is a lesson that runs on a computer that has a dynamic presentation and control of graphics and animations. For the Boeing 777, there were two types of CBT lessons; student-paced CBT lessons and instructor-led CBT lessons. With student-paced CBT lessons, the student takes the lesson at his or her own pace and does not require an instructor. Instructor-led CBT is the projection of specially developed CBT in the classroom, controlled by the instructor. Instructor-led CBT is used when animations are needed to more clearly instruct a concept that would be difficult with projection of graphic. Field Trips to the Factory. Periodically (about once a week), the students visit the aircraft factory to see the actual components they are learning about in the course. Maintenance Training Simulator (MTS). MTSs are much like the fixed base simulators used for training pilots (i.e., a fully functional flight compartment without a visual system or motion platform), except they contain additional maintenance features. These maintenance features consist of simulated aircraft components such as BITE modules, central maintenance computers, the external power panel, the fuel panel, and pneumatic connectors. Airplane malfunctions can be set by the instructor on an instructor station to simulate aircraft faults. The faults can be isolated by the students, from finding the fault in the flight deck to performing the BITE tests, all while using the real aircraft support documentation.

51

The MTS lessons typically focus on the maintenance performed on the flight line between flights, either during a turnaround or overnight/base maintenance. This concept, named line oriented scenarios, focuses on the material recently covered in the classroom and CBT. The students put the knowledge gained in the classroom and skill gained in CBT to work by performing real maintenance tasks in the MTS. ELECTRONIC PERFORMANCE SUPPORT TOOLS Because of the vast quantity and array of technical documentation that are necessary to perform maintenance on jet transports, digitizing the data and making it accessible from a computer became necessary. Beginning in the early 1990s, aviation manufacturers began digitizing their maintenance documents, thus making them accessible from a variety of devices such as PCs, laptops, or other specially built devices designed specifically for aircraft maintenance support. Because these devices aid in the performance of maintenance tasks, they became known as electronic performance support tools. Each maintenance electronic performance support tool is essentially an electronic library of maintenance documents. It consists of a digital database of technical data or technical documents that are accessed via a retrieval software program. Typically the tool is nothing more than a CD ROM containing the technical documents already described, loaded into a laptop computer. As electronic support tools evolved, many specially built devices were designed specifically for aircraft maintenance. Figure 20 shows a Boeing 777 portable maintenance access terminal. Besides the variability of types, electronic performance support tools also vary in what they do or can perform. Often they contain not just the technical documents that are used for a reference when performing a maintenance task, but also additional features such as training programs, case-based-

Figure 20. Boeing 777 portable maintenance access terminal.

52

JUNCTION FIELD EFFECT TRANSISTORS CIRCUITS

reasoning applications that record information, and artificial intelligence. For example, the case-based-reasoning tool may be used to give the mechanic the most probable cause for a aircraft fault. The cause would be derived from a database of maintenance history. Moreover, data about this particular fault and the resulting fix could be added to the case-base reasoning database thus continually improving its ‘‘reasoning.’’ Maintenance electronic performance support tools can be used by anyone who is involved in the planning or performing of scheduled or unscheduled maintenance on an aircraft. This includes maintenance technicians, controllers, engineers, dispatch personnel, and spares personnel. Because maintenance electronic performance support tools are portable, they may be used at the gate, on the aircraft, at the hangar, or on a person’s desk. Typically the paper-based maintenance documents used by mechanics are voluminous and are located in a crew room or line shack, often far from the aircraft. In effect, maintenance electronic performance support tools allow all the documents to be taken to the aircraft, making the mechanic efficient because he can take all the data he needs to the aircraft to perform his job. This can save many trips the mechanic must take to and from the aircraft to access the traditional paper/microfilm documentation, which saves a large amount of time.

2. ARINC Report 604, Guidance for Design and Use of Built-In-Test Equipment, Annapolis, MD: Aeronautical Radio Inc., 1988. 3. ARINC Report 624-1, Onboard Maintenance System, Annapolis, MD: Aeronautical Radio Inc., 1993. 4. ARINC Specification 629, Multi-Transmitter Data Bus, Annapolis, MD: Aeronautical Radio Inc. 5. A. J. Martin, Development of Onboard Maintenance Systems on Boeing Airplanes, Aerospace, August 1989. 6. United States Code of Federal Regulations, Title 14, Aeronautics and Space, Washington, DC: Federal Aviation Regulations, 1996. 7. Air Transport Association of America (ATA) Specification 100, Specification For Manufacturers’ Technical Data, Washington, DC: Air Transport Asso. Amer., 1981. 8. Air Transport Association of America (ATA) Specification 104, Guidelines for Aircraft Maintenance Training, Air Transport Association of America Maintenance Training Sub-Committee, 1996. 9. Air Transport Association of America (ATA) Specification 2100, Digital Data Standards for Aircraft Support, Washington, DC: Air Transport Assoc. Amer., 1997. 10. M. E. Irrang, Airline Irregular Operations, The Handbook of Airline Economics, New York: McGraw-Hill, 1995.

JACK HESSBURG RICHARD REUTER WILLIAM AHL Boeing Commercial Airplane Group

Benefits Electronic performance support tools offer many more benefits than just portability and relief from the use of paper and microfilm documents. Because they consist of digital data, they are easily updated and can even be on-line. This eliminates the expense of paper revisions and the labor to revise maintenance documentation. As electronic performance support tools have evolved, they also include many user friendly features that paper/microfilm cannot offer, such as indexing systems for ease of access and fast retrieval of information, or hyperlinking, which allows quick and direct movement from document to document. Future Considerations As technology has advanced, so have the types of electronic performance support tools. From nothing more than software on a CD loaded on a laptop in the mid 1990s, electronic performance support tools are expected to evolve to small wearable computers seen through dedicated goggles or safety glasses for viewing. Devices, such as a hand held computer with a touch sensitive liquid crystal display (LCD) with low frequency transceiver are expected to be on-line to the airlines computer system. They eventually will be on-line to the aircraft manufacturer and therefore always up-to-date. Peripheral devices such as barcode readers could be connected to these devices to record a multitude of information, such as the users name, the aircraft tail number of the aircraft being worked on, the serial number of the parts removed, and the maintenance task followed. BIBLIOGRAPHY 1. ARINC Specification 429-12, Mark 33 Digital Information Transfer System (DITS), Annapolis, MD: Aeronautical Radio Inc.

JFET. See JUNCTION FIELD EFFECT TRANSISTORS CIRCUITS; JUNCTION GATE FIELD EFFECT TRANSISTORS.

JOSEPHSON JUNCTIONS, HIGH TEMPERATURE SUPERCONDUCTOR. See HTS JOSEPHSON JUNCTION DEVELOPMENT.

AIRCRAFT NAVIGATION

353

AIRCRAFT NAVIGATION Historically, pilots flew paths defined by VOR (VHF Omnidirectional Radiorange) radials or by nondirectional beacon signals using basic display of sensor data. Such paths are restricted to be defined as a path directly to or from a navigation station. Modern aircraft use computer-based equipment, designated RNAV (Area Navigation) equipment, to navigate without such restrictions. The desired path can then be direct to any geographic location. The RNAV equipment calculates the aircraft position and synthesizes a display of data as if the navigation station were located at the destination. However, much airspace is still made available to the minimally equipped pilot by defining the paths in terms of the basic navigation stations. Aircraft navigation requires the definition of the intended flight path, the aircraft position estimation function, and the steering function. A commonly understood definition of the intended flight path is necessary to allow an orderly flow of traffic with proper separation. The position estimation function and the steering function are necessary to keep the aircraft on the intended flight path. Navigation accuracy is a measure of the ability of the pilot or equipment to maintain the true aircraft position near the intended flight path. Generally, navigation accuracy focuses mostly on crosstrack error, although in some cases the alongtrack error can be significant. Figure 1 shows three components of lateral navigation accuracy. Standardized flight paths are provided by government agencies to control and separate aircraft in the airspace. Path definition error is the error in defining the intended path. This error may include the effects of data resolution, magnetic variation, location survey, and so on. Position estimation error is the difference between the position estimate and the true position of the aircraft. This component is primarily dependent upon the quality of the navigation sensors used to form the position estimate. J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

354

AIRCRAFT NAVIGATION

Intended path Path definition Error

99.999% Integrity limit 2X NM

Path defined by Pilot or RNAV equipment

95% Accuracy limit X NM

Flight technical error Estimated position of aircraft

Intended path X NM 95% Accuracy limit

Position estimation error

2X NM 99.999% Integrity limit

True position of aircraft

Figure 2. RNP–X accuracy and integrity limits.

Figure 1. Aircraft navigation errors.

Flight technical error is the indicated lateral deviation of the aircraft position with respect to the defined path. RNAV systems in larger aircraft have provisions to couple a steering signal to a control system to automatically steer the aircraft to the intended path. In less equipped aircraft, the RNAV system simply provides a display indication of the crosstrack distance to the intended path, and the pilot manually provides the steering correction. RNP–RNAV STANDARDS In the interest of standardizing the performance characteristics of airborne navigation systems and the airspace, the concept of required navigation performance (RNP) for RNAV, denoted as RNP–RNAV, is being developed. Reference 1 provides the current state of the concept. Because the separation requirements for airspace depend on the proximity of obstacles, density of traffic, and other factors, the RNP–RNAV characteristic includes a measure, expressed in nautical miles (NM), that is correlated to the accuracy and integrity requirements for the airspace. To be more specific, the airspace or route will be defined as RNP-X, where X is the associated measure in nautical miles. This allows a consistent means of designation for airspace from the en route environment to the approach environment. The main navigation requirements for RNP–RNAV equipment are an accuracy requirement, an integrity requirement, and a continuity of function requirement. For RNP-X airspace, the accuracy requirement limits the crosstrack and alongtrack error of aircraft position to less than X NM 95% of the time. For RNP-X airspace, the integrity requirement limits the undetected position error to less than 2 times X 99.999% of the time. The continuity of function requirement limits the failure of the system to meet RNP–RNAV standards to less than 0.01% of the time. Figure 2 illustrates the accuracy and integrity limits for the RNP-X route.

United States the airways below 18,000 ft are designated as victor airways and have a prefix of V. Airways above 18,000 ft are designated as jet airways with a prefix of J. In other parts of the world, airways are prefixed with a letter (A1, G21, etc.) that is the first letter of a color (amber, green, etc.) Those airways are divided at different altitudes, and the upper airways are indicated with a prefix of U (UA1, for example). Airways have associated altitude restrictions to provide separation from terrain. In addition, published airways have certain conditional restrictions. The restrictions can be on the type of aircraft (only jet, for example) and on the direction of travel, and they can have restrictions that are effective for certain hours of the day.

AIRWAYS Published airways provide defined paths for much of en route airspace. Generally, airways are defined by great-circle segments terminated by VOR stations. In remote areas, nondirectional beacons (NDBs) are used in the airway structure. Figure 3 shows an aeronautical chart of airways. In the

Figure 3. Example of an airway chart.

AIRCRAFT NAVIGATION

355

For purposes of transitioning from one airway to another, the intersections of airways are often defined by named fixes. Navigation equipment can store the network of airways and intersections for use by the pilot in defining the path. This allows the pilot to enter the intended flight path in terms of the airway identifiers. Airborne equipment generally does not store directional or other conditional airway restrictions. For airways defined by VOR stations, the pilot is expected to navigate using the VOR at the closest end of the segment unless a changeover point (COP) is defined on the airway. The defined changeover point may not be at the midpoint of the airway segment to account for radio interference or other unique characteristics of the situation. Some airways are designated as RNAV airways and are available only to aircraft operating with RNAV equipment. Such airways do not have the restriction that a receivable VOR or NDB be used to define the great-circle path. It is expected that the RNAV equipment uses available navigation stations or GPS to compute the aircraft position. Because conventional non-RNAV airways are defined by VOR or NDB stations, traffic becomes concentrated near those stations. RNAV airways offer a significant advantage by allowing the airspace planner the ability to spread the aircraft traffic over a greater area without the installation and support of additional navigation stations. TERMINAL AREA PROCEDURES To provide a fixed structure to the departure and arrival of aircraft at an airport, published procedures are provided by the authorities. Such procedures are known as standard instrument departures (SIDs) and standard arrival routes (STARs). Figure 4 is an example of an SID chart. Generally, the instructions provided in SIDs and STARs are intended to be flown by the pilot without the aid of RNAV equipment. In order to incorporate the procedures into the RNAV equipment, the instructions must be reduced to a set of instructions that can be executed by the equipment. A subsequent section describes this process in more detail. Standard approach procedures are issued by the authorities to assist pilots in safe and standardized landing operations. The generation of the approach procedures accounts for obstacles, local traffic flow, and noise abatement. Historically, the approach procedures are designed so that RNAV equipment is not required. That is, the pilot can execute the approach using basic sensors (VOR, DME, ADF) until landing visually. For operations in reduced visibility situations, there are Category II and III instrument landing system (ILS) approaches that require automatic landing equipment. In addition, there are RNAV and global positioning system (GPS) approaches that require RNAV equipment. Modern RNAV equipment is capable of storing the defined approach path and assist the pilot in flying all approaches. Figure 5 is an example of an approach chart. NAVIGATION SENSOR SYSTEMS RNAV equipment receives information from one or more sensor systems and forms an estimate of the aircraft position. If more than one sensor type is available, the position estima-

Figure 4. Example of an SID chart.

tion algorithm will account for the quality differences and automatically use the data to generate a best estimate of position. Complementary filters or Kalman filters are commonly used to smooth and blend the sensor data. The common sensors used for position estimation are GPS, DME, LORAN, VOR, and IRS. The data from each of the sensor types have unique characteristics of accuracy, integrity, and availability. In addition, each of the sensor types requires unique support functions. Sensor Accuracy The accuracy characteristic of a sensor can be expressed as the 95th percentile of normal performance. For any specific sensor, the wide variation in conditions in which it can be used makes it difficult to generalize the accuracy with specific numbers. The following data represent the accuracy under reasonable conditions.

356

AIRCRAFT NAVIGATION

provide integrity is with redundant measurements. By comparison of the redundant measurements, an error in one of the measurements can be detected and in some cases removed from consideration. GPS has a function known as receiver autonomous integrity monitoring (RAIM), which provides integrity. This function can be used when sufficient signals of satellites are available. This is usually the case when the GPS receiver is receiving signals from five or more satellites. The status of RAIM is provided to the RNAV equipment and is important in approach operations using the GPS sensor. For RNAV systems that use VOR and DME signals, if there are not redundant signals available, the position solution is vulnerable to the effects of radio signal multipath and to the navigation database integrity. The DME signal multipath problem occurs in situations where the local terrain supports the reflection of the radio signal to or from the DME station. The navigation database integrity is difficult to ensure, especially for DMEs that are associated with military TACANs. Military TACANs are sometimes moved, and the information does not get included in the navigation database in a timely fashion. NAVIGATION COORDINATE REFERENCE

Figure 5. Example of an approach chart.

GPS has an accuracy of better than 0.056 NM in approach conditions, with some degradation allowed at higher speeds. DME range is accurate to about 0.1 NM with some degradation for longer ranges. The accuracy of a position estimate based on two or more DME ranges will be dependent upon the geometry of the DME stations relative to the aircraft. LORAN accuracy is about 0.25 NM when receiving a good ground wave signal. VOR bearing is generally accurate to within 2⬚. When used as a position sensor, the position estimate accuracy is dependent upon the range to the VOR station. IRS accuracy is dependent upon the time since alignment and the accuracy of the entry of the position at alignment. An accuracy of better than 2 NM/h since alignment is normal. Sensor Integrity Integrity is the ability of the system to warn the pilot of significant errors in a timely manner. The most common way to

The WGS-84 ellipsoid has become the standard for aeronautical navigation. This reference can be viewed as a surface of revolution defined by a specified ellipse rotated about the earth polar axis. The semimajor axis of the ellipse lies in the equatorial plane and has a length of 6378137.000 m. The semiminor axis is coincident with the earth polar axis and has a length of 6356752.314 m. Paths between two fixes on the WGS-84 spheroid are defined as the minimum distance path along the surface, known at the geodesic path between the two points. In general, the geodesic path does not lie on a plane but has a geometric characteristic of torsion. However, for reasonable distances, there is no significant error by approximating the path as a portion of a great circle of the appropriate radius. Most of the fixes defined in the world were specified in a reference system other than WGS-84. An effort is under way to mathematically convert the data from the original survey coordinate system to that of the WGS-84 coordinate system. At the same time, when possible, the survey of the location is being improved. COURSE OF THE GREAT CIRCLE PATH The basic path for airways is a direct path between two fixes, which may be a VOR station, an NDB station, or simply a geographical location. In terminal area procedures the most common path is defined by an inbound course to a fix. The RNAV equipment approximates such paths as segments of a great circle. Considering the case of a path defined as a radial of a VOR, the actual true course depends upon the alignment of the VOR transmitter antenna with respect to true north. The angular difference between the zero degree radial of the VOR and true north is called the VOR declination. When the VOR station is installed, the 0⬚ VOR radial is aligned with the magnetic north so the VOR declination is the same as the magnetic variation at the station at the time of installation.

AIRCRAFT NAVIGATION

Magnetic variation is the difference between the direction of north as indicated by a magnetic compass and true north defined by the reference ellipsoid. As such, it is subject to the local anomalies of the magnetic field of the earth. The magnetic field of the earth varies in a systematic manner over the surface of the earth. It is much too complex to be defined as a simple bar magnet. The magnetic field is also slowly changing with time in a manner that has some random characteristics. Every 5 years a model, both spatial and temporal, is defined by international agreement using worldwide data. A drift of magnetic variation of 1⬚ every 10 years is not uncommon on the earth. The model is defined in terms of spherical harmonic coefficients. Data from this model are used by sensors and RNAV systems to calculate the magnetic variation at any location on the earth. In particular, inertial navigation systems are references to true north and produce magnetically referenced data by including the local magnetic variation as computed from a magnetic variation model. Because the magnetic variation of the earth is slowly changing, a VOR whose 0⬚ radial is initially aligned with the magnetic north will lose this quality after a period of time. This discrepancy between the VOR declination and the local magnetic variation is one reason for ambiguity in course values. As one progresses from along the great circle path, the desired track changes due to the convergence of the longitude lines and due to the magnetic variation. Figure 6 shows the effect of position on true and magnetic courses. The true course at the fix, CT, is different from the true course, C⬘T, at the aircraft because the longitude lines are not parallel. The difference in the magnetic courses is the result of the difference in the true courses together with the difference in magnetic variation at the two locations. For the pilot, an important piece of information is the magnetic course to be flown to stay on the great circle path. With no wind, when on track, the current magnetic heading of the aircraft should agree with the displayed magnetic course. To achieve this goal, the RNAV equipment first computes the true course of the desired path and then adjusts it for local magnetic variation. On the aeronautical charts, the magnetic course of the path is defined as the termination point of the path. When the aircraft is some distance from the termination point, both the true course and the magnetic variation are different. This causes the FMS to display a magnetic course at the aircraft that is different than that of the chart. As explained above,

True north

True north

Magnetic north

Magnetic north CT C ′T

CM

C ′M

357

this difference is necessary to provide a display of course that is consistent with the magnetic heading of the aircraft as it progresses along the path. ARINC-424 NAVIGATION DATABASE The navigation database installed in the RNAV system stores information about airways, SIDs, STARs, approaches, navigational aids, and so on. Such information changes continually as navigational aids are removed or installed, airports are improved, and so on. To ensure that the pilot has current data, new data become effective every 4 weeks by international convention. Because the aircraft may not be available for database update at the day the new data become effective, most RNAV systems have provisions to allow the new data to be loaded several days before it is to become effective. In effect, the RNAV system stores two databases, and the day of flight is used to determine the database that is effective for the flight. An international standard for the interchange of navigational database information is encompassed in the ARINC specification 424 entitled Navigation System Data Base. This specification provides for standardized records of 132 ASCII characters. Record formats are provided to store a wide set of navigational information. RNAV systems have packing programs that process ARINC-424 records into packed data that are loaded into the airborne equipment. The packing programs select only those records that are applicable to the RNAV system and are in the desired geographic area. It must also be ensured that the selected subset of data is consistent; that is, all references to other records are satisfied in the subset. Finally, the selected data are packed in the format required for the particular RNAV system. The reduction of terminal area procedures to a set of instructions that can be automatically flown by the RNAV equipment is particularly complex. A considerable fraction of the ARINC-424 specification is devoted to this issue. A set of leg types have been established to encode terminal area procedures. Each leg type has a path definition and a termination definition. The intended flight path is encoded as a sequence of legs. The RNAV equipment will automatically fly the procedure by processing the sequence of leg types. As each leg becomes active, the path definition of that leg will form the current flight path intent, and the termination definition will provide information when the successor leg is to become active. Table 1 lists the 23 leg types defined by the ARINC-424 specification. Note that generally the first letter of the leg type can be associated with the intended path and the second letter can be associated with the termination of the path. Leg types CA, CD, CI, and CR are provided to handle instructions such as ‘‘fly 310⬚ track until . . .,’’ whereas leg types VA, VD, VI, VM, and VR will handle similar instructions such as ‘‘fly 310⬚ heading until . . ..’’ These leg types have no specified geographic path but will cause the aircraft to be steered to the proper track or heading from the current position of the aircraft whenever the leg becomes active. The other leg types are referenced to some geographic location. Limitation of ARINC-424 Coding

Figure 6. True and magnetic courses vary with position.

Using the ARINC-424 leg types, most terminal area procedures can be encoded in such a way that the RNAV equip-

358

AIRCRAFT NAVIGATION

Missed approach path

Table 1. ARINC-424 Leg Types Leg Type

Path and Termination Description

AF CA CD CI CR

Fly a constant DME arc path to the fix Fly the specified course to the altitude Fly the specified course to a distance from a DME Fly the specified course to intercept the following leg Fly the specified course until crossing a specified VOR radial Fly the specified course into the fix Fly directly to the fix Fly the specified course from the fix to an altitude Fly the specified course from the fix for a specified distance Fly the specified course from the fix to a distance from a DME Fly the specified course from the fix until manually terminated Fly the holding pattern until terminated at an altitude Fly the holding pattern course reversal, terminated after entry maneuver Fly the holding pattern until terminated manually An initial fix (no path defined) Fly a procedure turn course reversal Fly the great circle path defined by two fixes Fly the arc defined by a constant radius to a fix Fly the specified heading to an altitude Fly the specified heading to a distance from a DME Fly the specified heading to intercept of following leg Fly the specified heading until terminated manually Fly the specified heading until crossing a VOR radial

CF DF FA FC FD FM HA HF HM IF PI TF RF VA VD VI VM VR

ment can generally fly the procedure in a fashion that is similar to the pilot navigation. However, there are significant limitations to this concept. First, the concept assumes that the RNAV equipment has sufficient sensor data to accomplish the proper steering and leg terminations. Lower-end RNAV systems designed for smaller aircraft often do not have sensors providing heading or barometric altitude. Without a heading sensor, the system cannot fly the heading legs properly. Substituting track legs for heading legs is not always satisfactory. In the same way, legs that are terminated by an altitude (CA, FA, VA, and HA) require that the RNAV system have access to barometric altitude data. The use of geometric altitude determined by GPS data will introduce several errors. The geometric altitude ignores the nonstandard state of the pressure gradient of the atmosphere. The geometric altitude ignores the undulations of the mean sea level. Finally, the GPS sensor is accurate in the vertical axis to about 150 m, which is less accurate than altimeters. A second limitation to the concept of using the ARINC-424 leg types has to do with the diversity of instructions that may

Runway transitions

En route transitions Common route

Figure 7. Data structure for SIDs and STARs.

Approach transitions Final approach path

Figure 8. Data structure for approaches.

appear on the procedure chart. Because the chart is written with the pilot in mind, the chart may include logical instructions that cannot be coded with the 24 leg types of ARINC424 specification. An instruction such as ‘‘fly to an altitude or DME distance, whichever occurs first’’ cannot be encoded as a sequence of ARINC-424 legs. Current charts exhibit a wide variety of logical instructions involving altitude, DME distance, aircraft category, landing direction, and the like. Many of these instructions cannot be directly encoded as a sequence of ARINC-424 legs. ARINC-424 Procedure Database Structures SIDs can be defined in such a manner that a single identifier implies a single path. In other cases, a single identifier can be used to describe the departure paths from more than one runway. In such cases, the departure path specification must include the runway together with the SID identifier. In addition, the single identifier can further be used to describe the path to several en route terminations. The multiple optional paths are known as transitions. In the most general case, a runway transition path is linked to a common route and then linked to an en route transition path. The complete path is therefore specified by the SID identifier, the runway identifier, and the termination of the en route transition. To allow the encoding of the complete set of options, the ARINC-424 specification incorporates a database structure similar to that shown in Fig. 7. STARs can have the same structure as SIDs, with the first leg of the star beginning at the first fix on the en route transition. With the complete SID or STAR encoded in the navigation database, the RNAV system allows the pilot to select the proper runway and en route transition and links a single path from the selection. SIDs and STARs in the United States commonly use the branched structure. Outside the United States, this is generally not the case. That is, a single identifier is used to define the complete path from the runway to the final en route termination with no optional segments. The general structure for approaches is a set of en route transitions followed by a common route. The common route includes both the final approach path and a single missed approach path. Virtually all approaches throughout the world have this structure (Fig. 8).

BIBLIOGRAPHY 1. RTCA Inc., Minimum Aviation System Performance Standards for Required Navigation Performance RNP-RNAV, DO-236, current ed., Washington, DC: RTCA Inc.

AIR POLLUTION CONTROL Reading List ARINC, Inc., Navigation Data Base Specification 424, current ed., Annapolis, MD: Aeronautical Radio. M. Kayton and W. R. Fried, Avionics Navigation Systems, 2nd ed., New York: Wiley, 1997.

GERALD E. BENDIXEN Rockwell Collins, Inc.

AIR DEFENSE. See ELECTRONIC WARFARE. AIRPLANE MAINTENANCE. See AIRCRAFT MAINTENANCE.

359

ATTITUDE CONTROL

45

stant cruising altitude, to perform a turn without excessive side forces, or to land in bad weather or low visibility. All these tasks are performed with the help of an attitude control system. An attitude control system is a collection of attitude sensors, actuators, and an attitude control law. The objective of an attitude control system is to acquire information about the orientation of the vehicle from the attitude sensors, process this information through an attitude control law, and generate a series of commands that will change or keep the attitude of the vehicle at a specified orientation. An attitude control law is an algorithm (i.e., a series of commands or instructions), which calculates the necessary action to be taken based on the knowledge of the current and anticipated attitude. This algorithm is usually executed on a digital or an analog computer. When the control law simply follows a set of prespecified steps stored in the computer’s memory or given to the computer by a human operator we refer to the control algorithm as an open-loop control law. If the computer makes its decisions without external intervention, solely on attitude measurements from its sensors, the control law is referred to as feedback or closed-loop attitude control law. ATTITUDE CONTROL OF SPACECRAFT A rigid satellite or spacecraft in orbit offers the most obvious example of a rotating rigid body. Attitude control for spacecraft arises in the process of orienting the spacecraft along a specified, predetermined direction. According to Wertz (1), it consists of two problems—attitude stabilization, or maintaining a specific orientation, and attitude maneuver control, or, controlling the spacecraft from one attitude orientation to another. Attitude orientation is specified either with respect to an inertial reference frame or with respect to another moving object. For instance, attitude stabilization of a spacecraft with one axis toward the Earth implies a continuously changing orientation with respect to an inertial frame. There are two main methods for spacecraft stabilization: (1) passive methods, and (2) active methods. Passive Stabilization

ATTITUDE CONTROL Attitude control is the field of engineering science that deals with the control of the rotational motion of a rigid body about a reference point (typically the center of mass). Attitude control systems are commonly used in controlling the orientation of spacecraft or aircraft. As a spacecraft orbits the Earth, it may have to move in space in such a way that its antenna always points to a ground station for communication or its on-board telescope keeps pointing to a distant star. A fighter aircraft may be required to turn very fast and maneuver aggressively to shoot down enemy airplanes or to avoid an incoming missile. A civilian airplane may need to keep a con-

Passive methods require no power consumption or external control. The stabilization is achieved naturally through the physical properties of the motion. Two typical methods of passive stabilization are gravity-gradient stabilization and spin stabilization. Gravity-gradient stabilization is based on the natural balancing torque due to the gravity differential at two distinct points of a body at different distances from the center of the Earth. It is a particularly effective way for stabilization of elongated structures at low Earth orbit where the gravity pull of the Earth is stronger. The result of this stabilization method is to keep the long dimension of the structure along the local vertical (the direction to the center of the Earth). Spin stabilization on the other hand, takes advantage of the natural tendency of the angular momentum vector to remain inertially fixed in the absence of external torques. The term gyroscopic stiffness is often used to describe this property of the angular momentum vector. A child’s familiar spinning top is based on the same principle. Spin stabilization aims to keep the axis of rotation (spin axis) and the angular momentum vector parallel. This ensures that the spin axis remains

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

46

ATTITUDE CONTROL

inertially fixed. If the spin axis and the angular momentum vector are not parallel, the spacecraft is said to exhibit nutation, which manifest itself as a wobbling motion. In the presence of damping (i.e., energy dissipation) the vehicle spin axis tends to align itself with the angular momentum axis. In practice nutation dampers are used to introduce artificial damping in order to align the spin and the angular momentum axis and thus keep the spin axes constant in inertial space. Gravity-gradient or spin stabilization cannot be used to control the body about the gravity vector or the spin axis. In addition, it may not always be possible to use spin stabilization. For example, mission requirements may demand that the communications antenna always point toward Earth, or the solar panels point always toward the sun. In this case, it is necessary that the antenna and the solar panels be stationary with respect to an inertial frame. They cannot be part of a continuously spinning satellite. The solution to this problem is the use of dual spin spacecraft or dual spinners, which consist of two parts, the rotor and the stator. The rotor rotates about its axis and provides the angular momentum necessary for stabilization as with case of the spin-stabilized spacecraft. The stator remains fixed and contains all the scientific instruments that have to remain inertially fixed. Thus, dual-spin spacecraft combine scanning (rotating) and pointing (inertially fixed) instruments in one platform. This clever solution comes at the expense of increased complexity of the spacecraft design and its operation, however. A momentum bias design is very common for dual-spin satellites in low-Earth orbit, in which the rotor is mounted along the normal to the orbit plane. This allows the instruments to scan the Earth. Other common methods of passive stabilization include magnetic torques or use of solar panels. Active Stabilization Although simple and cheap, passive stabilization schemes have two main drawbacks: First, they achieve pointing accuracy of the controlled axis only up to a few degrees. Several applications (e.g., communications satellites, space telescopes, etc.) require accuracy of less than a few arc seconds (1 arcsecond ⫽ 1/3600 deg). Second, control systems based on passive schemes cannot be used effectively to perform large attitude maneuvers. Reorientation of the spin axis for a spinning spacecraft, for instance, requires excessively large control torques to move the angular momentum vector. Also, gravitygradient and magnetic torques are limited by the direction of their respective force fields and, in addition, are not strong enough to be used for arbitrary, large angle maneuvers. Both of the previous problems encountered in the use of passive stabilization schemes can be resolved using active stabilization methods. The most common active control methods incorporate use of gas actuators or momentum wheels. Both can be used to achieve three-axis stabilization, that is, active control of the spacecraft orientation about all three axes, as well as three-axis large angular (slew) maneuvers. Gas actuators use a series of gas nozzles distributed (usually in pairs) along the three perpendicular axes of the spacecraft. Gas jets are classified either as hot gas jets (when a chemical reaction is involved) or cold gas jets (when no chemical reaction is present). The gas jets (thrusters) are usually of the on–off type. Continuously varying control profiles can be

generated, however, using pulse-width pulse-frequency (PWPF) modulators. These modulators produce a continuously varying control torque by generating a pulse command sequence to the thruster valve by adjusting the pulse width and pulse frequency. The average torque thus produced by the thruster equals the demanded torque input. This will wear out the jet valves in the long run. A better choice for generating continuously varying torques is the use of momentum wheels. Gas jets achieve stabilization by generating external torques, which change the total angular momentum of the spacecraft. Alternatively, flywheels can be used to generate internal torques or redistribute the angular momentum between the main vehicle and the wheels. The total angular momentum of the vehicle plus the wheels remains constant in this case. This is akin to a gymnast throwing a somersault. While in the air, the gymnast’s angular momentum is constant. The gymnast changes position and rotates in midair by redistributing the angular momentum by extending or contracting the arms, bending at the waist, and so on. Momentum exchange devices (sometimes collectively referred to as momentum wheels) are also preferable for application of continuously varying torques. There are three main types of actuators that use momentum exchange for attitude control. 1. Reaction wheels do not rotate under normal conditions. When an angular maneuver is commanded or sensed, the reaction wheel spins in the opposite direction to the sensed or commanded rotation. Thus a reaction wheel provides a torque along the wheel spin axis. A minimum of three reaction wheels is necessary to control the attitude about all three axes. 2. Momentum wheels spin at a constant speed under normal conditions, and are used to increase stability about the corresponding axis. A dual-spin spacecraft, for example, is a special case of a spacecraft with a momentum wheel about the axis of symmetry. A minimum of three wheels are necessary to achieve stability about all three axes. 3. Control moment gyros (CMG) consist of a single spinning flywheel that is gimballed and free to rotate about two or three perpendicular axes. Contrary to the momentum wheel, the magnitude of the angular velocity vector remains constant. The torque produced is proportional to the change in the direction of the angular momentum vector. A more complete discussion on the use of momentum wheels in attitude control problems can be found elsewhere (1,2). Attitude Sensors As mentioned earlier, an attitude control system requires information about the body orientation. This information is provided by attitude sensors. An attitude sensor actually provides the relative orientation of the spacecraft with respect to a reference vector (e.g., a unit vector in the direction of the Sun, a known star, the Earth, or the Earth’s magnetic field). Therefore, three-axis attitude determination requires two or more sensors. The definitive reference for a more in-depth discussion of attitude sensors and actuators and their principles

ATTITUDE CONTROL

of operation is Wertz (1). This reference also includes a fairly detailed overview of hardware implementation issues. Sun Sensors. Sun sensors are the most common type of attitude sensor. Their field of view ranges from a few square arcmin (10⫺7 rad2) to approximately 앟 rad2 and their resolution ranges from several degrees to less than 1 arc-sec. Horizon Sensors. Horizon sensors can be used when the spacecraft is in a close orbit around a celestial body. For lowEarth orbiting satellites, for instance, the difference in brightness of Earth’s disk from the background darkness of space can be easily detected by a horizon sensor and provides a coarse attitude measurement. Magnetometers. Magnetometers use Earth’s magnetic field to locate the body’s orientation. They have poor resolution due to uncertainty of Earth’s magnetic field. They work better at low-Earth orbits, where the magnetic field is stronger and better modeled. Star Sensors. Star sensors provide attitude information of very high accuracy, but they are heavy and expensive. They are usually the choice for deep-space spacecraft where attitude measurements from Sun sensors or near-by celestial objects are either unavailable or inaccurate. Gyroscopes. Gyroscopes (or simply gyros) use a rapidly spinning mass on a gimbal to sense changes in the spacecraft orientation. The principle of their operation is based on the fact that any change in the angular momentum vector about a certain axis will be sensed as a resulting movement about a perpendicular axis. They are very accurate for limited time intervals, but their measurements may become inaccurate over long periods of time due to drift. In this case, they have to be combined with some other attitude sensor to reset them periodically. Gyroscopes can also be used to get angular velocity measurements; in this case they are called rate gyros. Apart from their use in spacecraft, gyroscopes are also used as attitude or angular velocity sensors in aircraft, missiles, or marine vehicles (e.g., submarines).

lates and rotates at the same time. For an observer located on Earth, the spacecraft can be thought of as a single particle, and its trajectory through space is primarily determined by its instantaneous position and velocity. Its orientation with respect to Earth is irrelevant (unless a communication antenna needs to be always pointing toward Earth) or very difficult to observe from such a large distance. On the other hand, as the same observer moves closer and closer to the spacecraft, the orientation of the spacecraft may be very important, and it certainly becomes much more obvious. Translational motion therefore deals with the motion of particles, that is, idealized points with zero dimensions but nonzero mass. Rotational motion, on the other hand, deals with the motion of rigid bodies, that is, physical objects with nonzero dimensions and nonzero mass. One can safely think of translation as the macroscopic, or far-away view of the motion of an object and of rotation as the microscopic, or closeby view of the motion. Dynamics of the Attitude Motion The dynamic equations of the attitude motion of a rotating body describe the behavior of the angular momentum or the angular velocity vector as a function of the externally applied torques or moments. The basic equation that governs the body behavior under external torques is Euler’s equation of motion (2). It states that the rate of change of the angular momentum vector H of the body at every instant is equal to the applied moment M H dH =M dt

(1)

where the angular momentum is defined by H= r × v dm

(2)

B

and the integration extends over the entire body. In Eq. (2) the vector v ⫽ r˙ denotes the inertial velocity of the mass element dm (see Fig. 1).

Z

DYNAMICS OF ROTATING BODIES The motion of a rigid body in three-dimensional space is determined by the forces and moments acting on the body at each instant of time. There are two main types of general motion of a rigid body: (1) translational motion, which typically deals with the velocity and position of the center of mass of the body, and (2) rotational or attitude motion, which typically deals with the (angular) velocity and (angular) position about the center of mass. The angular position about the center of mass is often referred to as the attitude or orientation of the body. The choice of the center of mass as the reference point to describe the general body motion is not restrictive, but it has the advantage of allowing the rotation and translation to be treated separately. That is, the translation of the body center of mass does not affect nor is it affected by the rotation of the body about the center of mass. To understand the two types of motion, consider the example of a spacecraft traveling in space. The spacecraft trans-

47

k cm

i

j R K

dm r O

I

J Y

X Figure 1. Inertial and body-fixed reference frames. The body-fixed reference frame is located at the center of mass cm. The vector r denotes the location of the element mass dm in the inertial frame and the vector ␳ denotes the location of the mass in the body frame.

48

ATTITUDE CONTROL

The time derivative in Eq. (1) must be taken with respect to an inertial reference frame. In addition, in the calculation of H we have a choice of the reference point about which to calculate the moments and the angular momentum. Equation (1) implies that we have chosen either a point fixed in inertial space or the center of mass. The center of mass offers the most convenient and natural choice. In this case, from Fig. 1 we have that the location of the mass element dm is at r ⫽ R ⫹ ␳, where R denotes the location of the center of mass. Differentiation of a vector V, as seen from an inertial frame, is related to differentiation as seen in a moving (body) frame through the relationship V V dV dV = + ω ×V dt 1 dt B

(3)

where ␻ is the angular velocity of the moving frame. The velocity of the mass element dm is thus given by v = R˙ + ω × ρ

(4)

Subsequently, the angular momentum vector is given by

˙ dm + ρ ×R

H=

B

ω × ρ ) dm ρ × (ω

(5)

B

The first integral in the previous expression vanishes, because the origin is at the center of mass

Jx =

(y2 + z2 ) dm,

Jxy =

B Jy =

(x2 + z2 ) dm,

Jxz =

xz dm

ρ dm = 0

(6)

B

The angular momentum vector with respect to the center of mass is thus ω × ρ ) dm ρ × (ω

H=

(7)

B

Since the position vector ␳ changes with time in an inertial frame, it is beneficial to choose a reference frame fixed in the body, since in this case the mass distribution does not change. Therefore, choosing a reference frame fixed in the body and located at the center of mass, we can express ␻ and ␳ in this body-fixed frame as ω = ωx i + ωy j + ωz k,

ρ = xi + yj + zk

(8)

Then Eq. (7) yields ω H = Jω

(9)

where J is the moment-of-inertia matrix and is given by

Jx  J = −Jxy −Jxz

−Jxy Jy −Jyz

 −Jxz  −Jyz  Jz

(10)

(11)

B

B

Jz =

xy dm B

(x2 + y2 ) dm,

Jyz =

B

yz dm B

The moment-of-inertia matrix depends on the shape of the body and the manner in which its mass is distributed. The larger the moments of inertia the greater resistance the body will have to rotation. When using Eq. (3) and recalling that in a body-fixed frame the inertia matrix is constant, it follows that Eq. (1) can be written as ω =M Jω˙ + ω × Jω

(12)

The inertia matrix J, also called the inertia tensor, is symmetric and positive definite. One can therefore choose a reference frame such that the matrix J is diagonal. This particular choice of body-fixed axes is called the axes of principal moments of inertia. The directions of these axes are exactly those determined by the eigenvectors of the matrix J. The components of Eq. (12) resolved along the principal axes are given by

Jx ω˙ x = (Jy − Jz )ωy ωz + Mx Jy ω˙ y = (Jz − Jx )ωz ωy + Mz



where

(13)

Jz ω˙ z = (Jx − Jy )ωx ωy + Mz where Jx, Jy, Jz are the three principal moments of inertia (the eigenvalues of the matrix J), 웆x, 웆y, 웆z are the components of the angular velocity vector along the principal axes, as in Eq. (8), and Mx, My, Mz are the components of the applied moment along the same set of axes, i.e., M ⫽ Mxi ⫹ My j ⫹ Mzk. Equation (12) or Eq. (13) is the starting point for most attitude control problems. Kinematics of the Attitude Motion The solution of Eq. (13) provides the instantaneous angular velocity of the body about its center of mass. It does not capture the instantaneous orientation of the body with respect to, say, the inertial reference frame. In particular, integration of the angular velocity vector 웆 does not, in general, give any useful information about the orientation of the body. The orientation of the body is completely determined if we know the orientation of the body-fixed frame with respect to the inertial reference frame, used in deriving Eq. (13). The rotation matrix R between the body-fixed and the inertial reference frames is used to completely describe the body orientation. The rotation matrix is a 3 ⫻ 3 matrix having as columns the components of the unit vectors of the inertial frame expressed in terms of the unit vectors of the body-fixed frame. In other words, i, j, k denote the unit vectors of the body frame and I, J, K denote the unit vectors in the inertial frame, a vector V having coordinates (Vx, Vy, Vz) and (VX, VY, VZ) with

ATTITUDE CONTROL

respect to the body-fixed and inertial frames, respectively, can be written as V = Vx i + Vy j + Vz k = VX I + VY J + VZ K

(14)

The matrix R establishes the following relation between the coordinates of V in the two reference frames     VX Vx     (15) Vy  = R VY  Vz VZ B

dR = S(ω)R dt where S(웆) is the skew-symmetric matrix (S ⫽ ⫺ST)   −ωy 0 ωz   S(ω) = −ωz 0 ωx  ωy −ωx 0

(16)

(17)

It can be shown that the matrix R is orthogonal, that is, it satisfies (18)

and it is also proper, that is, its determinant is ⫹1. Equation (16) can also be used to calculate the angular velocity if the rate of change of R is known ˙ T S(ω) = RR

rotations. Several definitions of the Euler angles are possible, depending on the choice of the axes of rotations and the particular order in which the rotations are performed. In fact, there are 12 different possible choices of Euler angle sets. These sets have been discussed in great detail elsewhere (3). One of the most common choices in aircraft and spacecraft attitude control problems is the (3-2-1) Euler angle sequence. According to this set of Euler angles, the orientation of the body-fixed reference frame with respect to the inertial reference frame is described by a sequence of the following three elementary rotations:

I

If ␻ = 웆xi ⫹ 웆y j ⫹ 웆zk denotes the angular velocity of the body frame with respect to the inertial frame (expressed in the body frame coordinates), the differential equation satisfied by R is given by (3,4)

RRT = RT R = I

49

(19)

1. First rotate the inertial reference frame about its Z axis through an angle ␺ to the new axes x⬘ ⫺ y⬘ ⫺ z⬘. 2. Then rotate about the y⬘ axis by an angle ␪ to the new axes x⬙ ⫺ y⬙ ⫺ z⬙. 3. Finally rotate about the x⬙ axis by an angle ␾ to the final body-axes x, y, z. This sequence of rotations that takes the inertial frame to coincide with the body frame after three rotations is depicted in Fig. 2. Note that the order of rotations is very important. The angles ␾, ␪, and ␺ are called the roll, pitch, and yaw angles. The elementary rotations of a reference frame about the axes x, y, and z are given, respectively, by   1 0 0   Rx (φ) = 0 cos φ sin φ  0 − sin φ cos φ   cos θ 0 − sin θ   1 0  Ry (θ ) =  0 (20) sin θ 0 cos θ   cos ψ sin ψ 0   Rz (ψ ) = − sin ψ cos ψ 0 0 0 1

We can use Eq. (16) to find the orientation of the body at any instant of time if the corresponding angular velocity vector ␻ of the body is known. In particular, the matrix differential equation in Eq. (16) can be integrated from the known initial attitude of the body to propagate the attitude for all future times. This process will require the integration of the nine linear but time-varying differential equations for the elements of the matrix R in order to obtain R(t) at each time t. Careful examination of the matrix R, however, reveals that the nine elements of this matrix are not independent from each other, since the matrix R must necessarily satisfy the constraints in Eq. (18). An alternative approach to solving Eq. (16) is to parameterize the matrix R in terms of some other variables and then use the differential equations of these variables in order to propagate the attitude history. Euler Angles. The minimum number of parameters that can be used to parameterize all nine elements of R is three. [Notice that Eq. (18) imposes six independent constraints among the elements of R.] The Euler angles are the most commonly used three-dimensional parameterization of the rotation matrix R. They have the advantage that they are amenable to physical interpretation and can be easily visualized. Using the Euler angles we can describe the final orientation of the body-axis frame by three successive elementary

Z = z′

z

z′′ ψ φ θ

y k j

φ ψ

θ

y′= y′′ Y

i

ψ θ

X

φ

x′ x′′= x

Figure 2. Euler angle sequence (3-2-1). We align the inertial and body frames by first rotating with an angle ␺ about the z axis, then rotating with an angle ␪ about the new y axis, and finally rotating with an angle ␾ about the x axis.

50

ATTITUDE CONTROL

The rotation matrix in terms of the (3-2-1) Euler angles can thus be expressed in terms of the three previous elementary rotations by R(ψ, θ, φ) = Rx (φ)Ry (θ )Rz (ψ )

(21)

and thus



cos ψ cos θ  R(ψ, θ, φ) = − sin ψ cos φ + cos ψ sin θ sin φ sin ψ sin φ + cos ψ sin θ cos φ sin ψ cos θ cos ψ cos φ + sin ψ sin θ sin φ − cos ψ sin φ + sin ψ sin θ cos φ

 − sin θ  cos θ sin φ  cos θ cos φ

(22)

The components of the angular velocity vector in the bodyframe are given in terms of the rates of these Euler angles by

ωx = −ψ˙ sin θ + φ˙ ωy = ψ˙ cos θ sin φ + θ˙ cos φ

(23)

The Euler parameters satisfy the constraint q20 + q21 + q22 + q23 = 1

(26)

q = q0 + q1 i + q2 j + q3 k

(27)

The quantity

defined from the Euler parameters is called the quaternion (4,7). It should be pointed out that there is often a confusion in the literature about this term. With a slight abuse of terminology, many authors refer to the Euler parameters q0, q1, q2, q3 as the quaternion although, strictly speaking, this is incorrect. The rotation matrix in terms of the Euler parameters is given by

R(q0 , q1 , q2 , q3 )  2 q0 + q21 − q22 − q23  =  2(q1 q2 − q0 q3 ) 2(q1 q3 + q0 q2 )

2(q1 q2 + q0 q3 ) q20 − q21 + q22 − q23 2(q2 q3 − q0 q1 )

ωz = ψ˙ cos θ cos φ − θ˙ sin φ Conversely, we can solve the previous equations and express the rates of the Euler angles in terms of the angular velocity components in the body-frame

φ˙ = ωx + ωy tan θ sin φ + ωz tan θ cos φ θ˙ = ωy cos φ − ωz sin φ ψ˙ = ωy sec θ sin φ + ωz sec θ cos φ

(24)

The previous equation indicates that there is a singularity when ␪ ⫽ ⫾앟/2. This singularity does not allow for a global parameterization of the attitude using the Euler angles. The previous (3-2-1) Euler sequence, for example, is defined only for ⫺앟 ⱕ ␾ ⱕ 앟, ⫺앟/2 ⬍ ␪ ⬍ 앟/2, and ⫺앟ⱕ ␺ ⱕ 앟. Other threedimensional parameterizations include the Cayley-Rodrigues parameters and the modified Rodrigues parameters (4). However, the singularity problem is always present, when using a three-dimensional parameterization of the rotation matrix R (5). Higher order parameterizations need to be used to avoid singularities. Euler Parameters (Quaternions). A four-dimensional parameterization of the attitude kinematics that does not have any singularities is given by the Euler parameters. The Euler parameters are defined via Euler’s theorem, which can be stated as follows (6): The most general displacement of a rigid body with one point fixed is equivalent to a single rotation about some axis through that point.

The corresponding axis is called the eigenaxis of rotation, and and the corresponding angle is called the principal angle. If the eigenaxis unit vector is e ⫽ e1i ⫹ e2j ⫹ e3k and the principal angle is ⌽, the Euler parameters are defined by q0 = cos , 2

qi = ei sin , 2

i = 1, 2, 3

(25)

 2(q1q3 − q0 q2 )  2(q2q3 + q0 q1 )  q20 − q21 − q22 + q23 (28)

and the corresponding kinematic equations are given by      q˙ 0 q0 0 −ωx −ωy −ωz q˙     0 ωz −ωy   1  1 ω x  q 1   =    (29) q˙ 2  2 ωy −ωz 0 ω x  q 2       q˙ 3 ωz ωy −ωx 0 q3 These equations are linear and do not involve any trigonometric functions as the corresponding kinematic equations in terms of the Euler angles. Integration of these equations to obtain attitude information can thus be performed very fast on a computer. In addition, the attitude description in terms of q0, q1, q2, q3 is global and nonsingular. For these reasons the Euler parameters have increasingly gained popularity in many attitude-control applications. The main disadvantage when using the Euler parameters is that they are difficult to visualize. The orientation needs to be transformed to an Euler angle sequence if they are to be meaningful, for example to a pilot or an engineer. The Eulerian angles (␾, ␪, ␺) in terms of the Euler parameters can be computed, for example, from

sin θ = 2(q1 q3 − q0 q2 ) tan ψ =

2(q1 q2 + q0 q3 ) q20 + q21 − q22 − q23

tan φ =

2(q2 q3 + q0 q1 ) q20 − q21 − q22 + q23

The Euler parameters are related to the (3-2-1) Euler angles by

q0 = cos(φ/2) cos(θ/2) cos(φ/2) + sin(φ/2) sin(θ/2) sin(ψ/2) q1 = sin(φ/2) cos(θ/2) cos(ψ/2) − cos(φ/2) sin(θ/2) sin(ψ/2) q2 = cos(φ/2) sin(θ/2) cos(ψ/2) + sin(φ/2) cos(θ/2) sin(ψ/2) q3 = cos(φ/2) cos(θ/2) sin(ψ/2) − sin(φ/2) sin(θ/2) cos(ψ/2) (30)

ATTITUDE CONTROL

Careful examination of Eq. (28) shows that both a given set of values of q0, q1, q2, q3, as well as and their negatives give the same rotation matrix R. Every orientation corresponds to two different sets of Euler parameters. This slight ambiguity has no significant effect in applications, however. STABILITY OF THE TORQUE-FREE MOTION When no external forces or moments act on the body, it rotates freely about its center of mass. Its motion is called torque-free (2). If we perturb this torque-free motion slightly by exerting, say, a small impulse, the subsequent motion may or may not be similar to the motion before the impulse was applied. If the ensuing motion is similar or close to the motion before the impulse, we say that the motion of the body is stable. If, on the other hand, the motion after the impulse departs significantly from the original one, we say that the motion of the body is unstable. The stability of a torque-free rigid body can be analyzed by setting Mx ⫽ My ⫽ Mz ⫽ 0 in Eq. (13)

Jx ω˙ x = (Jy − Jz )ωy ωz Jy ω˙ y = (Jz − Jx )ωz ωx

(31)

Jz ω˙ z = (Jx − Jy )ωx ωy Assuming a nonsymmetric body (Jx ⬆ Jy ⬆ Jz), equilibrium (or steady-state) solutions correspond to permanent rotations with constant angular velocity about each of the three axes. For the sake of discussion, let us assume that Jx ⬍ Jy ⬍ Jz. Recall that in the absence of any external torques the angular momentum vector H remains constant in inertial space. Since the body rotates, H does not appear constant for an observer sitting in the body-fixed frame. Nevertheless, the magnitude of H is constant. This is evident from Eqs. (15) and (18). Thus, H2 = H

2

= Jx2 ωx2 + Jy2 ωy2 + Jz2 ωz2

of the semiaxes of these two ellipsoids differ, the ellipsoids will, in general, intersect. Their intersection defines a series of closed curves ( polhodes), which are the paths of the the tip of the angular velocity vector as seen from the body-fixed frame. Figure 3 shows a complete set of polhodes plotted on the angular momentum ellipsoid. Equilibrium solutions correspond to the intersections of the axes with this ellipsoid. The closed curves around the permanent rotations about the x and z axes indicate that the motion is periodic and these rotations are stable. The x-shape curve in the neighborhood of the permanent rotation about the y axis (the intermediate axis of inertia) indicates that this is an unstable rotation. In fact, any small perturbation will cause the body to depart from this equilibrium point. The previous geometric analysis shows that permanent rotations about the minor or the major principal axis of inertia are stable, whereas rotations about the axis of the intermediate moment of inertia are unstable. The reader can easily verify this behavior by throwing a book into the air about each of its principal axes and observe its subsequent motion. Thus far, it was assumed that the kinetic energy is conserved. Often the kinetic energy is reduced steadily due to internal or external dissipative forces, for example, elasticity or drag. In this case, it can be shown that rotations about the minor axis are also unstable. The body tends to a minimum energy state which is a rotation about the major axis of inertia. In particular, for axisymmetric bodies pure spin about the symmetry axis is stable only if the symmetry axis is the major inertia axis. This was vividly demonstrated during the launch of the first U.S. satellite, Explorer I. The satellite was a prolate (pencil-shaped) axisymmetric object, designed to spin about its symmetry axis. Flexing of the four whip communication antennas, however, caused energy dissipation and decrease of the kinetic energy. The satellite ended tumbling end-to-end after just one orbit. Despite the fact that oblate (disk-shaped) axisymmetric bodies exhibit more stable motion about their symmetry axis,

(32)

where H is a constant. In addition, conservation of energy implies that T = 12 (Jx ωx2 + Jy ωy2 + Jz ωz2 )

51

k

(33)

is also constant. We can use these two expressions to determine the behavior of the angular velocity vector ␻ in the body-fixed frame. By dividing Eqs. (32) and (33) by their left-hand sides, we obtain ωy2 ωx2 ωz2 + + =1 2 2 (H/Jx ) (H/Jy ) (H/Jz )2

(34)

i

ωx2 (2T/Jx )

+

ωy2 (2T/Jy )

+

ωz2 (2T/Jz )

=1

j

(35)

These equations describe two ellipsoids in the body frame. Equation (34) is called the angular momentum ellipsoid and Eq. (35) is called the kinetic energy ellipsoid. Since the lengths

Figure 3. The closed curves on the angular momentum ellipsoid denote the path of the tip of the angular velocity vector. Rotations about the x and z axis are stable, whereas rotations about the y axis are unstable. Here y is the intermediate moment-of-inertia axis.

52

ATTITUDE CONTROL

most often the satellites have a prolate shape. This is because the shape of the carrying rocket is prolate and elongated satellites make more efficient use of the available cargo space. Thus, stable rotations of prolate satellites about their symmetry axis require the use of some sort of stabilization.

Jx ω˙ x = Mx

ATTITUDE-CONTROL LAWS AND STABILIZATION Attitude-control laws for a spacecraft can be designed based on a straightforward application of Eq. (12) or Eq. (13). In the case of passive stabilization, the control torques are generated through the interaction of the spacecraft with its environment (gravity or magnetic field, solar or aerodynamic pressure, etc.) These environmental torques are much smaller than the torques generated using active stabilization schemes. We consider here only active control methods. The external torques M acting on the spacecraft are thus almost solely due to the attitude-control system (i.e., due to gas jets firings or momentum exchange wheels). The environmental torques in this case are treated as disturbances. For gas jets, Eq. (12) can be used directly. When momentum wheels are used as attitude control devices, the equations of motion have to be modified to take into consideration the dynamics of the wheels. For a spacecraft with three momentum wheels along the three principal axes, the equations of motion are given by ω = −ω ω × Jw (ω ω + ν ) − Jw (ω˙ + ν˙ ) Jω˙ + ω × Jω

(36)

where ␯ is the vector of the angular velocities of the three wheels relative to the spacecraft, and Jw is the diagonal matrix with the (polar) moments of inertia of the wheels. The control inputs are the wheel accelerations ␯˙ . Equation (36) describes the balance between the angular momentum of the spacecraft and the wheels. It essentially states that the total angular momentum (wheels and spacecraft) remains constant. The dynamics of the wheels are given by T Jw (ω˙ + ν˙ ) = −T

(37)

where T denotes the torques developed by the electric motors of the wheels. These are internal torques, which do not affect the total angular momentum. A preliminary feedback ω × Jwν − M Jwν˙ = −ω

(38)

can be used to put the system in the standard form of Eq. (12) ˆω + M ω × Jω Jˆω˙ = −ω

control system must keep the angular velocity vector with respect to the inertial frame at zero. For small angular deviations and small angular rates, we can use the Euler angles to describe the orientation of the body frame with respect to the inertial frame. Since the angles and their rates are small, we can linearize Eqs. (13) and (24) to obtain

(39)

where Jˆ ⫽ J ⫹ Jw is the total inertia matrix of the combined spacecraft/wheel system. Thus, regardless of whether we use gas jets or momentum wheels, Eq. (12) can be used to predict the effect of the control torques M on the body. Typical Attitude-Control Algorithms One typical control objective is to maintain the inertial orientation of the spacecraft fixed. This implies that the attitude

Jy ω˙ y = My

(40)

Jx ω˙ z = Mz φ˙ = ωx θ˙ = ωy ψ˙ = ωz

(41)

The attitude motions about the three body axes are decoupled. The control system can independently control the motion about each individual axis. A control law of the form ˙ Mx = −k1 φ − k2 φ,

k1 > 0, k2 > 0

(42)

can be used, for example, to keep ␾ ⫽ 0. This control law will require an attitude sensor to measure the roll angle ␾ and a rate gyro to measure ␾˙ . If no rate gyro is available a control law using lead compensation can be used (8) Mx = −k(φ − ξ ) ξ˙ = −bξ + (b − a)φ

(43)

where a and b positive numbers with b ⬎ a. The transfer function of this controller is s+a Mx (s) = −k φ(s) s+b

(44)

Similar control laws can be constructed for the y (pitch) and z (yaw) axes. The previous procedure based on linearization cannot be used when the expected deviations from the rest position are significant or when the cross-product terms in the angular velocity equation are not negligible. For large or fast angular maneuvers we need to work directly with the exact, nonlinear equations. In this case, the alternative formulation of the kinematic equations in terms of the Euler parameters in Eq. (29) becomes very useful, since we avoid the singularity of the Euler angle description. Most importantly, because of their simple structure, these equations are easier to work with than the highly nonlinear differential equations in terms of the Euler angles. Assuming that the Euler parameters describe the attitude error between the current and desired orientation, we can use the control law proposed by Mortensen (9) M = −k1ω − k2q v ,

k1 > 0, k2 > 0

(45)

to reorient the body to the desired attitude and keep it there. In Eq. (45), qv denotes the vector portion of the quaternion, qv ⫽ q1i ⫹ q2 j ⫹ q3k. Note that the control law in Eq. (45) is linear, although the original equations of motion are nonlinear.

ATTITUDE CONTROL

Often we wish to control only the angular velocity of the body. If the target angular velocity is zero (i.e., it is desired to bring the body to rest), the linear control law ω, M = −kω

k>0

53

the spacecraft with respect to this frame, the equations of motion can be written as (8,10)

Jx ω˙ x = −(Jy − Jz )ωz − 32 (Jy − Jz )φ + Mx Jy ω˙ y = 32 (Jx − Jz )θ + My

(46)

(51)

Jz ω˙ z = (Jy − Jx )ωx + Mz φ˙ = ψ + ωx θ˙ = + ωy ψ˙ = −φ + ωz

can be used. The difference between the control law in Eq. (45) and the control law in Eq. (46) is that in the latter the final orientation of the body is irrelevant. To achieve an arbitrary angular velocity ␻d, the following feedback control can be used ω − ω d ) + ω × Jω ω M = Jω˙ d − kJ(ω

(47)

If, for instance, ␻d ⫽ 웆dk the previous control law will generate a pure rotation of the body about its z axis with angular velocity 웆d. A special case of this situation occurs when the final spin axis of the spacecraft is also required to point along a specified direction in the inertial frame (i.e., for a spin-stabilized vehicle). The linear control given by Coppola and McClamroch (10)

Mx = −(Jy − Jz )ωz (θ˙ + ωd φ) − Jx ωd θ˙ − k1 φ˙ − k2 φ My = −(Jz − Jx )ωd (φ˙ − ωd θ ) + Jy ωd φ˙ − k3 θ˙ − k4 θ

(48)

Mz = −k5 (ωz − ωd ) for some positive scalars k i, will keep the body z axis aligned with the inertial Z axis (assuming that 웆x, 웆y, ␾, ␪ are small), whereas the control law

sin φ cos θ − k 2 ωx 1 + cos φ cos θ sin θ My = −k1 − k 3 ωy 1 + cos φ cos θ Mz = −k3 (ωz − ωd ) Mx = −k1

(49)

(52)

These equations reveal that the pitching motion is decoupled from roll/yaw motions. The control law

Mx = −42 (Jz − Jy )φ − k1 φ − k2 φ˙ − (Jx + Jz − Jy )ψ˙ My = −32 (Jx − Jz )θ − k3 θ − k4 θ˙

for some positive numbers ki, can be used to make the spacecraft rotate about its y axis such that its z axis points always toward the Earth. Optimal Reorientation Maneuvers. Because of limited onboard resources (e.g., power consumption or propellant), a spacecraft control system may be required to achieve the control objectives in the presence of certain constraints. For instance, it is clearly desirable to design control algorithms that minimize the fuel consumption during a particular maneuver (assuming gas jets are used as attitude actuators). Another example is the reorientation of an optical telescope or antenna in minimum time. For small-angle reorientation maneuvers about individual principal axes, the linear equations in Eqs. (40) and (41) can be used. Linear quadratic methods provide optimal controls for a quadratic penalty on the error and the control input. These methods have been discussed elsewhere (12,13). Referring to Eq. (13), Windeknecht (14) showed that the control law that minimizes the quantity

for some positive ki, can be used to bring the spin-axis (assumed to be the body z axis) along the inertial Z axis from almost every (not necessarily small) initial state (11). Spacecraft in Orbit. Another important special case of the previous control laws is the stabilization of a spacecraft in a circular orbit of radius Rc, such that its z axis points always towards the Earth. The orbital angular velocity is =

rg Rc

J = H (tf ) + λ

tf

M (t) 2 dt

(54)

0

is given by M∗ = −

H (tf − t) + λ

(55)

Kumar (15) showed that the optimal control minimizing the quantity

tf

J =

(50)

In this case it is convenient to choose an inertial frame that is parallel to a local-vertical, local-horizontal frame attached at the center of mass of the spacecraft. The X axis of this frame points along the direction of the orbital velocity (local horizontal), the Z axis points along the center of the Earth (local vertical), and Y points along the negative of the angular velocity vector ⍀ of the orbit. Describing the orientation of

(53)

Mz = (Jy − Jx )ψ − k5 ψ − k6 ψ˙ + (Jx + Jz − Jy )φ˙ 2

tf

H 2 dt + λ

0

M 2 dt

(56)

0

is given by H M ∗ = −γ (t)H where

t −t 1 γ (t) = √ tanh f√ λ λ

(57)

(58)

54

ATTITUDE CONTROL

For tf 씮 앝 the previous expression reduces to the linear control H M∗ = − √ λ

(59)

The previous control laws minimize the energy required to perform the maneuver. Often it is more relevant to minimize the fuel expenditure. Fuel consumption is proportional to the magnitude of the control torque 兩M兩. The minimum-fuel control law which takes the system to the rest position is thus derived by minimizing

tf

J =

M dt

(60)

0

where the final time tf is not prescribed. The optimal feedback control law is given by ¯ M ∗ = −M

H H

(61)

¯ is a constraint on the available control magnitude, where M ¯ (16). 兩M兩 ⱕ M The control law in Eq. (61) is also the minimum-time control law for the system in Eq. (13). This control law does not deal with the final orientation of the spacecraft, however. Minimum-time reorientation maneuvers where the final attitude is also of interest have been treated extensively in the literature (17,18). Analytic solutions are extremely difficult to find in this case. For large-angle (slew) maneuvers, in particular, one almost always needs to resort to numerical methods using Pontryagin’s Maximum Principle (17,19). Nevertheless, a minimum-time three-dimensional maneuver is not a minimum-time rotation about the corresponding eigenaxis (20). Explicit calculation of the minimum-time control law is possible if we assume that the angular displacements are small. In this case the linearized equations in Eqs. (40) and (41) can be used, and the optimal control is bang–bang control (i.e, one that switches between the maximum and minimum

value of the torque). For instance, assuming that the maxi¯ y, the control mum available torque about the pitch axis is M law that will bring the motion about the y body axis to rest ¯ y to ⫹M ¯ y (or vice versa) in minimum time switches from ⫺M according to whether the initial state (␪,␪˙ ) is above or below the switching curve in Fig. 4. The switching occurs when ␪ and ␪˙ satisfy the switching condition 2 θ˙ = ±θ

¯y 2M Jy

(62)

which is the equation that defines the switching curve. A summary of the minimum-time attitude maneuver literature can be found in the survey article by Scrivener and Thomson (21). AIRCRAFT ATTITUDE CONTROL Although similar, attitude-control problems for aircraft are much more challenging than attitude-control problems for spacecraft. The main difference between an aircraft and a spacecraft is the fact that the former flies in the atmosphere. The principal forces and moments acting on an aircraft are generated by the interaction of the airplane with the air flow. These forces are the same ones used for attitude control. Moreover, since the same forces also affect the aircraft’s center of mass, the translational and rotational equations are coupled. The aerodynamic forces acting on an aircraft in the atmosphere are proportional to the air density and the square of the airspeed (the relative velocity of the airplane to the air flow). The main aerodynamic forces acting on an aircraft are the drag, which is opposite to the direction of the airplane’s velocity vector, and the lift, which is perpendicular to the velocity vector. Lift opposes gravity and is the force that makes airplanes stay aloft. Drag opposes the motion of the airplane through the air and is responsible for most of the fuel consumption. Other significant forces acting on an airplane are the force of gravity and the thrust from the engines. Aircraft Dynamics

· θ 2My/Jy My = – My

Typical path

θ

My = + My

Typical path Switching curve Figure 4. Bang–bang minimum time control of a single-axis attitude maneuver. If the initial orientation and velocity of the body is below the switching curve, the control logic will switch from the maximum to the minimum possible torque. The opposite is true if the initial condition is above the switching curve.

As for spacecraft problems, the orientation of an airplane is determined by the relative angular displacements between a reference frame fixed in the airplane and an inertial frame. For most problems in airplane dynamics an axis system fixed to the Earth can be used as an inertial reference frame. There are several choices for the body reference frame. The body axes are aligned such that the x axis is along the longitudinal fuselage axis, the y axis is along the right wing, and the z axis is mutually perpendicular to the x and y axes. The wind axes are defined such that the x axis is along the direction of the relative wind. The angles 움 and 웁, defined by performing a rotation about the body y axis, followed by a rotation about the new z axis, until the body x axis is along the velocity vector, are called the angle of attack and sideslip angle, respectively. A positive angle of attack corresponds to a negative rotation about the y axis. The sideslip angle is positive if the rotation about the z axis is positive. The wind axis is the natural choice for analyzing the aerodynamic forces and moments. The stability axes are defined by the angle 움 between the body x axis and the stability x axis. Although all the previous sets of axes are referred to in the literature as body axes,

ATTITUDE CONTROL

only the first one is body-fixed. The orientation of the stability and wind axes may vary with flight conditions, but in most cases 움 and 웁 are small, so the stability and wind axes are close to the body-fixed axes. The transformation from body to stability axes is given by

   x cos α    y =  0 z S − sin α

  sin α x   0  y cos α z B

0 1 0

(as is often the case), the off-diagonal terms in the inertia matrix Jxy and Jyz are zero. Following the standard notation in aircraft literature (22,23), we denote the three components of the angular velocity vector in body axes by p, q, and r, respectively and the components of the applied torque by L, M, and N. The equations of motion in Eq. (12) are then written as

(63) L = Jx p˙ − Jxz r˙ + qr(Jz − Jy ) − Jxz pq M = Jy q˙ + rp(Jx − Jz ) − Jxz (p2 − r2 )

whereas the rotation from stability to wind axes is given by

   x cos β    y = − sin β z W 0

  0 x   0 y 1 z S

sin β cos β 0

(64)

cos α cos β  − cos α sin β − sin α

The moments L, M, and N represent the roll, pitching, and yawing moments, respectively. They are defined in terms of dimensionless aerodynamic coefficients Cl, Cm, and Cn as follows

L = qSbC ¯ l



sin β cos β 0

sin α cos β  − sin α sin β  cos α   cos β sin β 0 cos α   = − sin β cos β 0  0 − sin α 0 0 1

M = qScC ¯ m

w , u

sin β =

0 1 0

 sin α 0  cos α

where q¯ is the free-stream dynamic pressure, defined by

(65)

v VT

(66)

where u, v, and w are the components of the relative airspeed velocity of the airplane in the body axes. The magnitude VT ⫽ (u2 ⫹ v2 ⫹ w2)1/2 of the relative velocity is called the true airspeed. If the xy plane is a plane of symmetry of the airplane

q¯ =

Ailerons Elevator

R

el

ψ , r, N Body z axis

(69)

b b C p+ C r + Cl β + Cl δa + Cl δr β δa δr 2VT l p 2VT l r c c M = qSc ¯ Cm 0 + Cm α α + Cm q q + Cm α α˙ + Cm δ δe e 2VT 2VT b p N = qSb ¯ Cn p p + Cn r r + Cn β β + Cn δ δa + Cn δ δr a r 2VT 2VT (70)

Rudder

θ, q, M

1 ρV 2 2 T

S is the airplane wing reference area, b is the wing span, and c is the wing mean geometric chord. In Eq. (69), ␳ is the air density (1.225 kg/m3 at the sea level). The dimensionless coefficients Cl, Cm, and Cn measure the effectiveness of the airplane’s aerodynamic surfaces in producing moments and depend on several factors, such as the aerodynamic angles 움 and 웁, control surface deflections, engine power level, airplane geometry, and configuration. For small deviations of these parameters, the moments L, M, and N can be approximated by the expansions given in Pachter and Houpis (24)

L = qSb ¯

Body y axis

(68)

N = qSbC ¯ n

The body, wind, and stability axes for positive 움 and 웁 are shown in Fig. 5. From Fig. 5 we have immediately that the angle of attack and the sideslip angle satisfy the following expressions tan α =

(67)

N = −Jxz p˙ + Jz r˙ + pq(Jy − Jx ) + Jxz qr

Subsequently, the rotation matrix from body to wind axes is given by



55

at

iv

ϕ , p, L

α e

w

in

d

β x axis (wind)

x axis (body) x axis (stability)

Figure 5. Body reference frames on an airplane. The stability axes differ from the wind axes by the sideslip angle 웁, and the body-fixed axes differ from the stability axes by the angle of attack 움. The angles 움 and 웁 change as the relative velocity of the airplane to the wind changes.

where 웃e denotes the deflection angle for the elevator, 웃r the angle for the rudder, and 웃a for the ailerons. The coefficients Clp, Cn웁, Cm움˙ , . . ., are called the stability derivatives. The stability of an airplane about an equilibrium configuration depends upon these coefficients. As shown by McLean (25), for large aircraft, such as civilian passenger airplanes and transports, which cannot generate large angular velocities, Eq. (67) can be approximated by

L = Jx p˙ − Jxz (r˙ + pq) M = Jy q˙ + Jxz (p2 − r2 ) N = Jz r˙ − Jxz ( p˙ − qr)

(71)

56

ATTITUDE CONTROL

Equation (67) can be inverted to obtain 2 Jz Jxz Jxz J 2 − Jz Jy + Jxz (Jx − Jy + Jz )pq − z qr + L + N Jz − Jx Jxz 2 1 q˙ = pr − (p − r2 ) + M Jy Jy Jy

p˙ =

r˙ =

2 Jxz Jx Jx2 − Jy Jx + Jxz Jxz pq − (Jx − Jy + Jz )qr + L+ N (72)

2 . Once the moments L, M, and N are where ⌬ ⫽ JxJz ⫺ J xz known, the angular velocity can be computed by integrating Eq. (72).

Euler Angles The orientation of an airplane is given by the three Euler angles ␾, ␪, and ␺ from Eq. (22), also referred to as roll, pitch, and yaw, respectively. The kinematic equations of the airplane’s rotational motion are thus given by Eq. (24), repeated below for convenience

φ˙ = p + q sin φ tan θ + r cos φ tan θ θ˙ = q cos φ − r sin φ ψ˙ = (q sin φ + r cos φ) sec θ

(73)

Equations (72) and (73) can be integrated to completely describe the attitude evolution of the aircraft. It should be pointed out, however, that the aerodynamic forces and moments depend on the altitude and speed of the airplane. The rotational equations are thus coupled with the translational (flight path) equations of motion. A complete, six-degree-offreedom system that includes the translational equations is required to accurately describe the current position and velocity of the airplane. The complete nonlinear equations can be decomposed into the longitudinal equations, which describe the motion in the xz plane, and the lateral equations, which describe the motion outside the xz plane. The longitudinal part of the airplane’s motion includes, in addition to ␪ and q, the forward and vertical velocity of the center of mass. The lateral equations, in addition to ␾, ␺, p, and r will include the side velocity of the center of mass. A more complete discussion of the airplane’s complete set of equations of motion may be consulted (see, for example, Ref. 26). Aircraft Actuators Control of an airplane is achieved by providing an incremental lift force on one or more of the airplane’s surfaces. Because these control surfaces are located at a distance from the center of mass, the incremental lift force generates a moment about the airplane’s center of mass. The magnitude of the moment is proportional to the force and the distance of the control surface from the center of the mass. The main control actuators used for changing an airplane’s attitude motion are the elevators, the rudder, and the ailerons. Additional configurations may include canards (small surfaces located ahead of the main wing) or thrust vectoring devices (for military aircraft). Figure 5 shows the main control surfaces of an airplane. Elevators. Elevators are relatively small surfaces located close to the tail of the airplane. Deflecting the elevators produces moments about the pitch axis of the airplane. Elevators

are thus, primarily, pitch-control devices. The transfer function between the elevator deflection 웃e and the pitch angle ␪ is given by

Kθ (s2 + 2ζθ ωθ s + ωθ2 ) θ (s) = 2 2 )(s2 + 2ζ ω s + ω 2 ) δe (s) (s + 2ζphωph s + ωph sp sp sp

(74)

The ␨ph, 웆ph and ␨sp, 웆sp are the damping ratio and natural frequency of the phugoid and short-period modes, respectively. Rudders. The rudder is a hinged flap that is part of the vertical surface located at the tail of the airplane. It is primarily a yaw-control device and is the main directional control device of the airplane. In addition to directional control, the rudder is used to compensate for unwanted directional yaw deflections caused by the airelons when an airplane is banked to execute a turning maneuver. Airelons. Airelons differ from the previous two control devices, because they incorporate two lifting surfaces. Ailerons are located at the tips of the main wings of the airplane. Roll control is achieved by the differential deflection of the ailerons. They modify the lift distribution of the wings (increase it in one wing and decrease it in the other) so that a moment is created about the x axis. Spoilers. Roll moment is also produced by deflecting a wing spoiler. Wing spoilers are small surfaces located on the upper wing surface and cause flow separation when deflected. Flow separation in turn causes a reduction in lift. If only one spoiler is used at a time, the lift differential between the two wings will cause a rolling moment. In some aircraft roll control is also produced by tail surfaces moving differentially. Roll. The rolling (lateral) motion is not, in general, decoupled from the yawing (directional) motion. The transfer functions from 웃a and 웃r to ␾ and ␺ are coupled. The transfer function from aileron deflection to roll angle ␾ is given by

Kφ (s2 + 2ζφ ωφ s + ωφ2 ) φ(s) = 2) δa (s) (s + 1/Ts )(s + 1/Tr )(s2 + 2ζD ωD s + ωD

(75)

whereas the transfer function from rudder deflection to yaw angle ␺ is given by

Kψ (s2 + 2ζψ ωψ s + ωψ2 ) ψ (s) = 2) δr (s) (s + 1/Ts )(s + 1/Tr )(s2 + 2ζD ωD s + ωD

(76)

Similar expressions hold for the transfer functions ␾(s)/ 웃r(s) and ␺(s)/ 웃a(s). These equations should be used with caution since, as mentioned earlier, the lateral/directional motion is inherently a multi-input/multi-output system. The quadratic term in the denominator in Eqs. (75) and (76) corresponds to the dutch roll mode. The first term in the denominator corresponds to the spiral mode and the second term to the rolling subsidence mode. For most aircraft Ts is much larger than Tr and the quadratic terms in the numerator and denominator in Eq. (75) are quite close. Equation (75) can therefore be approximated by Kφ φ(s) = δa (s) s(s + 1/Tr )

(77)

ATTITUDE CONTROL

The transfer function from 웃r to ␺ is more difficult to approximate. Often, the dutch roll approximation found in McLean (25) Kψ ψ (s) = 2 2) δr (s) (s + 2ζD ωD s + ωD

(78)

is good enough. The short period, the roll, and the dutch-roll modes are the main principal modes associated with the rotational motion of the aircraft and are much faster than the phugoid and spiral modes, which are primarily associated with changes of the flight-path (translational motion). The slow phugoid and spiral modes can be controlled adequately by the pilot. Control systems are required, in general, for controlling or modifying the rotational modes. In addition, the maneuverability of the aircraft is primarily determined by the rotational modes. Stability Augmentation and Aircraft Attitude-Control Systems An automatic flight control system (AFCS) typically performs three main tasks: (1) modifies any unsatisfactory behavior of the aircraft’s natural flying characteristics, (2) provides relief from the pilot’s workload during normal cruising conditions or maneuvering, and (3) performs several specific functions, such as automatic landing. In addition, an AFCS may perform several secondary operations, such as engine and aircraft component monitoring, flight-path generation, terrainfollowing, collision avoidance. Here we briefly outline the fundamental operations of only the first two tasks. Control systems that are used to increase the damping or stiffness of the aircraft motion so as to provide artificial stability for an airplane with undesirable flying characteristics are called stability augmentation systems (SAS). Typical uses of SAS are in increasing the damping ratio of the short period motion in pitch (pitch rate SAS), providing damping in the roll subsidence mode (roll rate SAS), modifying the dutch roll mode (yaw rate SAS), and increasing the maneuverability of the aircraft by reducing static stability margins (relaxed static stability SAS). The SAS typically uses gyroscopes as sensors to measure the body-axes angular rates, processes them on-board using a flight-control computer, and generates the appropriate signals to the servomechanisms that drive the aerodynamic control surfaces. In addition to stability augmentation systems, which are used to modify the characteristics of the natural modes of the airplane, attitude-control systems (ACS) are used to perform more complex tasks. In contrast to the SAS, they use signals from many sensors and control several of the aircraft’s surfaces simultaneously. As a result, attitude control systems are multivariable control systems and therefore more complex in their operation than SAS. Common ACS for a typical aircraft are pitch ACS, roll angle ACS, coordinated-turn control systems, wing levellers, and sideslip suppression systems. A more in-depth discussion of ACS can be found in McLean (25) and Stevens and Lewis (23). The aircraft dynamics change considerably with the flight conditions, such as speed and altitude. The control design process involves linearization of the nonlinear equations of motion about steady state (trim) conditions. Steady-state aircraft flight is defined as a condition where all motion (state) variables are constant or zero. That is, linear and angular velocity

57

are constant (or zero) and all accelerations are zero. Examples of steady-state flight conditions involving the rotational degrees of freedom include: (1) steady turning flight (␺˙ ⫽ ␪˙ ⫽ 0), (2) steady pull-up (␾ ⫽ ␾˙ ⫽ ␺˙ ⫽ 0), and (3) steady roll (␪˙ ⫽ ␺˙ ⫽ 0). A control system designed for a certain steady-state condition may perform very poorly at another condition or even lead to instability. A control system must therefore be adapted during the flight to accommodate the wide variations in aircraft dynamics occurring over the flight envelope. Typically, several controllers are designed for different conditions and then gain-scheduled during the flight. Gain scheduling amounts to switching between the different controllers or adjusting their parameters (i.e., gains) as the airplane’s flight conditions change. Dynamic pressure is commonly used to schedule the controllers because it captures changes of both altitude and speed. Other parameters, such as angle of attack are used as well. Care must be taken when switching controllers during gain scheduling to avoid unacceptable transients. Extensive simulations are required to ensure that the gainscheduled control system performs satisfactorily. The U.S. government periodically releases a series of publications (e.g., 27), with guidelines and specifications for acceptable performance of flight-control systems.

ATTITUDE CONTROL IN ROBOTICS One of the main problems in robotics is the derivation of algorithms to actively control the position and orientation of the end-effector of a robotic manipulator, whether it is a video camera, a gripper, or a tool. The position and orientation of the end-effector is completely determined by the position and linear or angular displacements of the robot joints. For the sake of discussion, we henceforth consider a robot consisting of revolute joints only. The case with prismatic joints can be treated similarly. For a robot manipulator made of n links interconnected by revolute joints, the joint variables 웁1, . . ., 웁n are the relative angles between the links. The orientation and velocity of the end effector is then completely determined by the angles 웁i and their rates 웁˙ i (i ⫽ 1, . . ., n). To describe the orientation of the end effector with respect to the inertial space, we choose a reference frame fixed at the end effector. We call this frame the end-effector frame or the task frame (28). The inertial frame (also called the base or world frame) is usually established at the base of the robot. The end-effector orientation is then given by the rotation matrix R between these two reference frames. Relative rotation of the robot joints induces an angular velocity of the task frame with respect to the base frame. Three Euler angles ␾, ␪, and ␺ (roll, pitch, and yaw) can be used to parameterize the rotation matrix R between the two frames. These angles are the same as the ones in Figure 2. For a gripper, the roll angle ␾ describes relative rotation about an axis extending forward from the manipulator (the roll axis). The pitch angle ␪ describes relative rotation about the axis parallel to the axis connecting the gripper’s fingers (pitch axis) and perpendicular to the roll axis. Finally, the yaw angle ␺ describes a rotation about an axis perpendicular to both the roll and the pitch axis. As discussed elsewhere (28), the x, y, and z directions of the tool frame are labeled frequently as a, s, and n, respectively. The terminology arises from the fact that the direction a (or x) is the approach direc-

58

ATTITUDE CONTROL

tion (i.e., this is the direction that the gripper typically approaches an object). The s (or y) direction is the sliding direction (i.e., the direction along which the fingers of the gripper slide to close or open). The n (or z) direction is normal to the plane defined by the a and s directions. The (a, s, n) frame attached to a gripper is shown in Figure 6. The roll, pitch, and yaw angles completely describe the orientation of the end effector. They are given by

The torques generated at the joints will specify a commanded time history for 웁i(t) and 웁˙ i(t). Equations (79) and (81) can be used to find the corresponding angular position and velocity of the end effector. This is the so-called forward kinematics problem. As an example, consider the general equation of a robotic manipulator (29) M(β )β¨ + C(β, β˙ )β˙ + K(β ) = τ

φ = f 1 (β1 , . . ., βn ) θ = f 2 (β1 , . . ., βn )

(79)

ψ = f 3 (β1 , . . ., βn ) where the functions f 1, f 2, and f 3 are determined by the specific geometry of the manipulator. Differentiating the previous equation with respect to time, one obtains

∂f φ˙ = 1 β˙ 1 + · · · + ∂β1 ∂f θ˙ = 2 β˙ 1 + · · · + ∂β1 ∂f ψ˙ = 3 β˙ 1 + · · · + ∂β1

∂ f1 ˙ βn ∂βn ∂ f2 ˙ βn ∂βn ∂ f3 β˙ ∂βn

(81)

where J(웁) is a 3 ⫻ n matrix and 웁 ⫽ (웁1, . . ., 웁n). The matrix J(웁) is often called the Jacobian kinematics.

s

θ

Joint 2

Lin

Link 2

a

ψ

y2

ϕ

k3

x2

x1 Z

β1

β3

z2

Joint 1

o

v = β¨ d − 2λ(β˙ − β˙ d ) − λ2 (β − βd ),

(83)

Y

Base (link o) X Figure 6. Typical robotic manipulator consisting only of revolute joints. The attitude of the gripper is given by the orientation of the (a, s, n) body frame. The geometry of the manipulator determines the orientation of this frame with respect the the joint angles 웁1, 웁2, and 웁3.

λ>0

(84)

will force 웁(t) 씮 웁d(t) as t 씮 앝. Very often, the inverse problem is of interest. For example, the desired orientation and angular velocity of the end effector may be known or specified by the robot’s mission requirements. In those cases, it may be necessary to find the required velocities and positions at the joints, given the angular orientation and velocity of the end effector. The problem of finding the joint variables for a given position and orientation of the end effector is called the inverse kinematics problem, and it is much more difficult than the forward kinematics problem. The solution of the inverse problem is obtained by inverting Eqs. (79) and (81) for given 웆 and (␾, ␪, ␺). In general, because n ⱖ 3, this problem has more than one solution. The best solution (웁, 웁˙ ) depends on the specific application. The minimum-norm (least-squares) solution of Eq. (81) is given by β˙ = J † (β )η

Joint 4

Joint 3

z1 β 2

τ = M(β )v + C(β, β˙ )β˙ + K(β ) where

ω = J(β )β˙

y1

These equations are derived using the classical Lagrange equations (e.g., 6). The matrix M(웁) is the mass matrix, the term C(웁,웁˙ ) 웁˙ contains the Coriolis acceleration terms, and K(웁) contains all conservative forces (e.g., gravity). A control law for a robotic manipulator will generate the torques ␶ at the robot joints. Assuming an actuator (i.e., motor) at each joint, a control law can be devised to track some prespecified trajectory 웁d(t) in terms of the joint angles 웁i. For example, the control law

(80)

We can use Eqs. (24) and (80) to obtain a relation between the angular velocity vector ␻ expressed in the end-effector frame as a function of the rates of change of the joint angles 웁i as follows

n

(82)

(85)

where J† (웁) ⫽ J T(웁) [J(웁)J T(웁)]⫺1 denotes the Moore-Penrose pseudoinverse of the matrix J(웁) and where ␩ denotes the vector of the Euler angles and the angular velocity. Equation (85) provides the minimum joint velocity 웁˙ , which gives the desired end-effector velocity 웆. CURRENT TRENDS Several methodologies exist for stabilizing or controlling a rigid spacecraft when at least three independent control inputs are available. Some of these methodologies have been presented earlier. More challenging is the case when one or more actuators (either gas jets or momentum wheels) have failed. The theoretical investigation of this problem was initially addressed by Crouch (30). Several control laws were

ATTITUDE CONTROL

subsequently proposed, both for the angular-velocity equations (e.g., 31), and the complete velocity/orientation equations (e.g., 32,33). Controlling flexible spacecraft also presents great challenges. Control laws using on–off thrusters, for example, may excite the flexible modes of lightweight space structures, such as trusses or antennas. Modern control theory based on statespace models has been used to control these systems with great success. An in-depth discussion on the effect of flexibility on spacecraft reorientation maneuvers can be found in the literature (12,34,35). Research into fail-safe control systems for aircraft has also been an active area. The main emphasis has been placed on the design of reconfigurable flight-control systems and, more specifically, attitude-control systems. The idea is to construct intelligent control systems with high levels of autonomy that can reprogram themselves in case of an unexpected failure, so as to fly and land the airplane safely. The use of multivariable modern control theory (23) along with the use of redundant sensors and actuators and smart materials promise to change the current method of designing and implementing control systems for aircraft. Traditionally, the airplane control surfaces are connected directly to the cockpit through mechanical and hydraulic connections. A pilot command corresponds to a proportional surface deflection. In many recent military and civilian aircraft, the commands from the pilot are sent electronically to the control computer instead. The computer generates the appropriate control deflection signals based on its preprogrammed control law. This method is called fly-by-wire, since the pilot does not have direct command of the control surfaces. The on-board control computer is responsible for interpreting and executing the pilot commands. Redundant computers or backup mechanical connections are used to guard against possible computer failures. The term fly-by-light is also used when the pilot and control commands are sent using fiberoptic connections.

BIBLIOGRAPHY 1. J. R. Wertz, Spacecraft Attitude Determination and Control, Dordrecht: D. Reidel, 1980 2. W. Wiesel, Spaceflight Dynamics, New York: McGraw-Hill, 1989. 3. T. R. Kane, P. W. Likins, and P. A. Levinson, Spacecraft Dynamics, New York: McGraw-Hill, 1983. 4. M. D. Shuster, A survey of attitude representations, J. Astronaut. Sci. 41 (4): 439–517, 1993. 5. J. Stuelpnagel, On the parameterization of the three-dimensional rotation group, SIAM Rev. 6 (4): 422–430, 1964. 6. D. T. Greenwood, Principles of Dynamics, Englewood Cliffs, NJ: Prentice-Hall, 1988. 7. E. T. Whittaker, Analytical Dynamics of Particles and Rigid Bodies, New York: Dover, 1944. 8. A. E. Bryson, Control of Spacecraft and Aircraft, Princeton, NJ: Princeton University Press, 1994. 9. R. E. Mortensen, A globally stable linear attitude regulator, Int. J. Cont. 8 (3): 297–302, 1968. 10. V. Coppola and H. N. McClamroch, Spacecraft attitude control, in W. S. Levine (ed.), The Control Handbook, Boca Raton, FL: CRC Press, 1996.

59

11. P. Tsiotras and J. M. Longuski, Spin-axis stabilization of symmetric spacecraft with two control torques, Syst. Cont. Lett. 23 (6): 395–402, 1994. 12. J. L. Junkins and J. Turner, Optimal Spacecraft Rotational Maneuvers, New York: Elsevier, 1986. 13. A. E. Bryson and Y.-C. Ho, Applied Optimal Control: Optimization, Estimation, and Control, Washington, DC: Hemisphere, 1975. 14. T. G. Windeknecht, Optimal stabilization of rigid body attitude, J. Math. Anal. Appl. 6 (2): 325–335, 1963. 15. K. S. P. Kumar, On the optimum stabilization of a satellite, IEEE Trans. Aerospace Electron Syst., 1 (2): 82–83, 1965. 16. M. Athans, P. L. Falb, and R. T. Lacoss, Time-, fuel-, and energyoptimal control of nonlinear norm-invariant systems, IRE Trans. Automatic Contr., 8: 196–202, 1963. 17. J. L. Junkins, C. K. Carrington, and C. E. Williams, Time-optimal magnetic attitude maneuvers, J. Guid., Contr., Dynam., 4 (4): 363–368, 1981. 18. J. R. Etter, A solution of the time-optimal Euler rotation problem, in Proceedings of the AIAA Guidance, Navigation, and Control Conference, Vol. 2, Washington, DC: AIAA, 1989, pp. 1441–1449. 19. E. B. Lee and L. Markus, Foundations of Optimal Control Theory. Malabar, FL: Krieger, 1986. 20. K. D. Bilimoria and B. Wie, Time-optimal reorientation of a rigid axisymmetric spacecraft, in Proceedings of the AIAA Guidance, Navigation, and Control Conference, Washington, DC: AIAA, 1991, Paper 91-2644-CP. 21. S. L. Scrivener and R. C. Thomson, Survey of time-optimal attitude maneuvers, J. Guid., Contr., Dynam., 17 (2): 225–233, 1994. 22. C. R. Nelson, Flight Stability and Automatic Control, New York: McGraw-Hill, 1989. 23. B. L. Stevens and F. L. Lewis, Aircraft Control and Simulation, New York: Wiley, 1992. 24. M. Pachter and C. H. Houpis, Flight control of piloted aircraft, in W. S. Levine (ed), The Control Handbook, Boca Raton, FL: CRC Press, 1996. 25. D. McLean, Automatic Flight Control Systems, New York: Prentice Hall, 1990. 26. B. Etkin, Dynamics of Flight: Stability and Control, New York: Wiley, 1982. 27. U.S. Air Force, MIL-STD-1797A: Flying Qualities of Piloted Aircraft, Washington, DC: Government Printing Office 1991. 28. M. W. Spong and M. Vidyasagar, Robot Dynamics and Control, New York: Wiley, 1989. 29. J. J. E. Slotine and W. Li, Applied Nonlinear Control, Englewood Cliffs, NJ: Prentice Hall, 1991. 30. P. E. Crouch, Spacecraft attitude control and stabilization: applications of geometric control theory to rigid body models, IEEE Trans. Auto Contr, 29 (4): 321–331, 1984. 31. D. Aeyels, Stabilization by smooth feedback of the angular velocity of a rigid body, Syst. Contr. Lett., 6 (1): 59–63, 1985. 32. H. Krishnan, M. Reyhanoglu, and H. McClamroch, Attitude stabilization of a rigid spacecraft using two control torques: a nonlinear control approach based on the spacecraft attitude dynamics, Automatica, 30 (6): 1023–1027, 1994. 33. P. Tsiotras, M. Corless, and M. Longuski, A novel approach for the attitude control of an axisymmetric spacecraft subject to two control torques, Automatica, 31 (8): 1099–1112, 1995. 34. D. C. Hyland, J. L. Junkins, and R. W. Longman, Active control technology for large space structures, J. Guid., Contr., Dynam., 16 (5): 801–821, 1993. 35. S. A. Singh, Robust nonlinear attitude control of flexible spacecraft, IEEE Trans. Aerospace Electron. Syst., 23 (2): 380–387, 1987.

60

AUTHORING SYSTEMS

Reading List T. R. Kane and D. A. Levinson, Theory and Applications. New York: McGraw-Hill, 1985. The basic equations for rigid-body dynamics. Special issue on attitude representations, J. Astronaut. Sci., 41 (4): 1993. An exhaustive presentation of different attitude representations. M. L. Curtis, Matrix Groups. New York: Springer-Verlag, 1979. A mathematical treatment of attitude motion, along with connections with group theory and Lie algebraic concepts. F. P. J. Rimrott, Introductory Attitude Dynamics. New York: SpringerVerlag, 1989. Complete treatment of the dynamics of spacecraft with momentum wheels. P. C. Hughes, Spacecraft Attitude Dynamics. New York: Wiley, 1986. Classic reference. Complete analysis of stability problems for single and dual-spin spacecraft. D. L. Mingori, Effects of energy dissipation on the attitude stability of dual-spin satellites, AIAA J. 7: 20–27, 1969. More on the dynamics of dual spin. R. J. Kinsey, D. L. Mingori, and R. H. Rand, Nonlinear control of dual-spin spacecraft during despin through precession phase lock, J. Guid., Contr., Dynam., 19 (1): 60–67, 1996. J. T. Wen and K. Kreutz-Delgado, The attitude control problem, IEEE Trans. Auto. Contr. 36 (10): 1148–1162, 1991. Theoretical analysis of attitude control. D. McRuer, I. Ashkenas, and D. Graham, Aircraft Dynamics and Automatic Control. Princeton, NJ: Princeton University Press, 1973. J. Roskam, Flight Dynamics of Rigid and Elastic Airplanes. Kansas: University of Kansas Press, 1972. Special issue on aircraft flight control, Int. J. Contr., 59 (1): 1994. Recent advances in aircraft control. R. M. Murray, Z. Li, and S. S. Sastry, A Mathematical Introduction to Robotic Manipulation. Boca Raton, FL: CRC Press, 1994. Mathematical treatment of attitude dynamics, rotation matrices. T. I. Fossen, Guidance and Control of Ocean Vehicles. New York: Wiley, 1994. Attitude-control applications to marine vehicles.

PANAGIOTIS TSIOTRAS University of Virginia

AUDIO, MULTIMEDIA. See MULTIMEDIA AUDIO. AUDITING. See ACCOUNTING. AUTHENTICATION. See CRYPTOGRAPHY; DATA SECURITY. AUTHENTICATION SYSTEMS. See FINGERPRINT IDENTIFICATION.

626

ELECTRONIC WARFARE

ELECTRONIC WARFARE Electronic warfare (EW) is the systems discipline that exploits an adversary’s use of the electromagnetic spectrum to

overcome threats that use communications, navigation, and radar systems. It is an important tool in pursuing military objectives and advancing national policy and sovereignty. EW provides the means to counter, in all battle phases, hostile actions that use the electromagnetic spectrum—from the beginning, when enemy forces are mobilized for an attack, through to the final engagement. EW exploits the electromagnetic spectrum through electromagnetic sensing, analysis, and countermeasures to establish operational advantage in a hostile encounter. The use of electronic warfare accelerated rapidly during World War II, and it has been used in most military conflicts since. The aircraft used by Nazi Germany to bomb the fogshrouded British Isles were guided by radio beacons from the European mainland. By using false guidance signals, the British were able to redirect the German bombing attacks from densely populated urban areas to less populated rural areas. In this same conflict, US bombers used chaff (packets of tinfoil cut into thin strips) jettisoned from the attacking US aircraft to reflect antiaircraft radar signals, thereby reducing the effectiveness of the German antiaircraft batteries and bomber force attrition. In the Pacific theater of operations during World War II, US Navy submariners detected and determined the bearing and location of Japanese ship radio transmissions for weapons targeting. In the Korean conflict, detection and location of North Korean antiaircraft radar signals provided targeting data for subsequent air strikes. In Vietnam, the exploitation of antiaircraft and missile radars was refined with the use of US Air Force Wild Weasel weapons—suppression aircraft that used sensors to detect and locate the weapons-associated threat signals to provide targeting information for ordnance delivery. Electronic warfare applications are described extensively in military accounts of the past half century. Military operations use EW as one means to gather tactical intelligence from noncooperative forces and to counter their electromagnetic, radio-, and radar-controlled weapons. Land, sea, and air forces use the electromagnetic spectrum for command and control, weapons targeting, and weapons control. Figure 1 shows multiple land, sea, and air platforms in a typical tactical environment. Also indicated are links for sensing, communications, and navigation in support of the military mission. Electronic warfare provides use of the electromagnetic (EM) spectrum by the host force and denial or limitation of its use by an adversary. Realization of this goal occurs when host force systems use the EM spectrum while adversary systems are denied its use. Countermeasures (CM) to threat systems that use the EM spectrum can be selectively applied on a time- and/or frequency-multiplexed basis so that host force use of the EM spectrum is uninhibited. Electronic warfare includes the operational functions of electronic support (ES), electronic self protection (EP), and electronic attack (EA). ES provides surveillance and warning information for EW system use. CM to threat systems, including jamming, false target generation, and decoying, are performed for EP (protection of the host platform against an electronically controlled threat). EA performs these same CM functions to protect a battle force composed of several platforms or battle units. The ES, EA, and EP functions are interrelated because EA and EP can be queued using ES informa-

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

ELECTRONIC WARFARE

627

Radar

Weapons guidance

Electronic warfare Communications Sonar

Figure 1. Tactical operational concept indicating systems that use the EM spectrum.

tion, and EA and EP can use some of the same sensing and CM equipment for distinct operational objectives. This article includes a description of the EW time line and the various phases of conflict. Also provided is a summary description of the signal environment in which EW systems operate. Those interested in more detailed descriptions of the EM communications, radar, and navigation technology against whose signals EW systems operate are referred to the appropriate sections of this encyclopedia. A discussion of EW functional areas ES, EP, and EA provides a functional framework for supporting EW technologies.

ment. The EW time-line stage in a specific engagement depends on the deployment of forces and the perceived imminence of hostile engagement. Note that the technologies used in the various stages of the engagement are dynamic, and EW systems and weapon systems technologies evolve to overcome susceptibilities. The boundaries and definitions of EW timeline stages are redefined with each new advance in weapon and EW technology. Electronic Support Electronic support provides operational intelligence that is related to radiated signals in the battle group or theater envi-

ELECTRONIC WARFARE TIME LINE Electronic warfare is used in a layered operational interaction with electronically controlled threat systems. The electronic warfare system provides its own force with data for self-protection and threat weapons suppression. Figure 2 graphically illustrates the EW functional time line. Electronic support provides operational intelligence relating to electronically controlled threat systems and communications systems in the battle group or theater environment. Electronic threat-warning information derives from ES surveillance data, recognizing that hostile force deployments or weapons-related transmissions constitute a threat. Air defense combines electronic and radar surveillance with tactics and countermeasures to control the air battle. EA and active EP, using countertargeting (CTAR) jamming, false target generation, and/or decoying, attempt to deny target acquisition by adversary sensors. CTAR endeavors to deny weapon’s sensors use of the spectrum, and decoys dispersed into the environment provide preferred target signatures to the threat weapon’s sensor. The EW battle time line provides the general context in which the discipline of EW is used in the tactical environ-

Surveillance and warning Air defense

Countertargeting Hostile force

Missile defense

Own force

Figure 2. Electronic warfare battle situation showing various phases of the engagement time line.

628

ELECTRONIC WARFARE

ronment. Surveillance includes monitoring of both combatants and commercial transports. Control of contraband and critical materials is an EW surveillance mission that provides critical intelligence data to the area commander. Surveillance of noncooperative combatant forces provides deployment intelligence in the area of observation. Early threat-warning information extracted from surveillance data occurs by recognizing hostile force weapons-related transmissions. Within the lethal range of hostile force weapons, battle space surveillance updates are required rapidly. Deployment and operational modes of hostile forces are monitored closely to determine imminence of hostile activity. In some environments, potentially hostile forces remain within weapons’ lethal range and a high level of vigilance is necessary to maintain security.

ment to threat platform sensors to prevent own force target acquisition by the hostile force. The terminal phases of an air defense engagement are characterized by heightened activity. The combatants, both hostile and own force, are confined to a smaller portion of the battle space. Weapons and decoys in flight add to the physical and EM signal density. Electronically, both own force and hostile forces struggle to exploit the EM environment to achieve their respective operational objectives. Countermeasures jamming and spoofing are used with full appreciation that coordinated jamming produces degradation of hostile force sensors, but that weapons with home-on-jam (HOJ) capability can exploit this action to the destructive detriment of the radiating platform. Countertargeting

Air Defense Air defense is used to maintain control of the battle group airspace and defend against threat aircraft and missiles. Battle group surveillance, implemented by the combination of EW, infrared/electro-optic (IR/EO), and radar sensors, provides environmental data required for air defense. Electronic combat techniques and weapons are used to counter an airborne threat. Air defense is an extensive, complex, electronic combat interaction between hostile forces. EW assets are a key tool of the battle force commander and of the individual elements within the command. These assets provide information for developing tactical intelligence in all phases of the engagement. The outcome of the air battle is by no means established by the quantity of EW assets possessed by each of the opposing forces, but depends greatly on how the EW assets are used in conjunction with other sensor systems, weapons, and air defense tactics. Aircraft ships and/or battlefield installations participate in air defense. Own force aircraft operating at altitude can engage a threat force at long line-of-sight ranges. Aircraft, together with ship and battlefield installations, provide coordinated air defense as the hostile force approaches own force locations. The EW objective in the early air defense or outer air battle is to prevent threat force detection and location of own force. Electronic combat actions that prevent or delay own force detection provide a distinct advantage by allowing additional time to develop tactics to counter the threat force. In addition, the threat force battle time line and interplatform coordination are perturbed. Fragmentation or dissolution of the hostile force attack can occur if own force electronic combat is effective in the outer battle. As the hostile force overcomes the outer battle electronic attack and approaches the own force within weapons range, air defense assumes the role of denying targeting information to the hostile sensors. The EW objective at this stage of the engagement is to prevent hostile force weapons launch by denying targeting data to their sensors. Electronic combat surveillance, warning, and countermeasure assets are used for countertargeting. Surveillance sensors assess hostile force deployment and provide information about the adversarial tactics being used. Warning sensors indicate the status of threat sensors as they attempt to acquire targeting data for weapons systems handoff. Countermeasure assets, including jamming, spoofing, and decoying, continue to provide a virtual environ-

Countertargeting (CTAR) is a subset of radar electronic countermeasures (ECM) used in electronic attack. CTAR provides specially modulated radio-frequency (RF) signal transmissions to counter hostile force long-range surveillance or targeting radar. The transmission modulation can be amplitudemodulated (AM) or frequency-modulated (FM) noise, or combinations of these, and they can be pulsed or continuouswave. CTAR transmission is used both to disrupt and interfere with the threat radar operation, thereby preventing it from correctly locating and identifying own force target(s). Countertargeting success criteria includes mission completion prior to threat force interdiction or weapon launch. Realistically, the results of a CTAR electronic attack against a hostile force are probabilistic, in that some opposing forces at some time during the battle time line succeed in launching missiles. CTAR can delay and reduce the coordination of hostile missile firings and, consequently, reduce the number of missiles fired and the attrition of personnel, ships, and aircraft. Terminal Defense Terminal defense against electronically controlled missiles and guns is the final phase of the EW battle time line. Weapons are launched in the terminal phase of hostile force engagement, and EP and EA capability is brought to bear on the weapons and their electromagnetic (EM) guidance and control signals. Onboard jamming and false-target radiation that is effectively used for countertargeting is less effective for terminal defense. Jamming or false-target radiation makes the target platform vulnerable to missiles with home-on-jam capability. Home on jam is an electronic counter countermeasure that exploits the target countermeasure’s radiation to steer the missile to the target. Consequently, off board countermeasures, or decoys, are used to lure the missile away from the high-value target. THE ELECTRONIC WARFARE ENVIRONMENT Threat Systems Electronic warfare interacts with an adversary’s EM systems for signal exploitation and potentially for electronic attack. Threat systems of EW interest include radar, communications, and weapons control. Some of the threat systems exploited by EW are briefly described in the following.

ELECTRONIC WARFARE

Radar. Radar uses radio-frequency transmissions ranging from high frequency (HF) to millimeter waves (30 MHz to 40 GHz) in pulsed and continuous-wave (CW) modes to illuminate targets and collect reflected echoes. Radar-transmissionreflected echoes are used to measure target characteristics and determine target location. Military forces use radar for both offensive and defensive weapon systems. Radar functions include target detection and identification, target acquisition, target tracking, and navigation. Weapons systems using radar may be land-based, airborne, shipboard, or in space. A typical radar system contains a transmitter that produces a high-powered RF signal, tunable over a band of frequencies; an antenna system that radiates energy and collects reflected echoes; a receiver that detects signal return; and signal processing electronics that extract target measurements, such as range, bearing, and speed. Target location information is provided to a weapon system to control and direct the weapon onto the target. Land-based radars function as ground-controlled intercept (GCI) systems, surface-to-air missile (SAM), antiaircraft artillery (AA) batteries, and space tracking systems. GCI is used to direct interceptor aircraft against attacking aircraft and to coordinate the air battle. SAM sites use early warning/surveillance radar, target acquisition radar, target tracking (TT) radar and/or illuminators for missile-guidance, beam-riding systems. AA radars have operating frequencies and data rates similar to SAM tracking radars and usually receive targeting information from SAM surveillance and target acquisition (TA) facilities. Advanced SAM systems handle ballistic missile defense with higher data rates against high-speed targets. Airborne intercept and control (IC) radars provide early warning and information for the command and control of forces operating in the tactical environment. Space surveillance and tracking radars usually use large, fixed, phased arrays operating in the HF (3 MHz to 30 MHz) to 1 GHz frequency range. Table 1 gives parameters of typical radars categorized by radar function. The reader is referred to radar and electromagnetic wave propagation articles within this encyclopedia. Radar advancements can be expected in the areas of phased-array antennas, complex modulations on the radar pulse, improved signal processing to extract enhanced data from the radar return, and frequency diversity to cover the less used regions of the spectrum. Advanced designs from the US, European, and Russian inventories can be expected because of operational needs for enhanced sensor performance and the availability of affordable technologies to provide additional capability.

tween surveillance sites and between combat units. Communications networks range from basic field radio networks to long-distance, wide-area systems and point-to-point, highdata-rate installations. Communications systems cover the spectrum from very low frequency (5 Hz) to the frequencies of visible light, and they can be either free-space transmissions or confined to a transmission line. Free-space transmission links may be line of sight or cover longer distances by reflecting from the ionosphere, atmospheric layers, or troposcatter, or by relaying via satellite. Command and control communication links, using HF direct microwave and satellite relay, disseminate voice and digital data transmissions to land forces, air forces, and ships. Land combat units use ultrahigh frequency (UHF) (300 MHz to 3 GHz), very high frequency (VHF) (30 MHz to 300 MHz), land lines, and cellular phones over shorter distances mainly for voice transmissions. Surveillance activities and weapons sites may exchange data via voice or digital data link over a transmission path appropriate for the link span. Such links are used to transmit surveillance radar reports to an operations center or directly to a SAM battery. Communicationlink data rates depend on link bandwidth, modulation technique, and signal-to-noise ratio. Individual transmission-link throughput rates are in the range of hundreds of megabytes per second. Computer technology has enabled increased communication-link capacity for handling and processing data. The high data rates attainable permit transmission from airborne observers and between precision weapons and launch platforms. Communications in hostile environments are transmitted via protected cable between fixed sites, thus providing protection from physical damage, security from intercept, and immunity from jamming. Mobile communications require freespace transmissions that are susceptible to intercept and jamming. Communications counter-countermeasures, complex modulation, encryption, and spatial radiation constraints are used to mitigate the effects of EA. The use of modulation techniques increases privacy, reduces interference, improves reception, and reduces the probability of detection. Spread-spectrum communication systems that use four categories of signal modulation (direct sequence-modulated, frequencyhopping, intrapulse FM [chirp], and time-hopping) provide some level of signal protection from detection, demodulation, and interference. However, this is at the expense of increased bandwidth. Passive Weapons Sensors. Electro-optical and infrared (EO/ IR) systems sense spectral energy that is radiated by an object or reflected from an object from a source such as the sun, moon, or stars. The electro-optical spectral regions are categorized according to atmospheric propagative characteristics or

Communications. Communications systems provide information exchange for command and control to coordinate be-

Table 1. Parameter Ranges Associated with Radar Functions Radar Parameter

Frequency Range PRF Range

Radar Function GCI

IC

30 MHz to 3.0 GHz 100 pps to 500 pps

3.0 GHz to 10.0 GHz 1000 pps to 3000 pps

629

Surveillance 30 MHz to 3.0 GHz 100 pps to 500 pps

TA

TT, AA

3.0 GHz to 8.0 GHz 1000 pps to 2000 pps

6.0 GHz to 10.0 GHz 2000 pps to 4000 pps

Space Surveillance 30 MHz to 1.0 GHz —

630

ELECTRONIC WARFARE

Nonimaging reticles Quadrant-fixed reticle

Target Error

V Time

Image decenter

Image

Spin-scan reticle

V Time Conscan

V

V

Time Time

Detectors (4) V No error

V Time

V Error

Freq. Time

Nutation Pseudoimaging

Rosette

Transverse line scan IFOV

Search FOV

Petal Constant Target

Search 1 FOV 1 V 0

Video Time

;; Full imaging Linear detector array

Figure 3. Common IR/EO sensor types including nonimaging reticules, line scanning detectors, and area array imagers.

Scanned array

spectral transmittance. The EO/IR spectrum used for passive weapons sensors spans the 0.2 애m to 15 애m wavelength range. Electro-optical/infrared guidance provides angle target tracking information only. EO/IR weapons system guidance sensors fall into three classes: nonimaging, pseudoimaging, and imaging. Generally, countermeasure techniques exhibit preferential effectiveness against guidance approach. Some countermeasures techniques may be effective against pseudoimaging sensors and less effective against nonimaging and imaging sensors. Other countermeasures techniques may be preferentially effective against nonimaging and imaging sensors. Figure 3 illustrates the most common seeker-design approaches. These approaches are quadrant, spin scan, conical scan (conscan), transverse-line scan, and rosette scan. In the quadrant approach, an intentionally defocused spot images on

Target

Target image 2 D array

a four-element square array. Tracking is achieved by balancing the signal on all four detectors. In spin scan, a spinning reticle provides phase and amplitude information with respect to a fixed reference. With conscan, the target image is nutated by using a scanning mirror or optical wedge imaged onto a fixed reticule or pattern of detectors. The nutated target image generates a modulated frequency proportional to the angular and radial offset from the center. In the transverse-line scan approach, a rotating or reciprocating mirror at a depressed elevation angle generates a scan line transverse to the missile axis, and the forward motion of the missile creates the orthogonal axis of the search pattern. With the rosette scan, a petal pattern is scanned over a small instantaneous field of view (IFOV) by two counterrotating optical elements.

ELECTRONIC WARFARE

Rosette-scan tracking is accomplished by balancing the signal output from all petals with the target present in the central apex of the rosette. The small IFOV of the transverseline scan and rosette scan provide high spatial resolution and the ability to resolve multiple sources within the scanned field of view. Focal-plane arrays, scanning-linear arrays, or twodimensional arrays of detectors in the image plane provide high-resolution ‘‘pictures’’ of the target space. Many imageprocessing algorithms are available to classify targets and establish track points. Figure 3 illustrates the basic features of common seekers. Passive electro-optic sensors are desirable targeting and weapons guidance systems because they radiate no energy to warn the target of an impending attack. These sensor systems are vulnerable to decoys, with thermal signatures similar to true targets and to high-intensity sources that can saturate the electro-optic sensor detector or cause physical damage. ELECTRONIC WARFARE FUNCTIONAL AREAS Threat systems use the EM spectrum extensively. This section discusses functional aspects of EW. The relationships that govern their systems’ application are described in the following section. These functional areas are electronic support (ES), electronic protection (EP), and electronic attack (EA). Electronic attack uses countertargeting (CTAR), jamming, false-target generation, and decoys to defeat the threat sensors. Electronic protection uses electronic support and electronic attack for own-platform self-protection. Electronic Support Electronic support provides surveillance and warning information to the EW system. ES is a passive, nonradiating, EW system function that provides a fast accurate assessment of the EM radiating environment. ES is the aspect of EW that involves techniques to search for, intercept, locate, record, and analyze radiated energy for exploitation in support of military operations. Electronic support provides EW information for use in EA and EP and in tactical planning. ES directly provides threat identification/detection and early warning. It also provides data for electronic countermeasures (ECM), electronic counter-countermeasures (ECCM), threat avoidance, target acquisition, and homing. Electronic support provides timely EM environment information for the EW system. The spatial and spectral environment over which ES operates may span a hemispherical spatial segment and a spectrum of tens of gigahertz. In tactical EW systems, signals in the environment are analyzed and reports of environment activity are provided on the order of a second after threat signal reception. Electronic Attack As an EW function, EA provides an overt active response capability against enemy combat systems with the intent of degrading, deceiving, neutralizing, or otherwise rendering them ineffective or inoperative. EA responds to threat systems to protect multiple platform or battle group units. EA includes measures and countermeasures directed against electronic and electro-optical systems by using the electromagnetic spectrum (radio, microwave, infrared, visual, and ultraviolet fre-

631

quencies). EA technical functions include radio and radar signal jamming, false target generation, and the use of decoys for threat system confusion and distraction. Electronic attack is reactive to environment threats. To function effectively, therefore, the EA system requires threat information from the environment, including threat classification, bearing and, if possible, range. These functions are performed by the ES system or by other surveillance systems such as radar or infrared search and track (IRST). Effective EA response selection requires knowledge of the threat class and operating mode. Threat signal data are derived from measuring signal parameters (frequency, scan type, scan rates, pulse-repetition frequency, or continuous-wave radiation characteristics). Absence of radiation may indicate that the threat uses a passive RF or an electro-optical sensor. The detected threat electronic parameters are compared to an extensive emitter database. The EW database, derived from intelligence sources, is used to identify the threat and correlate the threat and operating mode with effective EA techniques. Operational threat exploitation is often impeded by intelligence gaps and/or threat use of parameters reserved for wartime. Nondestructive Electronic Attack. Nondestructive EA produces electromagnetic signals at a predetermined radio, infrared, visual, or ultraviolet frequency with characteristics that temporarily interfere with the threat’s receiving system, that is, power level, frequency, and polarization. EA degrades or overcomes threat system operation by overpowering the target signal at the threat sensor. ‘‘Dazzling’’ is laser or highpower lamp EO/IR jamming. Dazzling saturates the detectors or focal-plane arrays of electro-optical (infrared, visual, ultraviolet) guided missiles and target-tracking systems. Deceptive EA presents a confusing signal to the threat sensor that degrades its performance to the point where it is no longer effective. Power levels used for deception are less than those required for jamming because deception does not require threat sensor saturation. Destructive Electronic Attack. Destructive EA physically damages or destroys the threat electronic system. Specially designed missiles such as the HARM missile, shown being released from an A-6 aircraft in Fig. 4, are equipped with radar-homing seekers that attack the threat radar antenna and nearby electronic equipment within the blast radius of the missile warhead. More recently, similar seekers have been fitted to loitering remotely piloted vehicles for a similar purpose. Advances in high-power microwave and laser technology have made directed energy more practical. At very high power levels, microwave energy destroys the components in a missile seeker or threat radar, rendering them inoperative. Highpower lasers also physically damage both RF and electro-optical threat systems. Electronic Protection Electronic protection provides EW protection for the host platform. Key environment surveillance and threat-warning information is provided by the ES system function (as it is for EA). EP responds to threats in the environment with information for evasive action and with the countermeasure responses described previously. EP is primarily directed against

632

ELECTRONIC WARFARE

Figure 4. HARM missile (shown after separation from an EA-6B aircraft) is an EW weapon for physically destroying the source of hostile radiation.

the terminal threat targeted on the host platform, and preferred EP techniques use decoys that are less susceptible to the home-on-jam weapon mode. ELECTRONIC WARFARE TECHNICAL AREAS Technical areas that support the ES, EA, and EP functional EW systems areas are discussed in this section. All aspects of EW are addressed by modeling and simulation because this is the most practical means for functional evaluation. System architectural analyses address the formulation of efficient EW system configurations to provide the operational functions required within the constraints of available equipment, techniques, and technology. Technical areas that address ES primarily are signal detection, measurement, and processing issues that deal with environment surveillance and warning. Technical areas associated with EA and EP include CTAR jamming and false-target generation, EO/IR CM, and decoys. Also included in these technical area discussions are technology challenges to EW technologies for future capability. Modeling and Simulation for Electronic Warfare Electronic warfare uses modeling and simulation extensively in three areas of investigation: research into new hardware; threat domination/exploitation; and tactics development. The effectiveness of an EW architecture or equipment suite is assessed by using a computer model and parametric studies run against the model. Estimates of a threat system’s capabilities are incorporated into the model as environment sources because acquiring foreign hardware and measuring its performance is difficult. Environment signal models stimulate the EW system model. The EA effectiveness modeled against the threat is measured, and tactics are developed to further reduce threat system efficiency. Modeling and simulation (M&S) combine detailed antiship missile models with ship models, antiair missile models with aircraft models, electromagnetic propagation models, and chaff RF decoy models. (Chaff RF decoys are described later).

Chaff effectiveness evaluation considers the spatial relationship between the missile seeker and the ship while accounting for radar clutter and multipath returns. Signals at the missile are processed through the seeker receiver and missile guidance and tracking logic. A chaff cloud(s) injected into the simulation provides a false radar target signal at the missile seeker. By varying the amount of chaff and/or the chaff round spatial relationship with respect to both the defended ship and the threat missile, chaff effectiveness and tactics can be evaluated. However, the accuracy of the M&S results depends on the accuracy of the models used. An accurate missile sensor and control model is necessary to determine the effects of the complex signal returns from the target ship and the chaff on the missile controls and resultant flight path. In a simulated engagement, detailed missile functions are required to provide an accurate assessment of chaff effectiveness. These functions include monopulse antenna processing, range and angle tracking, missile guidance, and aerodynamics. Multiple threat seeker modes, such as acquisition, reacquisition, track, home-on-jam (HOJ), and simulated coherent combinations of signal segments are also required in the model. Target ship, aircraft, and chaff radar cross section (RCS) must be accurately modeled. Typically, a multireflector target simulation is used to represent the RCS signature. Ideally, a model of thousands of scatterers would provide greater accuracy. However, careful selection of several hundred scatterers is adequate. The accuracy of the missile and target interaction depends on the propagative environment model including multipath. Typically, a ray-tracing algorithm models the propagation of RF energy. Useful models rely on a stochastic representation of clutter as a function of wind speed, grazing angle, frequency, polarization, and ducting. Modeling of an ocean environment can be extended to include reflection from wave segments. Models are verified by using measured field test data. Electronic Warfare System Architectures. The EW system architecture ties system functional elements into an efficient configuration optimized to the operational mission. Figure 5 shows a typical EW system architecture. The system performs signal acquisition and parameter measurement, direction finding, countermeasure generation, and decoy deployment. The system central processing unit (CPU) provides sensor and countermeasure coordination and EW system interface with other onboard systems. Fusing the measurements of EW sensors and processors is a complex technological challenge. This information includes radar, communications, EO/IR, direction finding, and signal analysis. Data fusion within the EW system requires algorithmic development and significant enhancement in computational throughput. The EW system includes antenna(s), receiver(s), and processor(s) elements that provide data on signals in the environment. System sensors detect and measure threat signal characteristics. Multiple sensor subsystems measure the characteristics of the signal. For example, a signal acquisition detects the presence of a signal and measures the envelope characteristics (frequency, time of arrival, and signal duration). Another sensor that may include multiple antennas and receivers provides signal bearing-angle data. Separate subsystem sensors measure intrapulse signal modulation and/or received polarization.

ELECTRONIC WARFARE

WB converter

Channelizer

Synthesizer

DF antenna

Tuner

CM RCVR ANT

Comm Nav Display

CPU

Encoder

Tuners

Decoys

Phase quantizer CM XMIT ANT

Receiver

633

Techniques generator

Transmitter

A countermeasures receiver may use an independent electromagnetic environment interface. The countermeasures receiver accepts signals from the environment and provides them to the techniques generator. Target signals designated by CPU algorithms are selected for countermeasure generation as are the countermeasure modulation techniques to be applied. The resulting jamming signals are amplified to the desired power levels and radiated into the environment. Decoys are part of the EW system architecture. This subsystem is controlled by the CPU based on sensor inputs. Decoys provide the important function of separating the countermeasure signal source from the host platform. In this operational mode, decoys provide alternative highly visible targets to divert a weapon from its intended target. Also required are the means, such as the coordination of jamming with the use of decoys, to neutralize the HOJ weapons threat. Surveillance and Warning Electronic support surveillance and warning perform the functions of noncooperative intercept and exploitation of radiated energy in the EM environment. Surveillance and warning detection relationships are those associated with communications systems. Additional signal detection constraints result because the signal’s spatial location and its characteristics may not be known. Signal unknowns require tradeoffs of detection sensitivity and environment search. Once detected and measured, environment signals require sophisticated signal processing for signal sorting, formation, and characterization before they can be correlated with signal intelligence libraries for classification. Some fundamental tradeoff relationships for detection and warning are discussed below. Threat Signal Detection. Threat signal detection occurs as the electronic support system is illuminated above the system sensitivity level with signals that satisfy the single-pulse detection criteria. Detection is performed as the ES system scans the environment. Detection metrics include incident radiation sensitivity, detection probability, false detection probability, corruption probability, simultaneous detection, and throughput rate. Aircraft are often used to carry electronic warfare battlefield surveillance equipment. The operating altitude of sur-

Figure 5. Electronic warfare system architecture indicating system functional elements required to provide ES, EA, and EP functions to the host platform and operational battle group.

veillance aircraft provides a long line-of-sight range to the horizon. The range to the electromagnetic horizon accounting for nominal atmospheric refractions is given by R=

3 h 2

1/2 (1)

where h is the aircraft sensor altitude in feet and R is the observer-to-horizon range in statute miles. The time required to survey the environment depends on the surveillance alert status, system sensitivity, instantaneous observation segment, and rate of environment search. The number of instantaneous environment segments in frequency and bearing establish the number of environment dwell periods required for an environment scan. The larger the environment segments, the more rapidly the system performs the scan. The dwell at a given environment segment is scheduled to span the signal event period. Time to intercept is modeled by TI =

(TD NM) PT

(2)

where TI is the time required to survey the environment, TD is the EW support system dwell period, N is the number of frequency segments in the environment, M is the number of spatial segments in the environment, and PT is the probability that the signal occurs above the sensitivity level. In Eq. (2), spatial environment segmentation, spectral environment segmentation, and detection probability combine multiplicatively to define the time required to survey the environment. Wide instantaneous bandwidths and a large instantaneous field of view reduce environment survey time unless equipment choices reduce system sensitivity and the corresponding probability of signal detection. Equations (3) and (4) describe receiver sensitivity and aperture gain functional relationships: S = (NF)(SNR)(kTB)

(3)

where S is receiver sensitivity, NF is receiver noise factor, SNR is the required sensitivity for detection and false alarm

634

ELECTRONIC WARFARE

criteria, k is Boltzmann’s constant, T is temperature in degrees kelvin, and B is bandwidth in hertz. (4)

where G is antenna gain, K is antenna efficiency (less than unity), and ␪ is antenna beamwidth in steradians. A tradeoff between sensitivity and time to intercept is implied in Eqs. (3) and (4). By using multichannel processing, the tradeoff can be resolved in either domain. A wideband channelizer provides instantaneous spectral coverage equal to the span of the channelizer frequency coverage, and receiver sensitivity is established by the bandwidth of an individual channel. Multichannel spatial processing provides the instantaneous spatial coverage of the sum of the channels being processed. System antenna gain is based on channel beamwidth. Detection sensitivity requires consideration of the desired detection range. Equation (5) defines the electronic support detection range in terms of the threat signal parameters and the electronic support antenna, receiver, and processor system parameters:

  RMAX =  

1/2 P G Gr λ

(4π )3

tS t N

2

kTBn L

  

(5)

MIN

where RMAX is the maximum detection range, Pt is the threat signal transmit power, Gt is the threat signal antenna gain, Gr is the antenna gain of the electronic support subsystem, ␭ is the wavelength of the threat signal transmission, (S/N)MIN is the minimum signal-to-noise ratio required by the electronic support subsystem for detection, k is Boltzmann’s constant, T is absolute temperature, Bn is the effective noise bandwidth of the electronic support receiver, and L represents the combined feed losses of the threat transmitter and the electronic support receiver. The probabilistic characteristic of signal detection is illustrated by considering the intercept of a threat at a signal level considerably above the receiver threshold level. Detection probability arises primarily from the independent probabilities that the ES system observes the environment in the spatial and spectral location of the threat emitter and that the threat emitter illuminates the receiver with the required power for detection. Also of importance is the probability of signal detection once the ES system is steered to the signal spatial and spectral location. Then detection probability PD is based on the signal characteristics, that is, the probability that the threat signal illuminates the EW system during the observation period. The time required to perform a detection TI is derived from the scan interval TS and is given by TI ⫽ TS /PD. False reports from the electronic support receiver are highly undesirable. Limited computational resources are needed to process each pulse received in an attempt to form an association with other pulse reports. The rate of false reports is established by the proximity of the detector threshold level to the noise level. Figure 6 shows the relationship between the single-event probability of detection, the probability of false signal report generation, and the signal-to-noise ratio. This figure shows that both the probability of detection

Visibility factor (signal-to-noise ratio) (dB)

G = 2Kπ/θ

20

10–14 10–12 10–10 10–8 10–6 10–5 10–4 10–3

Pf0 = 10–16 15

10 5

0

–5

10–2

10–1

–10 –15 0.001 0.01

0.1 0.5 0.9 Probability of detection

0.99 0.999

Figure 6. Detection probability and false detection probability for various signal-to-noise ratio conditions.

and the probability of false generation are strong functions of the signal-to-noise ratio. The probability of pulse interference POL depends on the duration TD of the signal and the rate R at which signals are expected. A reduction in POL results from adding parallel measurement channels. The functional relationship approximating POL is

POL

(TD R)N N! = N

TD R 1+ N N=1

(6)

where TD is the event duration, R is the event repetition rate, N is the number of parallel measurement functions provided, and POL is less than 0.9. Electronic Support Signal Processing. The ES signal processor derives signal information from the multitude of environment event measurements. Signal processing is the focal point of the ES subsystem where operationally relevant sense is made of large data inputs. ES processing includes sorting event data and correlating sorted event data with emitter libraries to establish the class or family of signals to which the emitter belongs. Beyond sorting, intensive processing is applied to identify intercepted emitters specifically and to locate them precisely within the battle space. Sorting, a key electronic support signal processing function, correlates event descriptors from the same emitter. Correlation is performed on the basis of both instantaneous and temporal signal parameters. Instantaneous parameter sorts are less computationally demanding than temporal deinterleaving. The initial signal sorting is histogramming based on instantaneous signal parameters. The signal parameters used for histogram-based sorting are those available from a single event or pulse measurement. They include external signal pa-

ELECTRONIC WARFARE

rameters, such as signal frequency, start time, duration, power level, and angle of arrival. Other instantaneous parameters used are measurements of signal modulation. Signals measurements with like parameters are binned together, and it is postulated that each bin contains event descriptor data from the same emitter. After sorting, event descriptors are placed in individual emitter-associated groups. The monopulse and interpulse characteristics of the event group measurements are quantified into a signal descriptor. The signal descriptors are classified into an emitter class by correlation with a library database. In some instances, high-resolution signal measurements identify specific emitters. As might be expected, identification parameter sets and the processing required to establish them are significantly in excess of that required for classification. Here, as in the case of classification, detailed signal descriptors are correlated with a library to define a specific emitter. The spatial distribution of threat signals in the environment is operationally important. Determining the threat signal bearing angle with respect to own platform is a key step toward establishing threat signal position information. Conventional techniques used for direction-finding measurements include the use of differential amplitude processing of squinted antennas (antennas aimed in different directions), differential phase measurements from a phased-array antenna, and differential time of arrival measurements from spatially separated receivers. Both hostile and benign operational scenarios require information about the location of both noncooperative fixed and mobile emitter installations. Electronic warfare target location exploits direction-finding data and navigational data to provide a signal location solution. Single or multiple platforms are used to generate location data. The accuracy of target location depends on the precision of the direction-finding data and the navigation measurement and on the length of the baseline between measurements and the range to the target. Figure 7 shows target location geometry. The major error location axis A is modeled by A = Rϕcsc

ψ

(7)

2

where R is the range from observer to the target emitter, ␸ is the direction-finding measurement error, and ␺ is the angle

Observer Rϕ /2

Target

ϕ /2

R Range

Baseline L

Figure 7. Emitter location geometry supporting Eq. (7), with observer track and signal measurement angles indicated.

635

subtended by the maximum difference in observation bearings with respect to the target, which provides location measurement error for the condition ␺ ⬍ 앟/2. The range R from the observer to the target is given by R=L

sin(π − θ − γ ) sin θ

(8)

where L is the separation between observations, ␪ is the angle between the baseline and the opposite bearing angle, and 웂 is the angle between the baseline and the adjacent bearing angle. Electronic Support Digital Signal Processing Technology. Electronic warfare system processing, both dedicated and programmable, assimilates environment data from the receivers and wideband processors. It uses these data to sort, classify, and identify the sources of emissions to represent the environment relevantly. The digital signal processor provides the means for applying an array of algorithms to both predetection and detection signal data to extract threat information for EW system use. Digital signal processing metrics include high-rate signal throughput processing in a compact module. Digital signal processing is the heart of the ES function. It provides the flexibility of applying an extensive array of algorithms to system data. Critical digital signal processing technology challenges include processing throughput and developing efficient processing algorithms. Although signal data can be refined by applying sequential algorithms, the ES response is time critical; it must provide the most accurate assessment of available data within the required response time. Great potential exists for advancing digital signal processing technology, but optimum ES performance can be expected from judicious allocation of processing tasks between wideband processors and the digital signal processor. An example of digital signal processing technology is LMISPE (little monopulse information signal processing element), a special-purpose signal processor designed to operate with high-quality superheterodyne RF receiver systems. LMISPE provides extremely accurate pulse analysis and parameter extraction for signal classification and specific emitter identification (SEI). It is contained in a single rackmounted enclosure. Surveillance and Warning Technology. Surveillance and warning are the sensor and environment processing functions for the EW system. Speed and accuracy of measurements and processing functions are the primary metrics for ES. Accurate throughput is important in providing sufficient time for effective threat response to the EA or platform commander. In addition, precision threat assessment provided to the EA subsystem facilitates optimum technique selection and conservation of EA power resource for engaging multiple threats. The ES performance challenge is further constrained by space limitations aboard platforms, particularly aircraft. Receiver technology performs environment sensing for the EW application. Receiver Technology. Electronic support throughput and physical displacement metrics are addressed in developing wideband, small-size monolithic microwave integrated circuit (MMIC) technology. MMIC monolithic integrated analog processing at multigigahertz operating frequencies provides a ca-

636

ELECTRONIC WARFARE

Figure 8. The MMIC receiver, a combination of monolithic microwave, analog, and digital circuits, performs signal selection and conversion to a convenient intermediate frequency.

pability suited to ES receiver applications. Advantages sought in the exploitation of this technology base include economies of size, weight, power, and cost. Increased receiver dynamic range for continuous environment intercept during active countermeasures transmission remains a receiver technology challenge. The MMIC receiver shown in Fig. 8 is an example of this technology. Wideband Processing. Wideband receivers provide high probability of signal intercept. Wide spectral segment processing is necessary to increase signal detection sensitivity and to provide copulse reception of multiple simultaneous signals and rejection of interference signals. Requirements for wide instantaneous bandwidth, rapid throughput, and small modules are wideband processing metrics. Acousto-optic channelization technology is being developed for wideband processing as a compact, economical means for performing high-resolution environment segmentation. Wideband-signal frequency demultiplexing is performed using Bragg regime acousto-optic diffraction and electronic signal detection and encoding. Functions performed by these acousto-optic processors include channelized correlation, convolution, and spectral processing. Acousto-optic channelizers are based on Bragg diffraction of light (Fig. 9). The Bragg cell serves as the optical deflection or optical modulator element within the processor. The Bragg cell is an optically transparent medium, such as a crystal, that is driven at the applied RF frequency by using a piezoelectric RF-to-acoustic transducer. The Bragg cell transduces the RF signal into acoustic waves that are collimated into the Bragg cell crystal. The propagating acoustic wave creates sequential regions of crystal compression and extension that correspond to the period of the acoustic wave. The acoustically induced diffraction grating in the Bragg cell interacts with a coherent optical source to perform RF input frequency demultiplexing. The deflected light beams output from the Bragg cell are focused onto a detector array where light is detected to indicate energy in segments of the applied RF spectrum.

Wideband Interconnections. Electronic warfare sensors require broad access to the electromagnetic environment to provide quick response to hostile electromagnetic activity. For convenience and efficiency, central stowage of signal processing functional elements is important. To assure signal visibility, environment apertures, antennas, and EO/IR sensors must occupy locations on the periphery of the aircraft, ship, or land vehicle. Wideband interconnects transmit electromagnetic environment data from the EW system apertures to processing subsystems. With the current RF bandwidth of the electronic warfare environment expanding through tens of gigahertz, just finding a medium that supports that level of frequency coverage is a challenge. At light frequencies, however, a 100 GHz spectrum spans less than a third of 1% of light frequency. In addition, low-loss-transmission optical fibers provide a nearly lossless means to transfer wide spectra across a platform. Indeed, wideband interconnect technology is developing the use of fiber optics. Usable optical fiber bandwidth is limited by dispersion. Conventional fiber exhibits dispersion of 20 ps/km/nm of bandwidth. A typical signal operating within a 10 MHz bandwidth would exhibit dispersion of less than 0.1⬚. Clearly, bandwidth limitations are elsewhere in the link. Detectors have also been developed to provide bandwidths on the order of tens of gigahertz. High RF operating frequency detection is performed by using small-geometry detectors that exhibit maximum power limitations. Limitation in maximum power levels applied to the detector restricts the output signal intensity range. Recent developments in distributed detector elements are extending detector power-handling capabilities. Dynamic range is a significant fiber-optic link metric because the EW sensor system must process low-power signals on the horizon in an environment with high-power local transmissions. Modulator and detector attenuation reductions are technological issues being addressed to enhance the dynamic range performance of fiber-optic links. Countertargeting Countertargeting (CTAR) is the technical area that provides the means for protecting the host platform or force from Bragg cell principle Laser Bragg cell

cte Defle

Acoustically induced diffraction grating

Transducer f1 Signal input f2

Defle

t d ligh

ction

angle

f1

f2

O p t i c a l s e n s o r

Undeflected light

Figure 9. The acousto-optic Bragg regime signal transform processing principle used for signal-frequency analysis, sensitivity enhancement, and direction-finding functions.

ELECTRONIC WARFARE

aperture, inverse synthetic aperture, high range-resolution, and moving target indication processing can accurately determine target location and identify the type of target.

Program start Initialized serial port and generators

Set generator 1 at center frequency and sens pwr +4dBm Set generator 2 at first sensitivity frequency/power reading

Set PRI/pulse width (100/0.1 50/0.5 10/1.0 20/10.0)

Send dwell check for data

Yes

Record data in file

Data available? No

No

Five repetitions completed?

Yes

Are all PRI/pulse widths done?

No

Yes Set generator 2 for next frequency/sens power.

Freq = 3.5GHz?

637

Yes

No Program finish Figure 10. CTAR functional diagram showing sequence used in engaging a surveillance or targeting radar signal.

weapons targeting by a hostile force. CTAR functions include obscuration, false-target generation, and confusion. Associated techniques include jamming and onboard and offboard false-target generation. Countertargeting operates against radars that feature a target-locating or surveillance mode, as shown in the functional sequence of Fig. 10. Airborne surveillance radar is generally used against ship and ground forces because the aircraft altitude provides extended surface target detection range. Conversely, when defending against aircraft with CTAR, the radar could be ground-based. Some radars are designed with the sole purpose of surveillance, whereas others are multimode and can track targets. By using imaging processing, modern surveillance radars that include synthetic

Countertargeting Techniques. Figure 10 shows the CTAR functional sequence. CTAR EA techniques are categorized as environment obscuration and jamming and false-target signal generation. CTAR provides either confusing or ambiguous data to adversary surveillance and targeting radar displays to confuse the human operators who interpret these presentations. Radar displays include plan position indicators (PPIs), A- or B-scopes, or combinations of these. Obscuration screens targets over selected portions of the display with a jamming signal power above that of the target signal in environment segments spanning both range and azimuth (see radar articles for descriptions of radar displays). The amplitude of the obscuration CTAR signal exceeds that of any target-reflected signal in the screened sector. Experienced operators can recognize obscuration and radar jamming and initiate procedures to mitigate its effects. The false-target CTAR technique, however, is a more subtle form of EA that is less apparent to the operator. Here, the CTAR signal creates false indications on the radar display that appear as real targets to the operator. When the display is cluttered with false targets, radar operator time is consumed sorting through them. Selecting a false target for missile engagement dissipates an expensive weapon. CTAR EA systems can be used to protect an entire military force. CTAR force protection systems are generally large and use human operators for system control. An example is the AN/ALQ-99 system installed on the EA-6B (Fig. 11), and EF111 EW aircraft. Some EA systems, such as the AN/SLQ-32 installed on surface ships (Fig. 12), are for self-protection and support EA functions. The EA system selects a specific technique from a large EA technique library. Selection is based on knowledge of the threat location, class, electronic parameters, and operating mode. The EA system, using an embedded receiver subsystem, rapidly adapts to threat signal operating mode changes. The threat changes operating mode as either a counter-countermeasures technique to circumvent EA or as part of the hostile targeting and homing sequence. Adaptive EA provides rapid changes in techniques as the threat sequences through operating modes.

Figure 11. EA-6B aircraft equipped with the AN/ALQ-99 EA system for airborne CTAR.

638

ELECTRONIC WARFARE

grammed pattern. False-target deception techniques are generated to emulate true target returns. The threat-radar operator, in response to deception, may conclude that all detected targets are genuine and simply select false targets for weapons engagement, or, if deception is suspected, time and computational resources must be used to identify the true target prior to engagement. In automated weapons systems, the EA subsystem may create so many false targets that the radar computer becomes overloaded. Because Doppler radar and missile seekers process large numbers of ambiguous radar returns to fix the true target, they are particularly vulnerable to coherent false-target techniques. An effective CTAR approach combines jamming and deception. Jamming creates a radial strobe that obscures the true target, whereas the deceptive CTAR provides false targets that project through the jamming strobe.

Figure 12. Shipboard installation of the AN/SLQ-32 EW equipment used for CTAR.

Both jamming and deception CTAR techniques may be used. RF jamming techniques are either ‘‘barrage’’ or ‘‘spot.’’ Barrage jamming covers a wider frequency band at lower power density levels and is used to jam either several radars at once or spread-spectrum systems where the precise frequency of the threat is uncertain. Spot jamming concentrates the entire jamming power within the bandwidth of a single threat radar receiver with correspondingly better results. In both cases, a radial jamming strobe will appear on the threat radar PPI scope, as shown in Fig. 13. If the ratio of jamming signal power to the reflected radar signal power (J/S) is insufficient, the real target will ‘‘burn through’’ the jamming signal and become visible within the jamming strobe. For greater jamming effectiveness, it is desirable to have sufficiently large J/S to prevent burn through in the main beam and the principal sidelobes (see jam-to-signal calculations later). Deception techniques are more varied and are generally threat-specific. Many deception techniques are directed against threat-tracking radars or missile-seeker radars. These techniques attack the threat radar target-tracking loops in range, angle, or Doppler. Deception techniques are often used in combinations and can be sequenced as the threat modes vary, or they can sequence according to a pro-

Figure 13. PPI radar scope without and with jamming, showing the effects of CTAR jamming on the threat radar display.

Countertargeting Effectiveness. Countertargeting effectiveness is assessed by comparing threat system performance in benign and CM environments. The ability of the threat system to detect, acquire, and target true targets, including parameters, such as target acquisition time and weapon release range, is assessed by evaluating threat performance against live targets on test ranges. Evaluating missile-seeker countermeasure effectiveness presents a more difficult problem. Computer simulations model the missile fly-out from an actual or surrogate threat system against a live target. A measure of CTAR effectiveness (MOE) is the ratio of the number of missiles that approach their target outside of the missile lethal range to those missiles that approach the target within lethal range. Software simulates multiunit engagements. US Navy ship EA is evaluated by flying test aircraft carrying captive instrumented seekers against the ships and recording the threat system performance. A statistical technique to assess CTAR effectiveness compares the number of missiles required to defeat an EAequipped aircraft versus the number required to defeat a nonEA-equipped aircraft. Similar statistics assess the number of antiradiation missiles fired versus the number of radar systems defeated. Additional effectiveness information can also be gleaned from intelligence sources. Obscuration Burn Through. A measure of CTAR obscuration effectiveness is the range at which the radar displays the target in the presence of jamming. This is called the burn through range. At this range, the radar is sufficiently close to the target that the processed target-reflected radar power exceeds the jamming signal display masking. The real target becomes visible superimposed on the jamming signal. Burn through is modeled in Eq. (9) by using the radar range equation and free-space propagation. The radar range equation provides the signal power S that is received at the radar after being transmitted to and reflected from the target. The freespace signal propagation equation models the jammer power J that is received at the radar from the jammer. The quotient of jammer to signal power constitutes a figure of merit known as the jam-to-signal (J/S) ratio. This ratio is unique for each radar and depends on radar processing gain and on the display format and screen phosphor. Operator proficiency also plays a significant role. Rearranging the terms of this equation to solve for range yields the burn through equation: Rb =

J P σB R

S

J

PJ 4πBR

(9)

ELECTRONIC WARFARE

where Rb is the burn through range, J/S is the ratio of jammer-to-signal power required to jam the victim radar, PR is the effective radiated power of the radar, PJ is the effective radiated power of the jammer, ␴ is the radar cross section of the target, BJ is the jamming signal bandwidth, and BR is the processing bandwidth of the radar receiver. This equation models the case with the jammer located on the radar target platform. Jammer-to-Signal-Power Relationships. The J/S power ratio at the threat radar is a concept central to predicting EA effectiveness. To degrade the threat radar, an interfering jammer power J of sufficient strength is required to overcome the target-reflected signal at the radar S. For effective EM noise jamming, the J/S required is 0 dB to 6 dB minimum, depending on the noise modulations used and the detailed characteristics of the threat. The minimum J/S ratio required for effective CTAR deception techniques varies from 0 dB for false targets, to 0 dB to 6 dB for range deception, to 10 dB to 25 dB for angle-tracking deception, and to 20 dB to 40 dB for monopulse deception. Equations (10)–(12) are based on two typical EA tactical situations. Self-protection CTAR [Eq. (10)] addresses the case with the target in the threat radar main beam. Support CTAR [Eq. (11)] addresses the case of the target in the threat main radar beam but with the EA jamming emanating from a separate platform and radiating into an arbitrary bearing of the threat radar antenna pattern. In both cases, the radar is assumed monostatic (i.e., the radar receiver and transmitter are collocated). J/S for self-protection EP CTAR:

J/S =

4πPj Gj Br R2 Pr Gr σ g2 Bj

(10)

where Pj is jammer power output; Gj is gain of jammer antenna in direction of radar; Br is radar receiver noise bandwidth; R is radar-to-jammer range; Pr is radar power output; Gr is gain of radar antenna in target direction; ␴ is target radar cross section; g2 is propagation one-way power gain (square of the ratio of field strength to free-space field strength due to direct and reflected ray combination), 0 ⬍ g2 ⬍ 4 (interferometer lobing); and Bj is the jammer noise bandwidth. J/S for support EA:

J/S =

4πPJ Gjr Grj Br R4t g2j Pr G2r σ Bj R2j g4t

(11)

where Gjr is the gain of the jammer antenna in the direction of the radar, Grj is the gain of the radar antenna in the direction of the jammer, Rt is the radar-to-target range, gj is the jammer-to-radar propagation factor, Rj is the radar-to-jammer range, and gt is the radar-to-target propagation factor. The remaining terms are as defined previously. Effect of target radar cross-sectional reduction: S=

Pr Gr σ λ2 g4 (4π )3 R4

(12)

where ␭ is the wavelength of the radar operating frequency. All of the remaining terms are as defined previously.

639

Equation (12) defines the signal at the receiver of a monostatic radar. Note that the power received at the radar is directly proportional to the target radar cross section ␴ and inversely proportional to the fourth power of the range R (R is the separation between the target and radar). Therefore, as the radar cross section is reduced, the signal at the radar is correspondingly reduced. If the cross section is sufficiently reduced, the target becomes indistinguishable from the radar noise and background clutter. Low observable platforms, such as the B-2 and F-117 aircraft, provide sufficiently low radar cross section to make radar detection difficult. The implication of radar cross-sectional reduction technology to CTAR is twofold: first, with sufficiently low radar cross section, EP may not be necessary, and secondly, if the cross section merely lowers the signal power at the radar, then a lower power, low-cost CTAR transmitter becomes sufficient to provide the J/S necessary to achieve the desired level of survivability. Countermeasure Technology. Countermeasure technology addresses the evolving threat in addition to the need for economic force protection. Significant advances in radar, communications, EO/IR weapons’ sensors, and weapons control present heightened challenges to maintaining effective EA capability. Radar Countermeasures Technology. Countertargeting equipment for use against advanced synthetic aperture radar (SAR) or inverse synthetic aperture (ISAR) surveillance and targeting radar requires wide instantaneous bandwidths and high processing speeds. Furthermore, because these radars use coherent processing, CTAR effectiveness consequently requires coherent radar signal storage and reproduction to enhance effectiveness. Digital RF memory (DRFM) technology is being developed to convert the analog radar RF signals into a digital format for convenient storage. As required, the radar signal is retrieved from storage and converted to RF for use in countermeasure waveform generation. Technology limitations and costs constrain currently available DRFM designs, each optimized for a specific application. Radio-frequency-tapped delay lines provide precise timing between portions of the CTAR waveform. Analog RF-tapped delay lines use surface acoustic wave (SAW) and acoustic charge-transport technology. Research is underway to create digital tapped-delay lines. Noise modulation is commonly applied to CTAR signals, and high-quality tunable noise sources are required. The output EA stage is the transmitter/antenna combination that generates and radiates the CTAR signal. Antennas for EA applications, once considered a dedicated asset, are currently envisioned as multifunction phased-array antennas with elements fed by solid-state amplifiers. Radio-frequency isolation between the countermeasures transmitter and the receiver is a common problem of countermeasures-equipped platforms. The countermeasure signal appears at the receiver antenna. When the transmitter and receiver are insufficiently isolated, the countermeasure signal interferes with lower level threat signal reception from the environment. Interference demands careful attention to antenna design, isolation, and platform siting. Radar Countermeasure Signal Source Technology. Electronic attack transmitters require signal sources that can be rapidly switched in azimuth, elevation, frequency, and polarization to generate multiple high-power beams with low sidelobes over large multioctave bandwidths. CTAR requirements for eco-

640

ELECTRONIC WARFARE

nomical compact transmitters are challenged by the lack of appropriate low-cost EM power sources. Furthermore, few commercial applications exist for wideband EM power-source technology. Research and development in this area is limited primarily to EA applications. Original EW power sources, tunable magnetrons, and cross-field amplifiers provided only narrow operating bandwidths. Traveling wave tubes (TWTs) evolved to fill the need for wide, instantaneous bandwidth. Over time, TWT bandwidths grew from a single-octave 2 GHz to 4 GHz band to multiple octaves at frequencies beyond 40 GHz. However, TWTs are expensive and unreliable. Although new mini-TWTs and microwave power modules have become available, their basic design remains vacuum-envelope-based. MMIC technology is steadily advancing, and it now provides solid-state chips with multioctave signal-generation capability, wide instantaneous bandwidth, and signal power levels approaching 5 W. With MMIC technology, solid-state active aperture arrays become achievable, and such arrays for EA applications are now being developed. Although MMIC active aperture array signal source promises good performance and reliability, the system remains expensive. Passive Electro-Optic/Infrared Electronic Warfare Electronic warfare in a passive EO/IR target acquisition and weapons sensors environment applies to a growing threat capability. The open-ocean blue-water scenario requires EO/IR EA and EP ship protection, typically 200 nautical miles or more from shore, against massive and coordinated attack. EO/IR EA applications have recently focused on littoral scenarios involving amphibious operations in support of peacekeeping operations for regional conflicts; providing humanitarian assistance in politically and militarily unstable regions; evacuating civilians from regions of conflict; and ensuring safe passage of commerce through disputed littoral waters and choke points. The traditional EO/IR threat, the long-range antiship missile, has been intensified in the littoral areas by a large vari-

ety of air-to-surface, air-to-air, and surface-to-air EO/IR missile weapons. These missiles can inflict severe damage to the smaller craft used for littoral warfare. Electro-optic system target detection range depends on detector sensitivity and resolution. A target image is defined by contrast with the background. Sensitivity determines whether the contrast is discernible. Resolution depends on the spatial environment angle illuminating the detector, which is a function of detector surface area and focusing optics. The distance at which target features are resolvable determines the maximum operating range of the system. The target signature detectability is not determined by the absolute temperature of the object but rather by the contrast between the target and background within a given spectral band. Environment backgrounds range from the cold, uniform background of space to thermally cluttered land areas. Solar interaction with the target and background reflection and heating further degrade the background contrast with the target. Typical target contrasts range from about 1 kW/sr (kilowatt per steradian) in the 2 애m to 3 애m atmospheric window for an aircraft engine to tens of kilowatts per steradian for ships in the 8 애m to 12 애m window. Target aspect, especially the location of hot spots, greatly influences the signature. Electro-Optic/Infrared Countermeasures. Electro-optic/infrared countermeasures are constrained by specular atmospheric propagative characteristics, as is the threat (Fig. 14). The contrast of the target to the background within the weapon sensor’s specular passband, the type of seeker spatial localization processing, and available practical radiation sources are also prime considerations. The missile fly-out and CM sequence of events occurs in several seconds. As part of an integrated electronic warfare suite, the EO/IR EA system is designed to engage a large number of missiles launched in a coordinated attack. Figure

Atmospheric transmission 0.2

0.5

1

0.2

Nd 0.5

10

Er

Gas 1

Er

Ho, Tm 2

20

CO2 CO2

Hf 5

10

20

BBO OPO

BBO OPO+SHG

Lasers plus frequency conversion

5

Diodes

Excimer Lasers

2

PPLN OPO

FHG 0.2

FHG

ZnGeP2 OPO

Nd + SHG 0.5

1

2

5

10

Wavelength (µ m) Figure 14. EO/IR atmospheric transmission spectral segments and laser and laser harmonics countermeasures source spectral regions.

20

ELECTRONIC WARFARE

641

60 Minimum range for intercept Decoy Time from ASM impact (s)

50 CIWS

Antimissile missiles

40 Range by which EO/IR EA Inform must be passed

30

Handoff from EO/IR EA

Handoff to EO/IR EA

20 RF decoy Antimissile missile launch IR decoy CIWS

10

0 0

5

10 ASM Range from the ship (km)

15 shows a typical time line of the CM response to an attack by a subsonic antiship missile. The time line indicates the interaction of EO/IR EA with other ship defense systems. To preclude detection by a threat EO/IR sensor, target signature can be reduced through a combination of convective, conductive, and radiative mechanisms. Exterior surfaces of ship stacks are cooled by convective air flow between the engine exhaust ports and the outer stacks. Engine plume and exhaust gases from all types of engines can be cooled by dilution with air. Radiation from hot spots can be reduced by spectral emissivity modifications or by obscuring the hot areas from view. On new platforms, low-observability design criteria have led to low-signature aircraft and ships. Onboard aircraft CM sources initially generated false target location and/or guidance degradation through weapon automatic gain control (AGC) manipulation. This technique remains highly effective against many threats. The onboard jammer sources can be chemically fueled IR sources or electrically powered incandescent and metal vapor lamps. As the wavelength passbands of antiair and antiship seekers gradually migrate to longer wavelengths, out to the 8 애m to 14 애m window, noncoherent sources will no longer be practical. Basic spin scan and conical scan (conscan) ‘‘hot spot’’ seekers are vulnerable to flare decoys. Almost universally, these flares are composed of magnesium and polytetrafluoroethylene and are designed with a radiant intensity several times that of the target. In the distraction mode, the decoy is an excellent target; in the seduction mode, the weapon’s seeker control signal is biased by the decoy or transferred to it. Because pseudoimaging seekers exhibit spatial and temporal processing capabilities, simple flares are relatively ineffective, and simple flares perform even more poorly against imaging sensors. Newer decoys overcome advanced seeker-discriminating processing with improved spectral characteristics that more closely match the target platform spectral emissions. Improved decoy spatial distribution in the form of clouds and multiple hot spots, temporal rise times, and persistence match target-signature increase rates and lifetimes, thus preventing time-history discrimination. Kinematics model realistic target movement.

15

20

Figure 15. Missile attack time line showing launch, acquisition, and homing phases of the missile as well as the CM attack on missile sensors and control circuits.

The small beam divergence of lasers can result in highradiance, low-power sources that provide the J/S power ratios needed for effective EA. Two laser sources, primary lasers and nonlinearly shifted lasers, are available for CM applications. Lasers shifted by nonlinear conversion include harmonic generation and tunable optical parametric oscillators (OPOs). Primary lasers do not produce spectral lines in all of the potential threat passbands of interest and are susceptible to notch-filter counter-countermeasure techniques. Although harmonic generating EA techniques provide additional wavelengths, they are also subject to counter CM. Promising sources for IR/EO CM are tunable OPOs pumped by diodepumped, solid-state lasers. Two nonlinear materials currently demonstrating the highest potential are periodically poled lithium niobate (PPLN) and zinc germanium phosphide (ZnGeP2). Figure 14 shows the primary lasers of interest and the wavelength coverage possible with PPLN and ZnGeP2 OPOs. Although noncoherent sources provide wide angular protection, high-resolution detection is necessary to point and track the threat system and effectively use laser power. Timely threat detection and warning ES is essential to the success of all nonpreemptive EA. Electro-Optic/Infrared Countermeasure Technology. Key EO/IR EA technologies required to counter threat performance improvements include higher throughput data processing using more capable algorithms, laser beam steering, and decoy launcher design. Needed processing improvements include faster signal processing, more efficient image processing, and false alarm reduction. High-performance, highspeed beam steering, preferably nonmechanical, is required to reduce response time in multiple threat environments. Improved decoy launchers to position decoys quickly and accurately within the scenario are also needed. Low observability technologies are being developed to decrease or mask the IR/EO signatures of targets. Target signature reduction increases the effectiveness of conventional countermeasure responses by reducing the jamming power required to counter the missile system effectively. Low observ-

642

ELECTRONIC WARFARE

ability enables applying new technologies to IR/EO countermeasures by reducing the size, weight, and power requirements of decoy and laser CM sources. For example, diode laser and diode-pumped nonlinear optical sources can be integrated with unmanned aerial vehicles to produce new classes of CM devices and tactics. Large-area spectrally selective sources and obscurants provide advanced capability against spatially and spectrally discriminating threats. Primary laser and laser-pumped nonlinear sources are important evolving technologies. Launchers and vehicles that provide rapid and precise CM placement with realistic kinematic performance are areas of increasing importance. Decoy Countermeasures Decoys are EW devices, usually expendable, deployed from the platforms to be protected. Decoys generate a jamming response to the threat or false targets. In either case, the decoy lures the threat away from the intended target toward the decoy. A jamming decoy generates a cover signal that masks the target signal. Thereby the threat sensor signal fidelity is degraded, making detection and tracking of the intended target more difficult. A jamming signal may also activate the antijam home-on-jam mode of the weapon system. As false targets, the decoys generate credible target signatures to provide weapon system seduction or distraction. Decoys create confusion that causes weapons to attack false targets. Decoys may be either passive or active. A passive decoy generates a countermeasure response without the direct, active amplification of the threat signal. Principal examples of passive decoys are chaff and corner reflectors in the RF spectrum and flares in the EO/IR spectrum. Decoy Operational Employment. Decoys provide EA capability across the entire EW battle time line. Decoys are used primarily for EP missile defense and self-protection missile defense but also for countersurveillance and countertargeting applications. Jamming is used in conjunction with decoys to obscure the target signal at the threat radar during decoy deployment. As decoys are deployed, jamming ceases and the threat radar acquires the decoy as a target or transfers radar tracking from the target to the decoy. Threat radar acquisition of the decoy as a target is probable because decoys present prominent signatures. Decoys used for missile defense perform either seduction, distraction, or preferential acquisition functions. A single decoy type may perform multiple functions, depending on deployment geometry with respect to the launch aircraft or ship and the stage of electronic combat. Decoys are used in a seduction role as a terminal defense countermeasure against missile weapons systems. A seduction decoy transfers the lock of the missile guidance radar or EO/IR sensor from the defending platform onto itself. The decoy that generates a false-target signature is initially placed in the same threat tracking gate, missile sensor range, and/ or angle segment as the defending target and is subsequently separated from the launching platform. The decoy signature captures the missile guidance sensor, and the target lock is transferred from the ship or aircraft to the decoy. Typically, the decoy is separated in both range and angle from the defending target to assure target-to-missile physical separation

Figure 16. ALE-129 RF chaff round with the bundle of reflector elements partially deployed from the canister.

greater than the missile warhead’s blast range. The seduction decoy missile interaction is typically initiated within 10 s of deployment. Distraction decoys are deployed prior to missileseeker acquisition and provide multiple false targets from which the seeker may select. Deployed distraction decoys provide a confusing environment to the missile seeker, causing it to attack a decoy rather than the intended target. The ALE-129 chaff decoy (Fig. 16) is representative of RF seduction decoys for aircraft defense. The NATO Sea Gnat MK-214 cartridge shown fired from a shipboard launcher in Fig. 17 provides surface defense against radar-guided weapons. Figure 18 shows a TORCH decoy deployed at sea for IR defense. Distraction decoys are observed for extended periods in the engagement scenario. Consequently, the distraction decoy must generate a credible signature that is sufficient to preclude short-term and extended missile decoy discrimination. The AN/SLQ-49 inflatable corner reflector (Fig. 19) and the rocket-launched NATO Sea Gnat MK-216 chaff cartridge (Fig. 20) are representative of distraction decoys for surface ship defense. The TALD decoy (Fig. 21) is an example of a distraction decoy used for aircraft defense.

Figure 17. NATO Sea Gnat MK-214 seduction RF decoy deployed from a shipboard rocket launcher.

ELECTRONIC WARFARE

643

Figure 18. TORCH EO/IR decoy deployed at sea.

Figure 22. AN/ALE-50 towed decoy deployed from a tactical aircraft in flight.

Figure 19. AN/SLQ-49 inflatable corner reflector decoy deployed at sea.

Frequently, persistent seduction decoys perform a distraction function after separating sufficiently from the defended platform. This ‘‘residual distraction’’ further minimizes the number of distraction decoys required in an engagement. An EA preferential acquisition decoy provides a signature to the missile seeker such that during acquisition the missile seeker senses the real target only in combination with the decoy signature. In the end game, the decoy signature in the missile field of view biases the aim point of the missile tracker away from the intended target. The preferential acquisition concept requires decoys positioned close to the defending platform. Decoys can be towed behind the target aircraft or tethered to the defending ship. The AN/ALE-50 (Fig. 22) is a towed decoy used for air defense preferential acquisition, and the EAGER decoy (Fig. 23) is being developed for ship defense preferential acquisition. Chaff Decoys. A chaff decoy is composed of multiple—tens of thousands to millions—of electrically conductive dipole filament elements deployed in the air to reflect and scatter radar signal radiation and create a false-target radar response. Figure 24 shows a typical deployed chaff decoy. The chaff decoy frequency response is determined by the length of the dipole elements, and the chaff radar cross-sectional (RCS) mag-

Figure 20. NATO Sea Gnat MK-216 distraction decoy deployed from a rocket launcher.

Figure 21. TALD decoy distraction decoy.

Figure 23. EAGER shipboard-tethered decoy in field trials.

644

ELECTRONIC WARFARE

Figure 24. Deployed chaff round shown as a burst of reflector elements against a sky background.

nitude results from the number of dipoles deployed. Figure 25 shows a radar PPI display of an environment containing numerous chaff clouds. The RCS of a chaff cloud is tuned for a given frequency (with the dipole length one-half the wavelength of the incident radar signal), and its RCS can be approximated by 2

RCS(m ) =

0.018c2 N f2

(13)

where c is the speed of light (3 ⫻ 108 m/s), f is the frequency in hertz, and N is the number of dipoles in the cloud. Corner Reflector Decoys. Corner reflectors are conductive geometric structures that are typically shaped in the form of a perpendicular triangular corner. The shape maximizes the reflection of incident radar signals and provides a large ap-

Figure 26. Multifaceted corner reflector deployed on a ship bow to provide a high cross-sectional reflection at several frequencies.

parent target signature. Figure 26 shows a multifaceted triangular corner reflector that provides wide angular coverage. The apparent RCS normal to a triangular corner reflector is given by 2

RCS(m ) =

4πL4 f 2 3c 2

(14)

where L is the length from the outside corner to the apex of the reflector, f is the frequency in hertz, and c is the speed of light (3 ⫻ 108 m/s). The 3 dB beamwidth of this type of corner reflector is 40⬚. Flare Decoys. Flares are typically incendiary devices that produce EO/IR radiation to generate a false target. Figure 27 is an IR image of a magnesium-Teflon flare deployed from an aircraft. Active Decoys. An active decoy uses direct threat signal amplification to generate the countermeasure response. In the case of RF systems, it is generally an RF amplifier (transistor or tube). In the EO/IR spectrum, a laser or flash tube amplifies the threat signal. Jammer and repeater decoys are active decoys. Repeater decoys receive, amplify, and retransmit the received signal to generate a false target. Multiple signals may be retransmitted to generate multiple target returns. Modulation techniques (amplitude and frequency) may also be ap-

Figure 25. Radar PPI display showing target reflections from multiple chaff decoys.

Figure 27. Flare IR decoy deployed from a tactical aircraft in flight.

ELECTRON IMPACT IONIZATION

plied to the signal before retransmission to enhance effectiveness. The apparent radar cross section of an active RF decoy is given by 2

RCS(m ) =

(Pd Gd 4πR2 ) Pr Gr

(15)

where PdGd is the effective radiated power (ERP) of the decoy, R is the range between the decoy and the radar in meters, and PrGr is the effective radiated power (ERP) of the radar. For a decoy operating with linear gain, that is, a decoy whose transmission signal power is directly proportional to the input signal level (up to the signal compression level), the RCS relationship simplifies to the relationship given by 2

RCS(m ) =

(Gt c 2 ) 4π f 2

(16)

where Gt is the combined electronic and antenna gains (receive and transmit) of the decoy, c is the speed of light (3 ⫻ 108 m/s), and f is the frequency in hertz. Decoy Effectiveness. A distraction decoy is deployed at an extended range from the defending platform and provides an alternate target for seeker lock-on. Distraction decoys require deployment before seeker lock-on to engage the radar in its acquisition process. Usually more than one distraction decoy is used to defend a platform. An estimate of the effectiveness of the distraction decoy is given by 1 Ps = 1 − N+1

(17)

where Ps is the probability that the missile will be distracted to the decoy and N is the number of distraction decoys deployed. Equation (17) assumes that all of the distraction decoys exhibit viable target signatures and are equally likely to be acquired by the missile sensor. The number of decoys deployed can be reduced with the same probability of success with knowledge of the seeker acquisition logic, for example, a near-to-far/right-to-left acquisition search. Seduction decoy effectiveness is primarily determined by the intensity of the decoy signature compared with the target being defended. However, the radar track bias, for example, leading edge tracker and discrimination algorithms, can significantly impact decoy effectiveness. In some cases, the radar track bias can be exploited to increase decoy seduction effectiveness. Decoy Countermeasure Technology. Diverse technologies are required to support decoy launch and station keeping and countermeasure generation. Because most decoys are singleevent, short-term items, cost plays a major role in selecting and developing technology for decoy use. Furthermore, because the defending platform must generally deploy a number of decoys throughout an engagement, decoy size and weight criteria also are critical. Attendant decoy platform technologies include aerodynamics, aircraft/projectile design, propulsion systems, avionics, and mechanical structures. Decoy payload technologies that will have significant importance in

645

future systems include broad bandwidth microwave and millimeter-wave components (e.g., antennas and amplifiers). Microwave and millimeter-wave output power sources are required with high power, efficiency, and duty cycle to support the projected threat environments. The future RF threat environment is expected to be densely populated with longpulse radar. Higher decoy radiated power at higher duty cycles will be needed to prevent decoy saturation as the number of simultaneous threat signals in the environment increases. Ultra high speed countermeasure frequency set on circuitry is necessary to queue jammer frequency rapidly. Signals with rapid frequency hopping and frequency chirping require rapid activation for effective countermeasures. Spatially large and efficient spectrally matched IR materials and radiating structures are needed to counter multispectral, imaging IR seekers. Safe, nontoxic, highly opaque, broad-spectrum IR and electro-optical obscuration materials are required to mask targets and confuse image-processing seekers. Efficient, primary power sources capable of high peak power and dense energy storage are needed to provide the increasing demand for electrical power used in decoy systems. Reading List J. S. Accetta and D. L. Shumaker (eds.), The Infrared and ElectroOptical Systems Handbook; D. H. Pollock (ed.), Vol. 7, Countermeasure Systems, Ann Arbor, MI: Infrared Information Analysis Center, and Washington, D.C.: SPIE Optical Engineering Press, 1993. B. Blake, Jane’s Radar and Electronic Warfare Systems, Surrey, U.K.: Jane’s Information Group, 1993. J. A. Boyd et al., Electronic Countermeasures, Los Altos, CA: Peninsula Publishing, 1978. E. J. Chrzanowski, Active Radar Electronic Countermeasures, Norwood, MA: Artech House, 1990. N. C. Currie, Techniques of Radar Reflectivity Measurement, Dedham, MA: Artech House, 1984. R. D. Hudson, Jr., Infrared Systems Engineering, New York: WileyInterscience, 1969. W. L. McPherson, Reference Data for Radio Engineers, New York: Howard W. Sams, 1977. R. J. Schlesinger, Principles of Electronic Warfare, Los Altos, CA: Peninsula Publishing, 1961. M. I. Skolnik, Radar Handbook, New York: McGraw-Hill, 1970. L. B. Van Brunt, Applied ECM, Vol. 1, Dunn Loring, VA: EW Engineering, 1978. W. Z. Wolfe and G. J. Zississ (eds.), The Infrared Handbook, revised ed., Ann Arbor, MI: Environmental Res. Inst. Michigan, 1985.

ANTHONY E. SPEZIO ALAN N. DUCKWORTH FRANCIS J. KLEMM STANLEY A. MOROZ JAMES M. TALLEY Naval Research Laboratory

670

HELICOPTER NIGHT PILOTAGE

HELICOPTER NIGHT PILOTAGE In 1971, the United States Army determined that, in order to survive on the modern battlefield, tactical helicopters had to fly very near the ground and hide behind terrain contour or trees. Flying at very low altitude, masked by hills and trees, was required in order to overcome the threat of enemy ground to air weapons. Flight to and from the battle area is at high speed and constant altitude above the ground, generally less than thirty feet above the terrain or local obstacles. This is called contour flight. Flight in the battle area is nap-of-the-earth (NOE). During NOE flight, at least part of the aircraft is below treetop level, and the aircraft flies around obstacles rather than over them in order to remain hidden. NOE and contour flight requires night imaging sensors with field of view (FOV) and resolution sufficient to allow the pilot to fly the aircraft near trees and other ground obstacles. The night pilotage task is very demanding on both the aviator and the helicopter night sensors. A helicopter night pilotage sensor should allow the pilot to fly ‘‘heads up and eyes out’’; the system should provide the same type of contextual information at night which allows the pilot to orient and fly the aircraft during the day with unaided vision. The sensor should provide an image that permits the pilot to perform precision aircraft movements in a confident and aggressive manner. The sensor should permit the pilot to discern terrain features for navigation, select low-level flight paths, and detect possible threats. A good pilotage sensor will also maximize the fraction of time that at least minimal performance can be gained from the sensor in order to execute a mission. NIGHT PILOTAGE SENSORS CURRENTLY IN USE Image Intensifiers The first fielded imaging aid used for low-level night pilotage was the AN/PVS-5 Night Vision Goggle which was adopted from ground use. The AN/PVS-5 goggle is shown in Fig. 1. This sensor uses image intensifier (I 2) tubes which amplify moonlight and starlight. The goggle amplifies visible light and provides a considerably brighter image to the pilot than would be available without the goggle. The goggle provides a binocular image (an image to both eyes) with 40⬚ circular FOV. To illustrate this field of view, a 19-inch television set viewed from 21 inches would provide about the same field of view to the eye as the goggles. The goggle image, however, is optically projected as a virtual image that appears to be outside the aircraft; this relieves eye strain and makes the image appear more natural. The image is unity magnification, meaning that objects appear life-sized. Under optimal light conditions, the AN/PVS-5 goggles have a limiting resolution of 0.7 cycles per milliradian (cy/ J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

HELICOPTER NIGHT PILOTAGE

671

Figure 1. The AN/PVS-5 goggle provides a good image with moonlight illumination. In use, it covers the entire upper portion of the face.

Figure 2. The ANVIS goggle provides a good image with moonlight or starlight illumination. The pilot can view instruments by looking under the goggle.

mrad) which is equivalent to a visual acuity of about 20/50. (When an optometrist says that you have ‘‘20/50 vision,’’ he means that you can read the same size letters at 20 feet as are legible to most people at 50 feet. The human eye resolution at the 20/20 level corresponds to the ability to resolve roughly one minute of arc.) Experience with the ground goggle showed it to be a significant aid for night flight. Two significant problems were encountered, however. In use, the ground goggle covers the entire upper portion of the face, so that the pilot viewed both the outside world and aircraft instruments through the goggle. The goggle optics could not be focused to simultaneously show both the nearby instruments and the outside world. The second problem with the ground goggle was that it provides a good image only when the moon is up; flying with these goggles was difficult under starlight illumination conditions. The development of an I 2 goggle specifically designed for aviation use was initiated in the late 1970s. The new goggle was designated the AN/AVS-6 Aviator’s Night Vision System (ANVIS). ANVIS mounts to the pilot’s helmet as shown in Fig. 2 and allows the pilot to view his instruments by looking under the goggle. ANVIS can also be rotated up to a stow position on top of the helmet, leaving the pilot’s vision completely unobstructed. ANVIS provides a good image under starlight illumination conditions. In addition to being more sensitive than the AN/ PVS-5 in responding to visible light, the ANVIS spectral band encompasses more of the ambient light available at night. ANVIS responds to near infrared light as well as to visible light. ANVIS provides a 40⬚, binocular, unity magnification image with better resolution than the original ground goggle. Under optimal illumination conditions, ANVIS limiting resolution is about 0.9 cy/mrad corresponding to a limiting acuity of 20/40. The AN/AVS-7 Heads Up Display (HUD) was added to ANVIS in the early 1990s; it is a small apparatus which clamps onto one of the ANVIS oculars. The HUD superimposes instrument symbology on goggle imagery, allowing the pilot to see important information like altitude, heading, and

gyro horizon without looking inside at the cockpit instruments. Figure 3 illustrates symbology superimposed on ANVIS imagery. The HUD allows the pilot to keep ‘‘heads up and eyes out,’’ because the pilot need not focus his eyes and attention inside the cockpit to view important instrument information. The primary problem with using ANVIS on helicopters is lack of compatibility with the cockpit instrument lighting. Modern image intensifiers amplify ambient light 2000 to 3000 times; cockpit lights can blind the goggles due to reflected glare off the canopy or off other objects in the cockpit. The problem is corrected by adding a spectral filter to ANVIS which rejects blue-green light, and only blue-green instru-

Figure 3. Flight symbology is superimposed on the ANVIS imagery; the pilot does not need to look inside the cockpit to see important aircraft status information.

672

HELICOPTER NIGHT PILOTAGE

ment lighting is used on the newer Army helicopters. Red light is avoided because ANVIS is quite sensitive to red light. Lighting requirements for ANVIS compatibility are discussed in ref. 1. Thermal Imagers In 1973, development was initiated on the first thermal imager for pilotage use. The AN/AAQ-11 Pilot’s Night Vision System (PNVS) was developed for the AH-64 Apache Advanced Attack helicopter. PNVS is a gimbaled thermal imager mounted on the nose of the helicopter. The position of the PNVS on the helicopter is shown in Fig. 4. The PNVS images 8 애m to 12 애m thermal energy (that is, heat) and provides a 40⬚ horizontal by 30⬚ vertical FOV. The pilot is in the cockpit, while the PNVS thermal imager is on the nose of the aircraft. The system hardware must provide some means of pointing the sensor where the pilot wants to look and some means to remote the thermal image back to the pilot in the cockpit. Figure 5 illustrates how this is accomplished on Apache. A helmet tracker slaves the sensor line of sight to the pilot’s head. The pilot wears a helmet-mounted display through which he views the thermal image. The helmet display projects a virtual image which appears to be outside the aircraft. The helmet-mounted display is monocular, viewed with the right eye only, and provides the same 30⬚ vertical by 40⬚ horizontal field of view as the sensor. The system therefore provides a unity magnification, thermal image of the world which the pilot can orient by moving his head. A second thermal imager is available on the Apache helicopter. The second thermal imager is one of several sensors in the AN/ASQ-7 Target Acquisition and Designation System (TADS); the TADS is the large, barrel shaped object located below the PNVS shown in Fig. 4. This imager is normally

Figure 5. Pilot wears a helmet mounted display in front of right eye; he uses this to view the PNVS thermal imagery. A helmet tracker turns the PNVS sensor to match the pilots head movement.

used by the copilot/gunner to locate and engage targets. However, the TADS thermal imager has three fields of view with the wide field of view identical to the PNVS field of view. The copilot/gunner can use the TADS image in a pilotage mode in exactly the same way that the pilot uses the PNVS. A helmet tracker senses the copilot’s head motion and moves the TADS to align the line of sight of the thermal imager. The copilot views the image via a helmet-mounted display. Heads-up instrument symbology is an integral part of the PNVS and TADS systems on the Apache helicopter. Both pilot and copilot can view important flight and status information superimposed on the thermal imagery. With symbology superimposed on his night vision imagery, the pilot does not have to focus his eyes inside the cockpit to determine critical information such as altitude, heading, or caution status. Combinations of Thermal Imagers and Image Intensifiers

Figure 4. The PNVS thermal imager mounted on the front of the Apache Helicopter. The TADS system is the barrel-shaped object with two windows mounted beneath the PNVS.

In 1987, an adapter was designed to permit the ANVIS to be mounted on the Apache copilot’s helmet. The adapter allows the ANVIS to be mounted simultaneously with the Apache helmet display, although ANVIS and the helmet display cannot be viewed simultaneously. When the copilot is using ANVIS, the TADS thermal imagery and symbology can be viewed on a panel display by looking under the ANVIS. The copilot can use the ANVIS imagery and periodically cross reference the thermal imagery as a safety check. If the copilot is using the helmet-mounted display and TADS thermal sensor, the ANVIS is placed in the stow position on top of the helmet. In the late 1980s, the Helicopter Night Vision System (HNVS), AN/AAQ-16, was fielded on some UH-60 Blackhawk Utility helicopters and on some CH-47 Chinook Cargo helicopters. The HNVS is a thermal imager which operates on similar principles to the PNVS and the TADS. The HNVS is mounted on the nose of the aircraft and is viewed via a panelmounted display in the cockpit. The HNVS is not head tracked, but can be pointed by a hand controller. The sensor has two fields of view. The wide FOV is 30⬚ vertical by 40⬚ horizontal; the narrow FOV is 5⬚ vertical by 7⬚ horizontal.

HELICOPTER NIGHT PILOTAGE

Both pilot and copilot use ANVIS to fly. The panel displayed HNVS imagery is used to cross reference and verify the information provided by the ANVIS. The aviators use HNVS as a backup, and as a cross reference for terrain avoidance, target location, check point verification, and during low illumination or poor visibility conditions where ANVIS vision is degraded. The newest Army helicopter, currently in development, is the RAH-66 Comanche; Comanche is a reconnaissance and light attack helicopter. The Comanche Night Vision Pilotage System will integrate an advanced, high -resolution thermal imager, an I 2 camera, and flight symbology into a single package. The pilotage sensors will be mounted on the nose of the aircraft in a manner similar to Apache; however, the nose turret will include both thermal and I 2 sensors. The pilot will wear a binocular helmet display rather than the monocular display worn by Apache aviators. The field of view of the NVPS with the new helmet-mounted display will be 30⬚ vertical by 52⬚ horizontal.

673

of fiberoptic bundles with the core etched away. The plate has millions of channels (holes) with photoemissive material on the inside of the channels. Each face of the MCP is metalized, and a high voltage is applied across the plate. As electrons strike the inside of the MCP channels, secondary electrons are emitted. Multiple secondary electrons are emitted for each cathode electron. The secondary electrons are accelerated by the voltage along the channel, the secondary electrons strike the channel wall and cause more electrons to be emitted, and the electron multiplication process is repeated. The amplified electrons from the MCP are accelerated to the phosphor, where a brighter version of the cathode image is formed. The fiberoptic twist erects this image. The eyepiece magnifies the image for presentation to the eye. ANVIS provides a scene to eye light gain of about 3000. In the absence of fog or obscurants, ANVIS performs well under clear starlight illumination. Generally, ANVIS provides good imagery with naked-eye visibility exceeding 200 m to 300 m and minimum light levels of 7E-5 footcandles (2). Thermal Imagers

SENSOR THEORY OF OPERATION Image Intensifiers The image intensifiers used in ANVIS amplify ambient light, moonlight, and starlight, at spectral wavelengths between 0.5 and 0.9 애m. A schematic of a goggle ocular is shown in Fig. 6; binocular goggles use two oculars to provide an image to both eyes. An inverted image of the scene is formed on the cathode by the objective lens. The cathode emits photo electrons; the shot noise associated with cathode photoelectrons dominates the performance of image intensifiers. Photoelectrons from the cathode are accelerated to the microchannel plate (MCP) by a voltage difference applied between the cathode and MCP. The MCP acts as an electron multiplier and provides most of the gain of the I 2 tube. A detail of the MCP is shown at the bottom of the Fig. 6. The MCP is a thin, glass plate made up

Objective lens

Cathode

MCP

Phosphor

Fiber optic twist

Thermal imagers like the Apache helicopter PNVS detect radiation in the 8 애m to 12 애m spectral band. This band is chosen because the atmosphere has a ‘‘window’’ where the transmission of thermal energy is good. Everything near room temperature radiates at these wavelengths. The emissivity of natural objects is generally above 70%; most human-made objects are also highly emissive. It should be noted, however, that thermal sensors derive their images from small variations in temperature and emissivity within the scene. Typically, the thermal scene is very low contrast even under good thermal viewing conditions. Scene thermal contrast is affected by the amount of solar heating during the day. Thermal contrast is decreased by the presence of clouds. Thermal contrast can be poor at night, particularly after extended periods of clouds or precipitation. In current thermal imagers like the PNVS, a linear array of infrared detectors is used. Figure 7 illustrates the theory

Eyepiece

Electrons Microchannel plate (MCP)

Figure 6. Theory of operation for an image intensifier. The microchannel plate is illustrated at the bottom.

674

HELICOPTER NIGHT PILOTAGE

Scan mirror Afocal optics

Imaging lens

Detector array

Light (thermal energy)

Figure 7. Theory of operation for a thermal imager.

of operation. The afocal optics provide a magnified image of the scene at the scan mirror. The linear array of detectors is scanned over the image by the oscillating mirror. The image is formed by rapidly sampling each element of the detector array as it is scanned over the whole image area. A video image is formed electronically from the detector samples; the video image is viewed via the helmet-mounted display. The linear array in PNVS has 180 detectors; interlace is used to generate 360 active lines in the image. Interlace is achieved by nodding the scan mirror alternately up and down a small amount after each sweep of the field of view. Detector noise dominates the performance of these imagers. PNVS provides usable imagery with tree to ground equivalent blackbody temperature differences greater than 0.3 K; performance with less than 0.1 K temperature difference is poor (2). PILOTAGE SENSOR PERFORMANCE The performance of a pilotage aid depends on the image delivered to the pilot during flight. Depending on the weather and other factors, the image can fade, become very noisy, and even disappear completely. The image quality of image intensifiers and thermal imagers is affected by ambient atmospheric conditions and the nature of the local environment. The I 2 image quality depends on available illumination from the moon and stars, on atmospheric visibility conditions, and on the diversity and contrast of ground objects in the local area. Thermal image quality depends on thermal contrast within the scene and on atmospheric transmission in the thermal spectral band. Thermal contrast is increased by solar heating during the day and is reduced by heavy or prolonged cloud cover or precipitation. User surveys were conducted in 1987, 1990, and after Desert Storm in 1992 (3–6). Structured flight evaluations have also been performed (2,3,4,7). These surveys and evaluations provide insight into the environmental conditions under which the pilotage systems perform well. While it is straightforward to define good and poor weather and environment conditions for ANVIS and PNVS usage, it is very difficult to define the conditions which are safe. An aviator will change the aircraft airspeed, altitude, and flight profile as needed to adapt to the conditions encountered. As night sensor imagery degrades, the pilot will also depend more on the instruments and the HUD symbology. The engineering trades for a night vision sensor relate to the ability of the

Electric reformat and display

sensor to deliver the desired visual information; these trades do not relate to the ability of the entire weapon system to accomplish a mission. When there is good thermal contrast in the scene, and in the absence of fog, heavy rain, or snow squalls, the PNVS thermal imager supports terrain (NOE and contour) flight. Good thermal contrast occurs when there has been clear weather with sunshine for at least a few hours during the day, heating up objects in the background scene. If there has been no sunshine during the day, or if there has been only a little sunshine followed by heavy rain or hours of drizzle, the thermal contrast will be poor, leading to poor visual flying conditions. Further, the thermal radiation which PNVS images is attenuated by heavy fog and by the atmospheric water vapor content found with heavy rain and persistent drizzle. Image contrast might be poor even when the scene is providing a usable thermal signature. Thus, poor local weather, such as patches of fog or squalls, may make terrain flight difficult at the midpoint of a flight, even though conditions are good at the starting point and destination. ANVIS performs well under clear starlight conditions but becomes marginal to unusable under overcast starlight conditions. Heavy fog will shut down ANVIS. Even a moderate fog can severely degrade imagery if flight is toward the moon; scattered light from the fog can severely degrade contrast and mask the view of the terrain. Also, ANVIS tends to ‘‘bleach out’’ or shut down when bright lights are in the field of view; this occurs around city lights, when flying toward the moon if it is low on the horizon, and under dawn and dusk conditions. Flying over land that has no features, such as the sand dunes of Saudi Arabia, presents a challenge; judging distance and closure to the ground requires scene detail. Areas devoid of distinguishable features, such as snow fields, lakes, and dry lake beds, will provide poor imagery for terrain flight. Under these circumstances, the availability of flight symbology is critical. Pilots express strong feelings that thermal sensors and image intensifiers are complimentary and that both are needed for night contour and NOE flight. The combination supports flight under a wider range of conditions than either alone, although environments certainly exist where even the combination will not support terrain flight. Also, each sensor brings a unique capability to the aircraft. The two sensors operate in different spectral bands and depend on different physical principles for performance. The

HELICOPTER NIGHT PILOTAGE

ability of the aircrew to detect wires and other obstacles is significantly enhanced. Even on poor thermal nights, the PNVS and HNVS provide a good capability to perceive and react to close in targets. Even on nights with poor illumination, ANVIS gives the ability to see town lights and therefore provides navigational aid; because ANVIS can see aircraft running lights, it also provides a better ability to fly formation as well as safety from collision with other aircraft. DATA RELATING TO DESIGN IMPROVEMENTS On the basis of feedback from pilot interviews, current night vision sensors like the ANVIS, PNVS, TADS, and HNVS provide a significant improvement in mission effectiveness over previous techniques of flying and fighting at night. Apache aviators stated that the thermal pilotage and targeting sensors on Apache (the PNVS and TADS systems) completely changed their capability to fight at night so that comparisons to previous aircraft were not meaningful. It is also clear from the pilot surveys, however, that further enhancement of night effectiveness can be gained from further hardware development. In recent years, the quality of both image intensified and thermal imagery has improved substantially. Even with advanced technology, however, optimizing the design of electrooptical pilotage sensors involves trade-off of resolution, field of view, and sensitivity. At any given level of technology, for example, an increase in the sensor field of view requires a decrease in sensor resolution or a decrease in sensitivity or both. Further, the optimum performance trade-off of imaging sensor parameters depends on specifying the visual task. Night helicopter pilotage involves many visual tasks. Flying a helicopter near the ground involves judging distance, judging closure to terrain or terrain objects, maintaining orientation of the aircraft, looking for a suitable flight path, searching for obstacles and threats, and other visual tasks. Over the years since the mid-1970s, responsible Army organizations have undertaken field surveys of operational users, flight evaluations, and flight experiments in order to develop design criteria for helicopter night pilotage systems. These efforts have focused on determining the fraction of time that existing pilotage sensors support mission accomplishment and on finding sensor design parameters which optimize flight handling. These efforts are summarized. User Feedback on FOV and Resolution In each of the three surveys taken between 1987 and 1992, the aviators were asked to answer questions, based on their total flight experience, about needed design improvements in field of view and resolution for ANVIS and PNVS. In an operational context, sensor resolution refers to image quality and therefore depends on the sensor sensitivity as well as the optical resolving power of the sensor. The results of all the surveys are consistent and can be summarized as follows. Based on total flight experience, pilots rate both the FOV and the resolution of ANVIS as acceptable. Pilots would choose to expand ANVIS FOV but not at the expense of current image quality. On the basis of total flight experience, pilots rated the PNVS FOV as adequate but the resolution as inadequate; they would improve image quality

675

Table 1. 1987 Survey: Pilot Rating of PNVS and ANVIS FOV and Resolution Sensor/Feature PNVS FOV PNVS Resol. ANVIS FOV ANVIS Resol.

Good

Adequate

Inadequate

5 1 9 13

35 18 17 13

9 30 3 3

before expanding FOV. The pilots are interested in increased PNVS FOV but only in combination with improved image quality. A summary of the responses to each survey is given below. The 1987 survey queried 49 Apache helicopter pilots, all with PNVS thermal imager experience; 29 of these aviators had ANVIS experience (3,4). When given an open choice of which sensor they preferred, 42 of 49 wanted both PNVS and ANVIS. The Apache crews were asked to give an overall rating for PNVS and ANVIS as to adequacy of FOV and resolution (image quality); they were to answer based on their total flight experience. Table 1 summarizes how many pilots rated FOV and resolution as good, adequate, and inadequate. In general, the pilots rated the PNVS FOV as adequate but the resolution as inadequate. They rated both the FOV and resolution of ANVIS as adequate. The large majority of Apache aviators, 45 out of 49, would improve PNVS resolution before expanding FOV. The opinion on ANVIS was about evenly split between improving resolution and FOV. However, two cautions were emphasized by the respondees. First, these numbers do not reflect a lack of interest in increased FOV if it accompanies improved image quality. Second, the user will not accept a smaller FOV than currently provided. The 1990 survey involved 52 ANVIS aviators from three units flying a variety of missions (5). Twenty of the ANVIS aviators regularly used the HNVS thermal imager in addition to ANVIS. Twenty-one PNVS aviators were also surveyed; eighteen of the PNVS aviators also used ANVIS. Again, when given an open choice of sensor, the overwhelming majority chose a pilotage system with both thermal and image-intensified imagery. The aviators were asked to give an overall rating for PNVS and ANVIS as to adequacy of FOV and resolution (image quality); they were to answer based on their total flight experience. Table 2 below summarizes their answers. Seventeen of the twenty-one Apache aviators would improve PNVS resolution rather than expanding FOV with the current resolution. Fifty of the ANVIS aviators would expand ANVIS FOV if the current ANVIS resolution could be maintained.

Table 2. 1990 Survey: Pilot Rating of PNVS and ANVIS FOV and Resolution Sensor/Feature ANVIS FOV ANVIS Resol. PNVS FOV PNVS Resol.

Good

Adequate

Inadequate

16 32 2 0

45 36 18 9

8 1 1 10

676

HELICOPTER NIGHT PILOTAGE

The 1992 survey was conducted after Desert Storm (6). No area is as devoid of distinguishable terrain features on such a scale as Saudi Arabia. The sand dunes lacked almost any vegetation and had rises and falls varying as much as 75 feet. The lack of features made the terrain relief imperceptible through the night vision sensors. This was a difficult area in which to use night vision sensors. Of 66 aviators surveyed, 70% judged ANVIS performance in Saudi Arabia to be good or adequate. What should be noted is that the 30% inadequate rating was never experienced elsewhere. Of the 34 Apache aviators surveyed, 70% rated the PNVS performance in Saudi Arabia as good or adequate. Thermal conditions were better at the beginning of the war, and image intensifier conditions were better at the end of the war. Aviators with a choice used both systems about half the time. The FOV of both systems was rated as adequate. Of the 34 Apache aviators, 55% rated the PNVS and TADS resolution as inadequate and 75% felt that improving resolution took precedence for a design improvement. Although image quality was a problem in Saudi Arabia, 60% of the 66 ANVIS aviators felt that improving FOV should take precedence based on their total flight experience; another 15% felt that improving FOV and resolution should take equal priority. Flight Experiments The flight experiment results can be summarized as follows. With normal eyesight acuity, performance improves with FOV up to a plateau between 40⬚ and 80⬚ depending on flight maneuver. However, degraded visual acuity strongly affects these results. Once a minimum FOV of about 40⬚ is achieved, performance is a strong function of image quality. Holding the sensor FOV to 40⬚ and optimizing image quality is usually the best design tradeoff. Increasing FOV by diverging ocular lines of sight (that is, both eyes see the center third of the total FOV, but the outer third on each side is seen by only one eye) does not improve performance and may hurt performance. Although the total FOV is increased, the data indicate that fixations and ocular tracking are limited to the central, overlapped region of the FOV. In some important respects, the sensor FOV becomes the small, overlapped region. Based on pilot assessment of flight trials, a detector dwell time (exposure time) of 16 ms is unacceptable in a pilotage system; a dwell time of 4 ms is not noticeable. Also, image processing delays (the time delay between capture of the image by the sensor and display of the image to the pilot) should be 33 ms or less. Delays of 100 ms lead to serious flight control problems. FOV and Resolution Trades. In 1975, the U.S. Army Aeromedical Research Laboratory performed a flight test comparing standard 40⬚ FOV, AN/PVS-5 goggles to modified goggles with a 60⬚ FOV (8). On the basis of the flight conditions, the limiting resolution of the 40⬚ and 60⬚ goggles was 0.6 and 0.4 cy/mrad, respectively. Participating aviators rated the 40⬚, higher resolution goggle as more suitable for terrain flight. Also, the 40⬚ goggles were associated with smoother, more gradual control stick movements than the lower resolution, 60⬚ goggles.

During 1985, a flight experiment was conducted by the NASA Ames Research Center to determine the visual cues essential for low speed and hover flight (9). This test was conducted in order to determine the importance of field of view and resolution on the fidelity of flight simulators. The variables in this flight test were field of view, the amount of macrotexture (large objects), and the amount of microtexture (fine detail) in the imagery. Field of view was varied by masking portions of the windscreen. Microtexture was varied with a set of liquid crystal goggles which selectively fogged the image. Macrotexture was varied by changing flight location and by laying objects like tires on the ground near the flight path. The test fields of view ranged from a 10 by 14⬚ rectangular window to a multiwindowed case encompassing 9000 square degrees. Two resolutions were used: 20/15 visual acuity, which is normal for these pilots, and 20/40 degraded visual acuity. Subject pilot ratings indicated that low speed and hover flight can be performed with reasonable workload using a 23 by 38 degree FOV with normal visual acuity. Also, when acuity was degraded, increasing field of view resulted in little improvement in pilot ratings. The effects of FOV and limiting resolution on flight handling were explored in two flight experiments performed by the Army’s Communications and Electronics Command in the late 1980s (10,11). Direct-view goggles were built to provide various combinations of FOV and resolution. These goggles are similar to ANVIS except they do not incorporate an image intensifier and are used during the day only. Pilots using these goggles were asked to fly preplanned NOE and contour flight profiles. Hover and lateral flight tasks were also evaluated. In both tests, trail runs were flown without goggles to establish baseline performance levels. The aircraft used was an AH-1 COBRA Attack helicopter with the subject pilot in the front seat. The aircraft and flight profiles were selected after consultation with test and user pilots. Six subject pilots participated, each flying three trials of each task. Measured data included altitude, airspeed, and head motion. After each trial of each task, pilots answered questions on workload, confidence, and aircraft handling qualities. Table 3 shows the combinations of resolution and FOV flown on a test range at Fort Rucker, Alabama in February, 1987. The term ‘‘ocular overlap’’ in Table 3 is described as follows. With 100% overlap, both eyes see the whole field of view. One technique to enlarge the display FOV while maintaining

Table 3. FOV and Resolution Combinations Flown in 1987 Experiment FOV in Degrees Unrestricted 40 40 40 40 ⫻ 60 60 60 60 ⫻ 75

Limiting Resolution

Ocular Overlap (%)

Normal eyesight Normal eyesight 0.9 cy/mrad 0.6 0.9 0.6 0.5 0.6

Normal 100 100 100 50 100 100 75

HELICOPTER NIGHT PILOTAGE

high resolution is to partially overlap the two oculars of a binocular display. With partial overlap, both eyes see the central portion of the FOV, but only one eye sees each edge of the FOV. For example, 50% overlap of a 60⬚ goggle means that both eyes see the central 30⬚ of the field of view. The right eye sees the right 15⬚ of the total field of view, and the left eye sees the left 15⬚ of the total field of view. This technique lets the optical designer reduce weight and volume by covering a large total FOV with smaller individual oculars. The test device with 40⬚ FOV and with 0.6 cy/mrad resolution represents current thermal imager capabilities under very favorable thermal contrast conditions. This combination also represents the capabilities of ANVIS night vision goggles under quarter moon illumination. With the exception of the device with 40⬚ FOV and normal eyesight resolution, the other combinations shown in Tab. 3 represent achievable performance in the 1990s time frame under good thermal contrast or high light level conditions. The following observations were made based on the Fort Rucker test: 1. When FOV was held constant at 40⬚, decreasing resolution resulted in a substantial increase in altitude, a slight decrease in airspeed, and significantly poorer pilot ratings. 2. Decreasing FOV to 40⬚ but retaining undegraded visual acuity had a very minor impact on altitude and airspeed. Pilot ratings for this combination were slightly below the unrestricted baseline but were better than all other combinations tested. 3. With the 40⬚ FOV, 0.6 cy/mrad device as a baseline, increasing either FOV or resolution with fully overlapped oculars improved performance and significantly elevated pilot ratings. When comparing the 40⬚ FOV with 0.9 cy/mrad goggles to the 60⬚ FOV with 0.6 cy/ mrad device, pilots had some preference for the wider FOV but exhibited no change in performance. 4. Increasing FOV by diverging ocular lines of sight (that is using less than 100% overlap of the images presented to the two eyes) did not improve performance when the 40⬚ oculars were used and caused poorer performance with the 60⬚ oculars. The 50% partial overlap of the 40⬚ oculars resulted in increased head motion and fatigue. Distortion for the 40⬚ oculars was less than 1%. However, distortion in the 60⬚ oculars reached 6%; high distortion will undoubtedly cause image convergence problems between the two eyes and lead to degraded performance. The FOV/resolution combinations tested at Fort Rucker represented performance projected to be attainable under favorable thermal contrast or high light level conditions. A second test was flown at Fort A.P. Hill, Virginia, to explore the resolution versus field of view trade-off when simulating less than ideal thermal contrast or light level conditions. The FOV/resolution combinations which simulated less than ideal conditions were chosen to make the flight tasks difficult but possible. The potential benefit of trading lower resolution at the edge of a sensor field of view for higher reso-

677

Table 4. FOV and Resolution Combinations Flown in 1988 Experiment FOV in Degrees

Limiting Resolution

40 40 40 60 60 60

0.9 cy/mrad 0.4 0.5 at edge/1.1 at center 0.6 0.3 0.2 at edge/0.9 at center

lution at the center was also evaluated. Table 4 gives the combinations evaluated in the second test which was flown during February and March, 1988. Four subject pilots participated; each subject flew four trails of each task. During this test, goggle configuration did not affect altitude and airspeed performance. Once the task was defined in the baseline flight, execution did not vary significantly in terms of the airspeed or altitude which was maintained. The highest workload and lowest confidence ratings were given to the 60⬚, 0.3 cy/mrad goggle simulators. In this test, the pilots consistently selected the higher resolution and smaller field of view devices over the larger field of view but lower resolution devices. If resolution at the edge of a 60 degree device was substantially poorer than resolution at the center, two of the pilots consistently rated the 40 degree field of view goggles higher even when the 60 degree goggles had equivalent or better resolution in the central portion of the field of view. The other pilots rated these 40⬚ and 60⬚ devices as equal. After test completion, the pilots were asked to explain this preference. The response was that, with the 60⬚ goggles, they would see an object ‘‘and then lose it.’’ This characteristic of the goggles was particularly bothersome during the 360⬚ hover turn out of ground effect but also affected performance during lateral flight, NOE, and contour flight. It is likely that ocular tracking is important in the performance of all these tasks and that poor resolution at the edge of the field of view would therefore lead to adverse pilot reaction. However, ocular tracking was not measured during the test. During 1994, a flight test was conducted to test the hypothesis that using an 18⬚ ocular overlap in a 52⬚ total FOV might result in abnormal eye and head movement patterns (12). A fully overlapped design was also flown for comparison. The flight test further determined if the difference would impact pilot performance of the prescribed flight tasks. Flight tasks included NOE, contour, out of ground effect hover, and lateral flight. On the basis of the eye tracking data collected during the flight, the partial overlap does constrain the eye at the center of the FOV and significantly reduces the amount of time that the eye uses the outer portion of the total FOV. Averaged across all pilots and tasks, the percentage of eye fixations that occur outside the central 18⬚ when using partial overlap was reduced by 60% (p ⫽ 0.0170) as compared to the full overlap (full ⫽ 24%, partial ⫽ 9%). There is no difference between tasks (p ⫽ 0.2836). Looking at horizontal eye movement, the mean rms amplitude across the five subjects for the partial overlap was only 70% of the rms for the full overlap. This 30% reduction was significant (p ⫽ 0.0136). No statistically significant difference

678

HELICOPTER NIGHT PILOTAGE

in rms amplitude was found between tasks (p ⫽ 0.5022) or for the interaction between overlap and task (p ⫽ 0.7769). The average head velocity for partial overlap increases by 12.5% and 6% for contour and NOE flights, respectively. The pilots indicated higher workload and lower confidence when flying the partial overlap as opposed to the full overlap. Some subjects reported nausea and fatigue after use of the partial overlap; this occurred whether the partial overlap configuration was flown first or second. There was no noticeable visual problem reported on the full overlap configuration. Overall, these results indicate a change in characteristic head and eye motion when the partial overlap is used. There is a 10% increase in average head velocity and a significant 45% increase in the fraction of time that the head is in motion. The data may suggest that the more frequent head dynamics may be substituting for the lack of the ocular tracking which is restricted (60% reduction) when the partial overlap design is in use. This appears to be consistent with the hypothesis that the eyes do not track across the overlap (binocular to monocular) boundary. The subjective data suggest that the partial overlap effectively segregates the overall 52⬚ FOV into an 18⬚ brighter central and two dimmer outer regions. This perceived decrease in brightness and acuity apparently derives from the lack of binocular fusion in the outer regions. The subjects indicated that luning at the overlap boundary hid scene cues; they subjectively rated the partial overlap FOV as being smaller than the fully overlapped FOV. It appears that the partially overlapped configuration limits ocular tracking, both because of the perceived loss in image quality at the overlap boundary and because of the loss of binocular fusion as the eye tracks over the boundary. The partially overlapped FOV configuration provides a functionally smaller FOV than the fully overlapped configuration. An experiment conducted in 1996 evaluated the impact of field of view on precision flight maneuvers (13). Subjects flew with FOV restricted to 40⬚ vertical and 20, 40, 60, 80, and 100⬚ horizontal. Normal eyesight acuity was not degraded. Maneuvers included pirouette, hovering turn, bob-up and down, precision landing, acceleration and deceleration, and slalom. Performance measures included accurate aircraft position and heading, head movement, pilot rating of flight handling qualities, and pilot rating of visual cues. Most of the measured data showed a general increase in performance with larger FOV. Flight data indicated that performance improves with FOV up to a plateau between 40 and 80⬚ depending on the flight maneuver. Subjective ratings of flight handling and visual cues increased with FOV up to a limit of 60 to 80⬚ depending on task. On the basis of all the collected data, it was the researcher’s opinion that the greatest overall performance gain occurred prior to the 60 to 80⬚ FOV range under the conditions tested. Image Blur Due to Head and Sensor Motion. A flight test was conducted to determine suitable exposure time for a staring camera operating at the standard video frame rate (11). Cameras which use ‘‘staring’’ detector arrays are being considered for use in night pilotage aides. Most staring sensors use detector dwell times equal to the field or frame time of the imager, typically either the 60 Hz video field time or the 30-Hz video frame time. In a pilotage sensor, however, considerable image movement can occur in a video field time due to aircraft and

head motion. The pilot will see a blurred image for the same reason that a photograph will be blurred if the exposure time is too long for the motion being captured. Two pilots flew an AH-1 Cobra from the front seat using helmets and helmet-mounted displays from the Apache helicopter with a small video camera mounted on the helmet. The camera FOV was 30⬚ vertical by 40⬚ horizontal and provided unity magnification through the helmet display. The test camera had a limiting resolution of about 0.5 cy/mrad and electronic gating to control the dwell time for each video field. Selectable exposure times ranged from 1/60 s (one field) to under a millisecond. The pilot’s visor was down and taped so that he flew solely by sensor imagery. The pilots performed hover, lateral flight, NOE, and contour tasks. The flight experiment was performed in January, 1989, at Fort A.P. Hill, Virginia. Image blur at 1/60 s exposure time was unacceptable. Blur was present with either aircraft or head motion, and the blur interfered with task accomplishment. With an exposure time of 1/120 s, image blur was noticeable with head motion but no conclusion was reached regarding impact on performance. No image blurring was noted at 1/240 s exposure time. Visual acuity is not degraded for ocular tracking rates up to about 30⬚ per second, and ocular tracking is probably important during pilotage. The exposure time for each snapshot taken by a video camera should be short enough that images crossing the sensor FOV at up to 30⬚ per second are not blurred. Note that acceptable exposure time depends on sensor resolution; exposure time should shorten as sensor limiting resolution improves. Impact of Image Processing Delays. In advanced helicopter pilotage systems, digital processing will be used to enhance imagery and add symbology. Digital processing adds a delay between when the image is captured by the sensor and when it is seen by the observer. This kind of delay is not present in currently fielded systems; the impact of this delay on flight performance is unknown. A flight test was conducted to qualitatively assess the performance impact of delaying pilotage video (14). Two aviators participated in the test and alternated as subject and safety pilot. The subject pilots wore Apache helmets and viewed a helmet-mounted camera through the Apache helmet-mounted display. The camera and display provided a 30⬚ vertical by 40⬚ horizontal, unity magnification image to the subject pilot. During the test, a cloth was draped over the subject’s visor so that all visual cues came from the helmet display. A video digitizer provided a variable delay between camera and display. All flights were in daylight and good weather. The project pilot established baselines for several, aggressive flight maneuvers using normal day, unaided vision. The maneuvers included rapid sidestep, pop-up, longitudinal acceleration and deceleration, rapid slalom, nap-of-the-earth, and contour flight. After practicing unaided and with the sensor hardware set for zero delay, the subject pilots repeated the maneuvers with the video delay increased after each iteration of the task set. Test results are based on subject and safety pilot assessments of flight performance. On the basis of the qualitative assessment of these two pilots, there appears to be no performance impact from a 33 ms image processing delay.

HETEROGENEOUS DISTRIBUTED COMPUTING

Delays of 100 ms or more impaired the subject pilot’s ability to make stable, aggressive maneuvers. All hover tasks were more difficult; sometimes a stable hover could not be achieved. Alternate strategies were developed for NOE and contour to compensate for the image processing delay. The subjects experienced the feeling that the aircraft motion was ahead of the visual scene. On the basis of this limited flight test, processing delays of up to 33 ms cannot be sensed by the pilot and appear to have no impact on flight performance. However, with an image processing delay of 100 ms, the pilot senses that aircraft movement is ahead of the displayed image. During these flights, and without prior training with delayed imagery, the 100 ms delay led to significant flight control problems.

EVALUATION Current night pilotage sensors like the ANVIS image-intensified goggle and the PNVS thermal imager provide a significant capability to fly helicopters at very low altitudes in order to hide behind hills, trees, and other terrain objects; this capability enhances the survivability of tactical helicopters on the modern battlefield. The availability of heads-up aircraft status symbology, that is, symbology superimposed on the night vision imagery, is a critical feature of these pilotage systems. Further, aviators report that their ability to perform night missions is greatly enhanced when both image-intensified and thermal imagers are available on the helicopter. Flight experiments and the results of user surveys provide guidelines for design improvements. NOE and contour flight can be accomplished with reasonable workload using a pilotage system with 40⬚ FOV and 0.6 cycles per milliradian limiting resolution; this resolution provides the pilot 20/60 visual acuity. Improving either FOV or resolution beyond these values will lessen pilot workload and lead to increased confidence. However, since the ability to resolve scene detail is important for terrain flight, night sensors should have sufficient sensitivity to provide 0.6 cycles per milliradian resolution under low thermal contrast or low scene illumination conditions. In advanced systems, this minimum level of image quality should not be traded for increased field of view.

BIBLIOGRAPHY 1. Anonymous, Lighting, Aircraft, Interior, Night Vision Imaging System (NVIS) Compatible, MIL-L-85762A, 1988. 2. D. Newman, ANVIS/PNVS Comparison Flight Test, Fort Belvoir: U.S. Army Night Vision and Electro-Optics Laboratory, 1982. 3. C. Nash, AH-64 Pilotage in Poor Weather, Fort Belvoir: U.S. Army Center for Night Vision and Electro-Optics, NV-12, 1987. 4. R. Vollmerhausen, C. Nash, and J. Gillespie, Performance of AH64 Pilotage Sensors during Reforger 87, Fort Belvoir: U.S. Army Center for Night Vision and Electro-Optics, NV-1-30, 1988. 5. T. Bui and J. Gillespie, Night Pilotage Sensor Field Assessment, Fort Belvoir: U.S. Army Center for Night Vision and Electro-Optics, NV-91-4, 1990. 6. G. Youst, J. Gillespie, and S. Adams, Desert Storm’s Night Vision and Electro-Optical Equipment Suitability Survey, Fort Belvoir: U.S. Army Night Vision and Electro-Optics Directorate, AMSELNV-0099, 1992.

679

7. D. Wood, Validation of the Night Vision Requirements for the Army Scout and Attack Helicopter Program, Fort Belvoir: U.S. Army Night Vision Laboratory, Experiment 43.7 Phase II, 1974. 8. M. Sanders, Aviator Performance Measurement during Low Altitude Rotory Wing Flight with the AN/PVS-5 Night Vision Goggles, Fort Rucker: U.S. Army Aeromedical Research Laboratory, 7610, 1975. 9. R. Hoh, Investigation of Outside Visual Cues Required for Low Speed and Hover, AIAA Paper 85-1808-CP, 1985. 10. D. Greene, Night Vision Pilotage System FOV/Resolution Tradeoff Study Flight Experiment Report, Fort Belvoir: U.S. Army Center for Night Vision and Electro-Optics, NV-1-26, 1988. 11. R. Vollmerhausen and C. Nash, Design criteria for helicopter night pilotage sensors, Proc. Amer. Helicopter Soc., 45th Annu. Forum, Boston: 1989. 12. T. Bui, R. Vollmerhausen, and B. Tsou, Overlap binocular fieldof-view flight experiment, SID Digest, XXV, 306–308, 1994. 13. L. Haworth et al., In-flight simulation of field-of-view restrictions on rotorcraft pilot’s workload, performance and visual cueing, Proc. Amer. Helicopter Soc., 52nd Annu. Forum, Washington, DC, 1996. 14. L. Biberman (ed.), Electro-Optical Imaging Systems and Modeling, Chapter 26, ONTAR Corp., North Andover, MA, In press. 15. R. Vollmerhausen, T. Bui, and B. Tsou, The affect of sensor field replication on displayed imagery, SID Digest, XXVI: 667–670, 1995. 16. G. Robinson, Dynamics of the eye and head during movement between displays: A qualitative and quantitative guide for designers, Human Factors, 21: 343–352, 1979. 17. M. Sanders, R. Simmons, and M. Hofmann, Visual Workload of the Copilot/Navigator during Terrain Flight, Fort Rucker: U.S. Army Aeromedical Research Laboratory, 78-5, 1977.

RICHARD H. VOLLMERHAUSEN U.S. Army Communications and Electronics Command

302

MISSILE CONTROL

MISSILE CONTROL A missile control system consists of those components that control the missile airframe in such a way as to automatically provide an accurate, fast, and stable response to guidance commands throughout the flight envelope while rejecting uncertainties due to changing parameters, unmodeled dynamics, and outside disturbances. In other words, a missile control system performs the same functions as a human pilot in a piloted aircraft; hence, the name autopilot is used to represent the pilotlike functions of a missile control system. Missile control and missile guidance are closely tied, and for the purposes of explanation, a somewhat artificial distinction between the two roles is now made. It must be remembered, however, that for a guided missile the boundary between guidance and control is far from sharp. This is due to the common equipment and the basic functional and operational interactions that the two systems share. The purpose of a missile guidance system is to determine the trajectory, relative to a reference frame, that the missile should follow. The control system regulates the dynamic motion of the missile; that is, the orientation of its velocity vector. In general terms, the purpose of a guidance system is to detect a target, estimate missile-target relative motion, and pass appropriate instructions to the control system in an attempt to drive the missile toward interception. The control system regulates the motion of the missile so that the maneuvers produced by the guidance system are followed, thereby making the missile hit or come as close as required to the target. The autopilot is the point at which the aerodynamics and dynamics of the airframe (or body of the missile) interact with the guidance system. Instructions received from the guidance system are translated into appropriate instructions for action by the control devices (e.g., aerodynamic control surfaces, thrust vecJ. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

MISSILE CONTROL

Guidance command

+ –

Controller

Actuator

Aerodynamic control surface

Missile dynamics

Figure 1. A block diagram describing the functional relations among the components of the missile control system.

Sensor

toring or lateral thrusters) that regulate the missile’s flightpath. A block diagram describing these missile control system operations is depicted in Fig. 1 where the function of each component is further explained as following. COMPONENTS OF MISSILE CONTROL SYSTEMS Sensor Units Sensor units measure some aspects of the missile’s motion. Gyroscopes and accelerometers are the two primary sensor units used in any missile control system. They provide the information of rotational and translational motions of a missile, respectively. 1. Gyroscope. A gyroscope is a mechanical device containing an accurately balanced rotor with its spin axis passing through the center of gravity. When the rotor rotates at a high speed, it assumes the rigidity characteristics that resist any force tending to displace the rotor from its plane of rotation. The tendency of a gyroscope to maintain its spin direction in the inertial space allows us to measure, with respect to the spin direction, the angular motion of the missile on which the gyroscope is mounted. Some recent gyroscopes, such as fiber-optical gyroscopes and ring-laser gyroscopes, do not use a spinning rotor. They calculate the body rate by use of the Sagnac effect. Fiber-optical gyroscopes have an especially high specification with reasonable cost. 2. Accelerometer. The basic principle of operation of an accelerometer consists of the measurement of the inertial reaction force of a mass to an acceleration. The inertial reaction force of the mass causes a displacement of the mass, which is suspended in an elastic mounting system within the missile, and the acceleration of the missile can be read from the displacement of the suspended mass. Velocity and position information can be obtained by integrating the accelerometer signal. One must avoid placing the accelerometer near an antinode of the principal bending mode of the missile; otherwise, the vibration pick-up at this point may result in destruction of the missile. 3. Altimeter. The altimeter, which is an instrument used to measure altitude, is another sensor unit frequently employed in cruise missile systems. There are two common types of altimeters. A pressure altimeter, which is simply a mechanical aneroid barometer, gives an approximate altitude from which a more accurate value can be calculated; on the other hand, radio altimeters give absolute altitude directly. In radio altimeters, a

303

transmitter radiates a frequency-modulated wave toward the earth, and the reflected signal is received on a separate antenna and combined with the signal taken directly from the transmitter. The frequency difference between the transmitted and the reflected signals indicates the height of the missile. Radio altimeters can be used to maintain automatically a missile at a preset altitude. Controller Units Controller units can be regarded as the ‘‘brain’’ of a missile, which tell a missile how to deflect the control surfaces or how to alter the thrust direction. The controller is in the form of preprogrammed logic and/or numerical operations installed in the on-board computer of a missile. There are two inputs to the controller units. One is from the sensor units, which provide the information about the actual motions of a missile, and the other input is from the guidance system, which provides the information about the commanded motions of a missile. The commanded motion and the actual motions are compared and manipulated in the controller units via a series of logic and/or numerical operations in order to output an intelligent decision, which renders the actual motions of a missile to match the commanded motions as closely as possible when fed into the actuator units. The series of operations involved in the controller unit is called control law. The most widely used control laws include amplification, integration, and differentiation of the error signal between the commanded motions and the actual motions. 1. Amplification. The amplification of error signal improves the robustness of the missile control system against uncertainties present in missile dynamics. 2. Integration. The integration of error signal effectively increases the closeness between the commanded motions and the actual motions. 3. Differentiation. The differentiation of error signal provides the trend of error propagation and decreases the required time for the actual motions to track the commanded motions. With the increasing computation power of on-board computers, more advanced control laws can be implemented in the missile control loop to improve the agility of a missile. This point is addressed in more detail later. Actuator Units Actuator units are energy transformation devices. They receive the command from controller units and transfer it into enough power to operate control surfaces in order to direct

304

MISSILE CONTROL

the missile to the right heading. There are three methods of operating control surfaces: (1) by a pneumatic piston, (2) by a hydraulic piston, or (3) by an electric motor. The selection of actuating power depends on factors such as the speed, size, altitude, range, and weight of the missile. 1. Pneumatic Actuator. In a pneumatic system, air from a pressure source passes through suitable delivery tubes, valves, and pressure regulators to do work upon some mechanical units such as a piston or a diaphragm, which is connected to the missile control surfaces. Unlike a hydraulic system, a pneumatic system does not reuse its transfer medium after it has performed work on the load. For that reason, the air must be stored at a pressure much higher than that necessary for actuating the load. Therefore, a pneumatic system that depends on tanks of compressed air is obviously limited in range. The performance of a pneumatic system is also limited by the property of air compressibility. Because air is compressible, the movement of a pneumatic actuator is slow because of the time it takes to compress the air in the actuator to a pressure sufficient to move it. 2. Hydraulic Actuator. The operation of a hydraulic system is similar to the pneumatic system. The most prominent difference between the two systems is that the medium of transfer in the pneumatic system is a gas, whereas the medium of the transfer in the hydraulic system is a liquid. Hydraulic fluid is practically incompressible and will produce a faster reaction on an actuator, especially when the actuator must move against large forces. This asset is evidenced by the fact that large, high-speed missiles are controlled by hydraulic actuators. The main drawback of a hydraulic actuator is its weight and the maintenance problems. A hydraulic system normally weighs more because of the need for a pump, a reservoir, filters, and an accumulator. Also a hydraulic system is hard to maintain, requiring filling and bleeding operations. 3. Electric Actuators. Generally, motors are used as the actuators in the electrical energy transfer systems. Direct current (dc) motors develop higher stall torque than alternating current (ac) motors and, therefore, are used more often for driving heavy loads encountered in highspeed missile control. An ac motor is inherently a constant-speed device that is not suitable for the requirements of a servo motor where variation in rotation speed is necessary. This factor also makes the dc motor more applicable than the ac motor as an electric actuator in missile control. The use of all-electric missile control would simplify manufacture, assembly, and maintenance. Also, it would be easier to transmit information or power to all parts of the missile by wires rather than by hydraulic or pneumatic tubing. To enforce actuating efficiency, different methods of energy transfer (e.g., electropneumatic and electrohydraulic actuators) can be combined. The preceding introduction describes the components and the operations of a missile control loop. A more detailed and fundamental introduction to the elements of missile control system can be found in Refs. 1 and 2. A missile’s heading is

changed by the action of actuators, which exert forces on control surfaces or on exhaust vanes. Altering missile heading by deflecting control surfaces is called aerodynamic control, whereas altering missile heading by deflecting exhaust vanes or by changing the jet direction is called thrust vector control. A control surface is not effective until the airflow across the surface has attained sufficient speed to develop a force. When missile speed is not high enough during the beginning of launch, the aerodynamic control is not effective, and its role is taken over by thrust vector control. The following two sections are dedicated to missile aerodynamic control and missile thrust vector control. MISSILE AERODYNAMIC CONTROL To control a missile accurately via aerodynamic forces, two general types of control surfaces (i.e., primary and secondary controls) are used. Primary control surfaces include ailerons, elevators, rudders, and canards; secondary control surfaces include tabs, spoilers, and slots. An understanding of missile aerodynamics is needed before a discussion of how these two groups of control surfaces work. Missile Aerodynamics Missile aerodynamics, like other flight vehicle aerodynamics, is basically an application of Bernoulli’s theorem, which says that if the velocity of air over a surface is increased, the pressure exerted by the air on the surface must decrease, thus keeping the total energy constant. The top surface of a missile wing section has a greater curvature than the lower surface. The difference in curvature of the upper and lower surfaces builds up the lift force. Air flowing over the top surface of the wing must reach the trailing edge of the wing in the same time as the air flowing under the wing. To do this, air passing over the top surface must move at a greater velocity than air passing below the wing because of the greater distance the air must travel via the top surface. The increased velocity means a corresponding decrease of pressure on the surface according to the Bernoulli’s theorem. Therefore, a pressure differential is created between the upper and lower surface of the wing, forcing the wing upward and giving it lift. Besides the wing, any other lifting surfaces and control surfaces of a missile exhibit exactly the same function. The three-dimensional motion of a missile can be described in the body-axis coordinate system as shown in Fig. 2. The longitudinal line through the center of the fuselage is called the roll axis (x axis), the line that is perpendicular to the x axis and parallel to the wings is called the pitch axis (y axis), and the vertical line is considered as the yaw axis (z axis). The origin of the body-axis coordinate system (x, y, z) locates at the center of gravity. The three-dimensional missile motion can be resolved into two planar motions: pitch plane motion and yaw plane motion, where pitch plane is normal to the pitch axis, and yaw plane is normal to the yaw axis. The angle, measured in the pitch plane, between the projected missile velocity and the roll axis is called the angle of attack (AOA) denoted by 움. The angle, measured in the yaw plane, between the projected missile velocity and the roll axis is called the angle of sideslip denoted by 웁. The resultant force on the wing or body can also be resolved into two components: the component in the pitch plane is called normal force, and

MISSILE CONTROL

305

Yaw rotates on vertical axis

Rudder Rudder tab

Roll rotates on longitudinal axis

Center of Gravity

Pitch rotates on lateral axis

Wing

elevator tab elevator aileron tab aileron

y, Y, V

Canard

α

λ

Iyy, M, q

Ixx, L, p

β

x, X, U

θ Izz, N, r z, Z, W Relative wind Figure 2. Schematic demonstration of the nomenclature used in missile dynamics. The locations of the primary control surfaces (rudder, elevator, aileron, and canard) and the secondary control surface (tabs) are shown. The definition of the roll, pitch, and yaw motions is also shown.

the component in the yaw plane is called side force. The normal force can be further resolved into two components: the component perpendicular to the projected missile velocity (in the pitch plane) is called lift and the component along the projected missile velocity is called drag. In many tactical missiles (e.g., short-range air-to-air missiles), the wing providing the lift force is not prepared. They keep a suitable AOA in the flight, where the lift force is produced by control fins or stability fins. Some fundamental control-related missile aerodynamics are surveyed in the following list. Readers who are interested in advanced missile aerodynamics can refer to Refs. 3 and 4 for details. 1. Lift Force. Lift force is the force by which aerodynamic control surfaces can change the attitude of a missile. Lift force depends on the contour of a wing, AOA, air density, area of the wing, and the square of the airspeed. The common equation for lift is given as L = CL

ρ AV 2 2

(1)

where L is the lift; CL is the lift coefficient, which depends on the wing contour and the AOA; ␳ is the air density; A is the area of the wing; and V is the airspeed. The lift coefficient CL is determined by wind-tunnel tests and is plotted versus AOA as a characteristic curve for the particular airfoil. As the AOA increases, the lift coefficient increases linearly to a certain maximum value, which is the point where the air no longer flows evenly over the wing surface but tends to break away. This breaking away is called the stalling angle. After the stalling angle is reached, the lifting force is rapidly lost, as is the airspeed. For a fixed AOA, the lift

coefficient CL depends on the wing span and the profile of the wing. Increasing wing span or using the leadingedge slot or trailing-edge flap to increase the camber of wing profile may effectively increase the lift coefficient. 2. Drag Force. Drag is the resistance of air to forward motion and is an adverse factor of control effectiveness. It is the force that must be overcome by the thrust. The drag force in formula is D = CD

ρ AV 2 2

(2)

where CD is the coefficient of drag obtained from characteristic curves of airfoils via wind-tunnel tests. For a small AOA, CD changes very little with the AOA. As the AOA increases, CD increases. The drag coefficient is usually quite small when compared with the lift coefficient. There are three sources of air drag. The skin friction of air on the wing is called profile drag; the air resistance of the parts of a missile that do not contribute to lift is called parasite drag; and the part of airfoil drag that contributes to lift is called induced drag. CL, CD, and other aerodynamic coefficients can be evaluated from empirical techniques, computational fluid dynamics (CFD) modeling, or by the processing of wind tunnel test data. It should be noted that various degrees of uncertainty are associated with each of these methods, with wind tunnel measurements usually being accepted as the most accurate. 3. Wingtip Vortex. The asymmetric wingtip vortex, which has a remarkable effect causing row-yaw instability at a high AOA, is always a challenge to missile control system design. As air flows about a wing, the pressure of

306

MISSILE CONTROL

the air immediately above the upper surface is less than the air pressure immediately below the surface. With the air at a higher pressure below the wing, air will spill by the wingtips to the upper surface. This flow of air from the lower surface combines with the normal flow of air, causing a swirl of air at the wingtips. This swirl is called a wingtip vortex. At each side of the wingtip, the action of the vortex is to throw the air inward and downward. Induced drag is related to the downflow caused by the wingtip vortices. 4. Downwash. Because of the camber shape of the wing airfoil, air flow over the wing is deflected downward toward the elevator. This angle of deflection is called the downwash angle. When missile tail control is considered, the downwash effect caused by the wing must be seriously taken into account because downwash can significantly reduce the effective AOA of the tail surface and reduce the elevator ability of pitch control. 5. Shock Wave Effect. Shock wave is a prominent aerodynamic phenomenon when missile speed is at the transonic or supersonic ranges. As the speed of a missile increases, there comes a point at which the air can no longer get out of the way fast enough. The air tends to pile up or compress in front of the missile, setting up what is known as shock waves. In a shock wave, the pressure of air varies sharply, seriously altering the forces and pressure distribution on a missile. When shock waves are formed on the wings or control surfaces, the air flow across the shock waves tends to separate, causing drag to rise suddenly much as in a lowspeed stall. At certain missile speeds, especially near the transonic range, the deflection of control surfaces may deteriorate the shock wave effect, which produces a peculiar vibration called flutter on control surfaces and can make control surfaces ineffective and even disintegrated. Missile Stability A stable missile can recover from the perturbed states spontaneously without control. Such stability is made possible by devices that stabilize a missile about its three axes. Accordingly, these devices are called stabilizers. The simplest stabilizer is the feathered fins at the rear of an arrow because it provides for a stable line of flight. Three types of stabilizers are required to stabilize a missile about its three axes. 1. Pitch Stabilizer. Missile stability about the lateral (pitch) axis is achieved by a horizontal surface at the tail of the missile. This horizonal surface consists of two parts: the stationary part as the pitch stabilizer and the movable part as the elevator. The degree of pitch stability can be quantitatively expressed by an index called static margin, which is the distance of the center of pressure (c.p.) to the center of gravity (c.g.). The c.p. is the point through which the combined aerodynamic forces caused by body, wings, and control surfaces are acting. If c.p. is behind the c.g. (i.e., the static margin is positive), the missile is said to be statically stable. In this case, any perturbation of the body away from the direction of the velocity vector results in a moment about the c.g. that tends to decrease this perturbation.

This of course is the reason why feathers are placed at the rear end of an arrow to move the c.p. aft. If a missile has no autopilot (i.e., no instrument feedback), a sizable static margin, say 5% or more of the overall length, has to be allowed to ensure stability. However, if the static margin is excessively positive, the missile is unnecessarily stable, and control moments will be relatively ineffective in producing a sizable maneuver. On the other hand, although a missile with negative static margin is statically unstable, it may exhibit great agility when autopilot is installed. It is worth noting that the static margin of a missile is not a fixed value, because of the c.p. variation for different flight conditions and the c.g. variation caused by propellant usage. A challenging missile control problem is to ensure the stability of the airframe for all possible c.p. and c.g. locations. 2. Yaw Stabilizer. Missile stability about the vertical (yaw) axis is usually provided for by a vertical fin. If a missile tends to turn to the left, the pressure on the right side of the fin is increased. This increased pressure resists the rotation and forces the tail in the opposite direction. In some missiles, the fin may be divided and have a movable part called the rudder that is used for directional control. Besides the fin, the vertical sides of the fuselage also act as stabilizing surfaces. Another way to increase the yaw stability is via sweepback of wings. 3. Row Stabilizer. Missile stability about the longitudinal (row) axis is achieved by a dihedral and by the positioning of the wing. A dihedral angle is the angle formed by a reference line through the wing surface and the lateral axis of the missile. Dihedral produces stability by causing a change of lift on the wing surfaces. As a missile starts to roll, it will sideslip slightly and thus create a relative wind component. This component increases the lift on the lower wing and decreases the lift on the higher wing. Hence, an opposite torque is generated to stop rowing. The positioning of the wings at the time a missile is constructed is another means of obtaining stability about the row axis. A missile has greater row stability if the wings are placed above the center of gravity than if they are placed below the center of gravity. Primary Control Surfaces Ailerons, rudders, elevators, canards, and their various combinations are considered primary controls. These control surfaces are shown schematically in Fig. 2. As these control surfaces are deflected, they present a surface to the existing air flow at an angle that will cause a force to exist. This force pushing against the control surface moves the wing or tail to which the control surface is attached in a direction opposite to the control surface movement. 1. Ailerons. A conventional aileron is attached to the outer trailing edge of the wings to control the missile row motion in a manner that when one aileron is lowered, the opposite one is raised. 2. Elevators. Elevators are attached to the pitch stabilizer on the tail to control pitch motion. They are raised and lowered together.

MISSILE CONTROL

3. Rudders. A rudder is attached to the rear part of the vertical stabilizer and is used to maintain directional (yaw) control. 4. Canards. A canard is basically a forward wing located ahead of the center of gravity of the missile for the purposes of stabilization and pitch control. One type of canard structure consists of a fixed stabilizing plane with a surface control attached to the trailing edge. Another type of canard structure uses a pivoted mechanism that allows the entire stabilizing plane to rotate up or down. 5. Dual-Purpose Control Surfaces. The preceding control surfaces can be properly combined to give multipurpose control functions. Feasible combinations include elevons, ailevators, and rudder-vators. As the names indicate, they consist of control surfaces that accomplish two purposes. For instance, an elevon takes the place of an elevator and an aileron, giving control of pitch and roll. 6. Variable-Incidence Control Surfaces. This type of control rotates the position of an entire wing rather than just part of it. The variable incidence control can overcome the problem of flutter and the need for structural strength of control surfaces and yet have a control that is sensitive and effective at various speed ranges. The variable incidence control can be used on the wing, horizonal stabilizer, or vertical stabilizer. Secondary Control Surfaces Primary control surfaces can be looked upon as the main controlling factor of the missile’s path; however, by using secondary control surfaces, a missile can be controlled much more accurately and efficiently. A secondary group of aerodynamic control surfaces is composed of tabs, slots, and spoilers, which are schematically demonstrated in Fig. 2. For the convenience of compact illustration, all six primary control surfaces and the three secondary control surfaces are put together in one missile, as shown in Fig. 2; however, a missile may not be equipped with all types of primary and secondary control surfaces. For example, missiles in general do not have both tail and canard controls, and conventional missiles do not have secondary control surfaces, which are almost exclusively used in large cruise missiles. 1. Tabs. Tabs are small pieces of movable or fixed metal attached to the trailing edge of the primary control surfaces. They help to trim the missile or to alleviate the loading of the primary control surfaces, but they do not in themselves determine the direction of missile motion. Tabs can be divided into three types: fixed, trim, and booster. A fixed tab can be bent uniformly in the required direction to trim the missile. A trim tab is movable and controllable, and is used to trim the missile with varying attitude, speed, or altitude. A booster tab, sometimes known as a servo tab, is used to assist in moving primary control surfaces with large area. 2. Slots. A slot is a high-lift device located along the leading edge of the wing. The slot is ineffective in the region of a normal AOA, but when a missile reaches a high AOA, the slot can be opened to allow air to spill through and hence delay the formation of turbulence flow over the top surface of the wing.

307

3. Spoilers. As the name indicates, a spoiler is used to generate turbulence flow and ‘‘spoil’’ the lift on a wing. When not used, spoilers are recessed into the upper camber of the wings and allow the flow of air over the wing to be smooth and uninterrupted. If, however, a gust of wind has caused the right wing to drop, the control system instantly calls for the spoiler on the left wing to extend. As the spoiler extends, the lift on the left wing is spoiled and reduced a considerable amount. The wings then tend to return to the original position. MISSILE THRUST VECTOR CONTROL A completely different method of steering a missile is to alter the direction of the efflux from the propulsion motor. This method is known as thrust vector control (TVC). TVC is clearly not dependent on the dynamic pressure of the atmosphere and is generally used in the phase of flight where missile speed is so low that the airfoil sections do not have enough aerodynamic stabilizing effect. On the other hand, TVC is inoperative after propulsion motor burn-out, but at this time aerodynamic forces become large enough to take over the role of TVC. There are several methods of directing the thrust of a rocket motor, and each has advantages and disadvantages, which may or may not recommend it for a particular application. References 1 and 5 provide more information on TVC. 1. Exhaust Vanes. Exhaust vanes are surfaces that are installed directly in the exhaust path of the jet engine. When the position of the vane is changed, it deflects the exhaust and causes the thrust to be directed in opposition to the exhaust vane. The operation of exhaust vanes is sketched in the middle part of Fig. 3. Because of the severe erosion problem caused by the tremendous heat in the exhaust, the life of exhaust vanes is generally short. Graphite and more recently tungsten and molybdenum have been used as the materials of the exhaust vanes. To reduce the complexity of the actuator design, the actuating mechanism of an exhaust vane often shares with that of aerodynamic control surfaces; therefore, when control surfaces move in the ambient air path, an exhaust vane moves, simultaneously and with the exact same manner, within the exhaust path of the jet engine. The device ‘‘jetavators’’ is the outcome of such a design idea, which can control jet and elevator simultaneously. Perhaps the oldest TVC is the exhaust vane used in the German V2 in World War II. Many surface-to-surface missiles, including the American Pershing, have used exhaust vanes to control the jet direction. 2. Gimbaled Engine. By mounting the combustion chambers in gimbals and controlling its position by servos, the direction of thrust can be altered. The operation of gimbaled engine is sketched in the lower part of Fig. 3. Two serious objections to this method are that all the various fuel lines must be made flexible, and the servo system that actuates the jet must be extremely strong. However, gimbaled liquid-propellant engines have been used successfully for many years. For example, the Viking research vehicles have been successfully flown many

308

MISSILE CONTROL Jet control

Auxilary directional thrust

Deflection charges

Deflects jet stream

Jet vans

deflection of up to 12⬚ has been obtained by injecting hot gas bled directly from the combustion chamber. 5. Reaction Control Thruster. An easier system of jet control is accomplished by placing several small thrusters at various points about the missile body. Control is accomplished by using one or another of these jets as desired, thus giving different directions of thrust. The operation of reaction control thruster is sketched in the upper part of Fig. 3. This method eliminates the use of the outside control surfaces, affording a cleaner missile surface. When reaction control thrusters are used, there will be an interaction of the jet plume with the free stream flow. This jet interaction is very nonlinear with the AOA and dominates the effective moment produced by the reaction thrusters. The produced moment may be larger or smaller than the jet thrust force times its moment arm, depending on the height by which the jet penetrates into the free stream. Reference 6 discusses missile attitude control using reaction control thruster. 6. Jet-Driving Control Surfaces. This method employs jet or air injection over aerodynamic surfaces for actuating augmentation. MISSILE CONTROL CONFIGURATION

Gimbaled engine

Changes direction of thrust

Figure 3. Three thrust vector control methods. The upper part sketches the operation of reaction control thruster; the middle part sketches the operation of exhaust vane; and the lower part sketches the operation of gimbaled engine.

According to the aforementioned various missile control methodologies, we can now give a classification of missile configuration with respect to the location of controls. If the controls are located well behind the center of gravity of the missile, the term tail control applies. If the controls are placed forward of the center of gravity, the term canard control applies. When the control is mounted on the main lifting surface near the center of gravity, the term wing control applies. What type of control surface to be used depends on the type of missile configuration in question. Regarding missile configuration, Refs. 1, 5, and 7 serve as good references. Wing-Control Configuration

times using this type of control during phases of flight wherein aerodynamic control is inadequate. 3. Moving Nozzles. Instead of moving the entire combustion chamber, we can also alter the direction of thrust by changing the orientation of the nozzle. This can be accomplished by using a flexible nozzle or a ball-andsocket nozzle. A flexible nozzle is formed by attaching the nozzle to the motor case by means of a flexible rubber mounting that is very stiff axially but relatively compliant in the pitch and yaw planes. Thrust deflection of 4⬚ to 5⬚ is feasible by this method, but a large resistance to movement is encountered when an increasingly larger deflection angle is required. Another way of attaching the nozzle to the propulsion motor is via a ball-and-socket joint with some form of low-friction seal. Although there will be some coulomb friction in this type of connection, the actuation torque will not increase with the deflection angle. 4. Injection Method. By injecting a liquid or gas into the motor venturi, we can obtain a sideways component of resultant thrust. The maximum jet deflection by using inert liquid as the injection fluid was found to be 4⬚. Jet

A wing-control configuration consists of a relatively large allmoving wing located close to the center of gravity of the missile and a set of tail or stabilizing surfaces at the aft end of missile. This all-moving wing serves as an aforementioned variable-incidence control surface. This type of control is used mostly in an air-to-air missile because of its extremely fast response characteristics. If the right and left moving wings are controlled by separate servos, they can be used as ailerons and elevators; the word elevons as mentioned earlier is applied to such a dual-purpose control surface. There are two main advantages in using wing-control configuration: • Air Inlet Consideration. Instantaneous lift can be developed as a result of wing deflection via a pivoted mechanism with little increase of missile AOA. This low value of AOA is advantageous particularly from the standpoints of inlet design for air-breathing power-plant and guidance-seeker design. For example, if the propulsion system is a ram jet, the air inlet is likely to choke if the body AOA is large, say 15⬚ or more. The use of wing control can greatly reduce the chance of inlet choke and maintain the engine efficiency by keeping the body AOA

MISSILE CONTROL

to a minimum. This point will be further addressed in the later sections. • Servo Location Consideration. The servos used in wingcontrol configuration are located near the center of the missile body. There are some occasions when the servos are most conveniently placed near the center of the missile. For example, if a medium-range missile has two separate motors, a boost motor and a sustain motor, the former may occupy the whole of the rear end of the missile and the sustainer motor may occupy most of the remaining rear half of the body. In such a case, there is just no room to install servos at the rear. If the missile carries a homing head, the servos cannot be placed at the front either. However, there are some distinct penalties involved in the use of wing control. • Pitch control effectiveness from the wings is generally very low as a result of short pitching moment arm because the lift developed is located close to the center of gravity of the missile. • Large aerodynamic hinge moments are required because of the large wing area. • Relatively large loss will be induced in tail effectiveness as a result of downwash. • Nonlinear aerodynamics is resulted from downwash caused by both wing deflection and AOA. • Severe adverse rolling moments is induced on the tail surfaces from combined effects of AOA and wing deflection. Canard-Control Configuration A canard-control configuration consists of a set of small control surfaces called canards located well forward on the body and a set of large surfaces (wing or tail) attached to the middle or aft section of the missile. Its advantages and disadvantages follow. Advantages. Advantages of canards include the following: • Canards, because of their small size, do not generate a significant amount of downwash to affect the longitudinal stability adversely. Thus relatively large static-stability margins can easily be obtained by simple changes in wing location. • Canard configuration has the inherent simplicity of packaging because the control system is small. Disadvantages. Disadvantages include the following: • Roll stabilization is difficult when the canard surface is used because of their size and downwash effect on the wings. Usually a separate set of lateral controls such as wing-tip ailerons is needed for canard configuration. • Relative high control-surface rates are required to obtain the desired rate of response because AOA must be generated before any lift is developed.

309

Tail Control Configuration Many missiles employ tail control for its convenient packaging. Usually it is desirable to have the propulsion system placed centrally in the missile so that the center of gravity movement caused by propellant usage is minimized. It is convenient and sometimes essential to have the warhead and fuse at the front together with any associated electronics including the guidance receiver. This leaves the control system to occupy the rear end with the propulsion blast pipe passing through its center. Advantages. Advantages of tail control include the following: • The tail loads and hinge moments can be kept relatively low as the total AOA on the tail is reduced. • The wing-tail interference effects are reduced because the forward main lifting surface is fixed (i.e., no downwash caused by wing deflection). Therefore, the aerodynamic characteristics are more linear than those for wing-control design. Disadvantages. Disadvantages include the following: • With this type of control, it is obvious that the tail deflection must be opposite in direction to the AOA. This feature results in relatively slow response characteristics because the initial lift is in a direction opposite to the desired one. • Deficiency of tail surfaces to provide the desired lateral control. Wing Arrangements Wing arrangements have a significant influence on the types of missile control to be used. Three types of wing arrangements are discussed here. 1. Cruciform. The most commonly used configuration in missile design is the cruciform, which possesses four wing surfaces and four tail surfaces. There are several major advantages in the use of this type of configuration: (i) fast response in producing lift in any direction, (ii) identical pitch and yaw characteristics, and (iii) simpler control system as the result of item (ii). One of the most important aspects associated with a cruciform design is the orientation of the tail surface with respect to the wing planes. The significant conclusion from considerable experience and experimental data was that an in-line tail surface (i.e., all the four tail surfaces are in the same orientations as the four wing surfaces) provides the best overall aerodynamic characteristics for most missile applications. The other possible wing-tail geometrical relation is called interdigitated configuration where there is a 45⬚ separation between the wing and tail orientation. For a cruciform missile, the most difficult parameter to determine accurately is the induced rolling moment. The rolling moments arise whenever the missile simultaneously executes pitch and yaw maneuvers that are unequal in magnitude. Such maneuvers result in unequal or asymmetric flow patterns

310

MISSILE CONTROL

over the aerodynamic lifting surface; consequently, rolling moments are induced on the airframe. Hence, roll stabilization or control is a critical issue for cruciform missiles. 2. Monowing. The monowing arrangements are generally used on cruise-type missile (i.e., missiles design to cruise for relatively a long range like crewed aircraft). This type of design is generally lighter and has less drag than the cruciform configuration. The wing area and span are, however, somewhat larger. Although the monowing missile must bank to orient its lift vector in the desired direction during maneuvering flights, the response time may be sufficiently fast and acceptable from a guidance-accuracy standpoint. The induced-roll problem for the monowing configuration is substantially less severe than that associated with the cruciform configuration. A separate set of lateral control surfaces, such as flaps, spoilers, and wing-tip ailerons, is generally used in a monowing design. This stems from the fact that the canard or tail surfaces that are usually employed for pitch control on monowing design are generally inadequate for lateral control. 3. Triform. This type of wing arrangement, which employs three wings of equal area spaced 120⬚ apart, is seldom used because no noticeable advantage can be realized. Results of a brief preliminary analysis indicate that the total wing area of the triform is equal to that used on a cruciform arrangement and that consequently no noticeable change in drag may be realized. In addition, little or no weight saving will be realized, even though one less arrangement or fitting is required because the total load remains the same. MISSILE CONTROL STRATEGY Because the missile control system (autopilot) is commanded by the missile guidance system, the autopilot command structure is dependent on guidance requirements for various mission phases. • Separation (Launch) Phase. A body rate command system is typically used during launch because of its robustness to the uncertain aerodynamics. • Agile Turn. During an agile turn, directional control of the missile’s velocity vector relative to the missile body is desired. This amounts to commanding AOA or sideslip, and regulating roll to zero. • Midcourse and Terminal Phases. An acceleration command autopilot is commonly employed in these two phases. • End of Homing Phase. At the end of terminal homing, the missile attitude may be commanded to improve the lethality of the warhead. Among these four autopilot structures, separation, midcourse, and endgame autopilots are in general well understood and have been implemented in production missiles. Autopilot designs for agile turns are significantly less well understood. Reference 8 gives a detailed discussion of the challenges involved in agile turn, and several solution techniques were provided there.

Up to now, the existing missile control strategies in various mission phases include two major categories: skid-to-turn (STT) strategy and bank-to-turn (BTT) strategy. It is interesting to note that the progress in control strategy for crewed aircraft is from BTT to direct sideslip control (i.e., STT), whereas the progress in missile control strategy is from STT to BTT. The applications and limitations of STT and BTT will be introduced in the following sections. Skid-to-Turn Strategy In STT the missile roll angle may be either held constant or uncontrolled; in either case, the magnitude and orientation of the body acceleration vector is achieved by permitting the missile to develop both an AOA and a sideslip angle. The presence of the sideslip imparts a ‘‘skidding’’ motion to the missile; hence the name skid-to-turn. The STT missile autopilot receives the guidance command interpreted in terms of the Cartesian system. In the Cartesian system, the missileguidance system produces two signals, a left–right signal and an up–down signal, which are transmitted to the missile-control system by a wire or radio link to rudder servos and elevator servos, respectively. If a cruciform missile adopts STT control strategy, the two servo channels can be made identical because of the identical pitch and yaw characteristics of a cruciform missile as mentioned earlier. Hence, in STT missiles, both pitch control and yaw control are called lateral control, which is different from the definition of aircraft control. The other control loop of the STT missile is roll control, which is used to stabilize the missile roll position. For a perfect performance of the STT missile, it is assumed that the missile will remain in the same roll orientation as at launch during the whole flight. In this ideal case, up–down signals, if sent to the elevator servos, should then result in a vertical maneuver only; and left–right signals, if sent to the rudder servos, should result in a horizontal maneuver only. However, a missile, except for a monowing missile, is not designed like an airplane and there is no tendency to remain in the same roll orientation. In fact, it will tend to roll for many reasons such as accidental rigging errors, asymmetrical aerodynamic loadings, and atmospheric disturbances. Two methods ensure that left–right commands are performed by rudder servos and up–down commands are performed by elevators. The first method applies a quick roll servo (with bandwidth larger than that of lateral servos) to stabilize the roll dynamics and to recover the missile to the original roll orientation. The second method allows the missile to roll freely but installs a roll gyro and resolver in the missile to ensure that the commands are mixed in the correct proportions to the elevators and rudders. However, roll stabilization (the first method) is generally more preferred for the following reasons: • There are many occasions when roll position control is necessary, for example, to ensure that the warhead or altimeter always points downward. • If the missile is free to roll, high roll rates may cause cross-coupling between the pitch and yaw channels and tend to unstabilize the system. An STT missile with properly controlled roll motion may provide the following advantages: • Same degree of vertical and horizontal maneuverability can be achieved.

MISSILE CONTROL

• With STT control it is possible to resolve three-dimensional target and missile motion into two independent planar motions and to consider the pitch and yaw channels as an independent two-dimensional problem. Hence, both guidance law and control system design can be done via two-dimensional analysis. This simplification makes it possible to apply the classic control theory, which treats single-input single-out (SISO) system to the missile autopilot design. Bank-to-Turn Strategy The concept of BTT stems from the motion of crewed aircrafts, which use ailerons to bank (roll) to the left or right. During a left or right turn, a small amount of rudder is also applied in an attempt to make the air flow directly along the longitudinal axis of the aircraft. Hence, in BTT motion, there is no sideslip and no net side force. From a passenger’s point of view, this method of maneuvering is the most comfortable because the total force experienced is always symmetrically through the seat. When BTT concept is applied to missile control, the missile is rolled first so that the plane of maximum aerodynamic normal force is oriented to the desired direction and the magnitude of the normal force is then controlled by adjusting the pitch attitude (AOA). If we consider the guidance command for an STT missile as being expressed in the Cartesian coordinates (x, y) where x is the right–left command and y is the up–down command, then the guidance command for a BTT missile can be considered as being expressed in the polar coordinates (r, ␾) where ␾ is the angle to roll and r is the distance to be steered in the pitch plane. Therefore,BTT strategy is sometimes called polar control or ‘‘twist-and-steer’’ control. Although BTT control has been used in crewed aircraft for a long time, the interest in BTT missile control only began in the late 1970s. The principle motivation for developing the BTT missile autopilot stems from the successful application of ramjet propulsion technology to missile system. Several ramjet missiles were developed in the late 1970s, including ramjet interlab air-to-air technology (RIAAT program, Hughes), advanced common intercept missile demonstration (ACIMD program, Naval Weapons Center), advanced strategic air-launched multi-mission missile (ASALM program, McDonnell Douglas and Martin-Marietta). These BTT programs are thoroughly surveyed in Ref. 9. All these ramjet missile programs require autopilot to prevent missile maneuvers from shading the inlet (i.e., the AOA needs to be small and positive) and to limit sideslip 웁 in order to increase engine efficiency and thereby maximize range. The conventional STT strategy cannot satisfy these limitations on 움 and 웁. The applicability of the ramjet missile requires investigation in the following areas: 1. Monowing Configuration. Ramjet missiles have two inlets external to the main body and there is room for only one pair of wings (i.e., monowing). 2. Variable-Incidence Wing Control. Because the inlets could accept only a small AOA as a result of interference from the body, the use of variable-incidence wing control, which can provide instantaneous lift without increasing the AOA of the body, is very suitable for ramjet engines.

311

3. BTT Autopilot Design. If a ramjet missile has two fixed wings and is controlled with four cruciform tails, the best solution is to adopt a BTT autopilot, which can ensure small values of AOA and sideslip angle. Only the technique in item 3 is discussed here. The design of a highly maneuverable BTT autopilot poses a severe challenge to the control designer. High maneuverability means not only high aerodynamic acceleration but also the ability to change the orientation of the acceleration rapidly. This means that the roll rate can be expected to be much larger (perhaps by an order of magnitude) than they would be in a STT missile. The large roll rates induce substantial cross-coupling between the pitch and the yaw axes, whereas in a typical STT missile this cross-coupling is negligible. The main advantage of BTT strategy is its adaptability to ramjet missile control, but there are many difficulties that cannot be conquered by the techniques used in STT strategy: • The cross-coupling between the pitch and yaw axes requires the designer to consider both axes together as a single multi-input/multi-output (MIMO) system. The classic SISO control approach becomes inadequate for BTT application, and modern MIMO control theory needs to be considered. • The cross-axes couplings are proportional to the roll rate, which is a dynamic variable. This means that the dynamics of the pitch and yaw axes are not only cross-coupled but also nonlinear. Therefore, a single fixed-coefficient linear autopilot may be unable to cover the whole flight envelope, and linear autopilot with gain scheduling or nonlinear autopilot design should be taken into account. • The three-dimensional motion of a BTT missile cannot be resolved into two planar motions. Hence, the guidance law design for a BTT missile needs detailed three-dimensional analysis. In summary, a BTT missile can be considered as a MIMO system with nonlinear dynamics and with three-dimensional kinematics, whereas a STT missile can be well approximated as an integration of three SISO systems with linear dynamics and with two-dimensional kinematics. Reference 8 summarizes some status and concerns of BTT missiles. How modern control theory can be used to design BTT autopilots is discussed in Ref. 10. MISSILE AUTOPILOT DESIGN Equations of Motion The equations of motion of a missile with controls fixed may be derived from the Newton’s second law of motion, which states that the rate of change of linear momentum of a body is proportional to the summation of forces applied to the body and that the rate of change of the angular momentum is proportional to the summation of moments applied to the body. Mathematically, this law of motion may be written as         mU Hx L X d  d        (3)  mV  = Y  , Hy  = M dt dt Hz mW N Z

312

MISSILE CONTROL

where (X, Y, Z) and (L, M, N) are the resultant forces and moments caused by aerodynamic forces, gravity, and propulsive forces, along the body axes (x, y, z). (U, V, W) and (Hx, Hy, Hz) are the components of the velocity and angular momentum of the missile about the x, y, and z axes, respectively. The two main reasons for the use of body axes in the dynamic analysis of the missile are (1) the velocity along these axes are identical to those measured by instruments mounted in the missile and (2) the moments of inertia (i.e., Ixx, Ixy, etc.) are independent of time. Equation (3) and (4) can be expressed in terms of the moments of inertia and the missile angular velocity P, Q, and R as follows:



0   R −Q

−R 0 P

    U˙ + QW − RV X    ˙ m V + RU − PW  = Y  ˙ + PV − QU Z W    ˙ P Ixx −Ixy −Ixz   ˙  −I I −I Q  xy yy yz    + −Ixz −Iyz Izz R˙      P Q Ixx −Ixy −Ixz L      Iyy −Iyz  Q = M −P  −Ixy R 0 −Ixz −Iyz Izz N

(4a)

of inertia about the y axis is generally equal to that about the z axis (i.e., Iyy 앒 Izz). Hence, the resulting equations become

m(U˙ + QW − RV ) = X

(6a)

m(V˙ + RU − PW ) = Y ˙ + PV − QU ) = Z m(W ˙ xx = L PI

(6b)

˙ yy + PR(Ixx − Izz ) = M QI ˙ zz + PQ(Iyy − Ixx ) = N RI

(6c) (6d) (6e) (6f)

These are the general equations used in the analysis of STT control strategy, especially for agile STT missiles with substantial induced roll. When rolling rate P is relatively small when compared with Q and R, further simplification of Eq. (6) is possible by dropping the terms relating to P, and the result is the three decoupled servo channels used in the conventional STT autopilots. 1. Pitch dynamics:

(4b)

˙ − QU ) = Z, m(W 0

Iyy Q˙ = M

(7a)

Izz R˙ = N

(7b)

2. Yaw dynamics: For a missile with monowing configuration, the xz plane is a plane of symmetry. Consequently, Iyz ⫽ Ixy ⫽ 0 from the definition of moment of inertia. Hence, Eqs. (4) may be simplified as follows:

m(U˙ + QW − RV ) = X

(5a)

m(V˙ + RU − PW ) = Y ˙ + PV − QU ) = Z m(W ˙ xx + QR(Izz − Iyy ) − Ixz (R˙ + PQ) = L PI ˙ yy + PR(Ixx − Izz ) + Ixz (P2 − R2 ) = M QI

(5b)

˙ zz + PQ(Iyy − Ixx ) − Ixz (P˙ − QR) = N RI

(5c) (5d) (5e) (5f)

These differential equations govern the motion of a monowing missile with BTT control. It can be seen that these equations are nonlinear and cross-coupled; none of the equations can be isolated from the others. Taking Eq. (5b) as an example, the term ⫺mPW says that there is a force in the y direction caused by the incidence in pitch (i.e., 움 ⫽ W/U) and the roll motion P. In other words, the pitching motion (W) of the missile is coupled to the yawing motion (Y force) on account of roll rate P. Equation (5a) does not really concern us because, in most cases, we are interested in the acceleration normal to the velocity vector as this will result in a change in the velocity direction. In any case, in order to determine the change in the forward speed U, we need to know the magnitude of the propulsive and drag force. Nevertheless, except for power phase, the variation of U is generally very small. For a missile with cruciform configuration, further simplifications can be made because (1) the xy plane (as well as xz) is also a plane of symmetry (i.e., Ixz ⫽ 0) and (2) the moment

m(V˙ + RU0 ) = Y, 3. Roll dynamics: Ixx P˙ = L

(7c)

where the forward speed U is assumed to be a constant U0 ˙ is generally small. It can be observed that each because U servo channel is decoupled, linear, and SISO (i.e., each channel has a single input and a single output: the pitch dynamics with elevator input and AOA [움(t) ⫽ W(t)/U0] output, the yaw dynamics with rudder input and sideslip [웁(t) ⫽ V(t)/U0] output, and the roll dynamics with aileron input and roll rate P output). This formulation is rather simplified, but very promising results had been recognized in an STT autopilot application. In general, the resultant forces X, Y, Z, and moments L, M, N in Eq. (6) are nonlinear functions of U, V, W, P, Q, R, and of the control surface deflections. However, a linear control system is designed under the condition that the missile is exercised through small perturbations about some trim conditions (equilibrium conditions). From the viewpoint of autopilot design, a linear Taylor expansion of the resultant forces and moments about the trim conditions is adequate. We will use the symbol with subscript zero (0) to stand for trim condition and the symbol with a lowercase letter to denote the perturbation quantities. For example, V is expressed by V(t) ⫽ V0 ⫹ v(t) where V0 is the steady-state side speed and v is the perturbed side speed which is a function of time. The other variables can be expressed in the same way. The deflection angles of aileron, elevator, and rudder, will be denoted by 웃a, 웃e, and 웃r, respectively.

MISSILE CONTROL

Forces and moments can also be expanded in a perturbed form. For example, assume that the side force Y(V, R, 웃r) is a function of V, R, and 웃r. It can be expanded as

∂Y ∂Y ∂Y v+ r+ δr ∂v ∂r ∂δr = Y0 + yv v + yr r + yδ r δr

Y (V, R, δr ) = Y (V0 , R0 , δr 0 ) +

U0 + zq mq

2. Yaw dynamics: v˙ yv = r˙ nv

−U0 + yr nr

(8)

zδ e w + δe mδ e q

(9)

v y + δ r δr nδ r r

(10)

3. Roll dynamics: p˙ = lp p + lδ a δa

(11)

The Laplace transfer function from the aileron input 웃a to the roll rate output can be found from Eq. (11) as −lδ a /lp p = δa Ta s + 1

(12)

where ⫺l웃a /lp can be regarded as the steady state gain and Ta ⫽ ⫺1/lp can be regarded as the time constant of the roll channel. The Laplace transfer function from the rudder input 웃r to the body yaw rate r can be obtained from Eq. (10) as n δ r s − n δ r + n v yδ r r = 2 δr s − ( yv + nr )s + yv nr + U0 nv

follow-up units as a complete missile control system as described at the beginning of this article. Classic Control Design

where Y0 is the steady-state side force; yv ⫽ (⭸Y/⭸v (V0, R0, 웃r0), yr ⫽ (⭸Y/⭸r) (V0, R0, 웃r0), y웃r ⫽ (⭸Y/⭸웃r) (V0, R0, 웃r0) are called aerodynamic derivatives evaluated at the specified trim condition. Aerodynamic derivatives with respect to state variables are also called stability coefficients such as yv, and yr; derivatives with respect to control surface deflection are also called control coefficients such as y웃r. Remaining forces and moments can be linearized in a similar way as in Eq. (8). Substituting these linearized quantities into Eq. (7) yields the control equations for a STT missile as 1. Pitch dynamics: w˙ zw = q˙ mw

313

(13)

Figure 4 depicts the block diagram of a lateral autopilot performing side force control, where a rate gyro measuring yaw rate and an accelerometer measuring side acceleration are used as feedback sensors. The missile’s aerodynamic transfer function in Fig. 4 are obtained from Eq. (10). The controller is in the form of proportion and integration (PI). The problem of autopilot design is to design properly the seven parameters KP, KI, Ka, Kg, ␰s, 웆s, and Ks such that the actual missile side force y follows the commanded side force yd as quickly as possible. Among the seven parameters, the two controller gains KP and KI can be further tuned to satisfy different flight conditions. The remaining five parameters have fixed values and cannot be tuned on line. The selection of the seven parameters is aided by such tools as root locus, Bode, Nyquist, or Nicholls plots that enable visualization of how the system dynamics are being modified. The performance specifications of the side force response may be given in the frequency domain (e.g., bandwidth and gain/phase margins) or in the time domain (e.g., overshoot, damping ratio, rise time, and settling time). The classic control design process of missile autopilot can be summarized in the following steps. Detailed procedures and practical design examples can be found in Refs. 5 and 11. How aerodynamic derivatives affect the missile autopilot design is discussed in Ref. 12. A useful review of classically designed autopilot controllers may be found in Ref. 13, where the relative merits of proportional and PI autopilot controllers are discussed and the novel cubic autopilot design is introduced. 1. Based on the system requirements analysis, the designer selects a flight control system time constant, a damping ratio, and an open loop cross-over frequency that will meet the system requirements for homing accuracy and stability. 2. The autopilot gains are calculated. The gains such as KP and KI in Fig. 4 are obtained in a variety of linearized flight conditions and must be scheduled by appropriate algorithms to account for the changing environment.

(14)

3. A model of the flight control system is developed. Initially the flexible body dynamics are neglected and the rigid body stability is analyzed to determine if adequate phase and gain margins have been achieved. If not, the response characteristics are modified and the design is iterated.

It can be seen that the characteristics of the open-loop responses in Eqs. (12) and (14) are determined by the related aerodynamic derivatives. For example, to ensure that the open-loop yawing motion (i.e., without control) is stable, we must have yv ⫹ nr ⬍ 0. If the open-loop motion is unstable or is near the margin of instability, then autopilot must be installed to form a closed-loop system that integrates missile dynamics, sensor units, controller units, actuator units, and

4. When the low-frequency design is complete, the flexible body dynamics are incorporated into the frequency models, and the stability is reexamined. For typical tactical homing missiles, the flexible body model should include the first, second, and third resonant mode dynamics of the pith and yaw channels and at least the first mode of the roll channel. Depending upon the characteristics of the roll airframe structure, additional modes may have to be modeled.

Let ␰ and 웆n be the damping ratio and the undamped natural frequency of the yaw channel, respectively, then we have 2ξ ωn = −( yn + nr ),

ωn2 = yn nr + U0 nv

314

MISSILE CONTROL

Rudder angle Side force command + yd –

Actuator

Controller Kp +

–Ks

KI

s2

s

ϖ s2

Rate gyro + +

Missile dynamics

Kg

yδ rs2 – yδrnr s – U0(n δr yv – nv yδr)

+ 2ξ s

s

ϖs

+1

ξr

Yaw rate r

Side force response

s2 – (yv + nr)s + yvnr + U nv

y

0

nδ rs + nv yδ – nδ r yv yδ rs2 – yδ rnr s – U0(nδ r yv – nv yδ r

Is Accelerometer + Kα

Lateral acceleration

+

Figure 4. An autopilot structure performing side force command tracking. Both missile and rudder servos are modeled as second-order dynamics; the gyro and accelerometer are modeled as constant gains; and the controller is in the form of proportion and integration with tuning gains KP and KI.

5. In cases where the stability margins do not meet the design criteria the autopilot design is modified through adjustment of the autopilot gains and/or the inclusion of structural filters that adjust the gain or phase in the area of the natural resonances. Modern Control Design Classic control techniques have dominated missile autopilot design over the past decades. Autopilot design for future missile systems will be dominated by the requirement of ultimate agility in the entire flight envelope of the missile. Critical issues in the next generation autopilot will include (1) fast response to the commanded accelerations, (2) high maneuverability and guaranteed robustness over a wide range of mission profiles at all speeds and altitudes, (3) performance robustness against uncertainties in the aerodynamic derivatives, in the thrust profile, in the effectiveness of the control surfaces, and in the varying mass and moment of inertia, (4) cancellation or attenuation of highly nonlinear and coupled missile dynamics as a result of high AOA. The development of eigenstructure assignment, linear quadratic regulator (LQR) control, robust control, nonlinear control, adaptive control, and intelligent control techniques have revolutionized missile control system design considerably. They provide power tools to realize the aforementioned critical issues. Reference 14 provides an excellent discussion of various applications of modern control theory to flight control systems. Eigenstructure-Assignment Autopilot Design. Eigenstructure assignment is the multivariable extension of the root locus method. The behavior of a MIMO system is characterized by eigenvalues and eigenvectors. The eigenvalues determine stability, and the eigenvectors characterize the shape and coupling of different modes. The technique is concerned with the placing of eigenvalues and their associated eigenvectors by feedback, to satisfy directly closed loop damping, settling time, and decoupling specifications. A review of eigenstructure assignment for aerospace applications can be found in

Ref. 16. The technique has been applied to the control of the extended medium-range air-to-air missile in Ref. 17. LQR Autopilot Design. LQR control theory is a well-established control system design technique (18). The LQR control gains are all obtained simultaneously from the minimization of a suitable performance index (usually the integral of a quadratic cost function). The design is synthesized in the time domain as opposed to the complex frequency domain. Reference 14 demonstrates the effectiveness of LQR design techniques for the missile flight control problem—describing the application of various LQR formulations to the design of single-plane lateral acceleration autopilot controllers. Reference 19 further considers the advantages obtainable by combining classical PI and modern LQR methodologies for a multivariable airframe model with high frequency structural modes. Robust Autopilot Design. Robust control methods provide the means to design multivariable autopilots that satisfy performance specifications and simultaneously guarantee stability when the missile deviates from its nominal flight condition or is subject to exogenous disturbance. Several investigations have been undertaken specifically to research missile autopilot robustness. Early work was directed toward specific configurations and problems (20), with more recent work using the robust control system synthesis techniques of quantitative feedback theory (QFT) (21), H앝 control (22), 애-synthesis (23), normalized coprime factor loop-shaping H앝 control (24), and linear matrix inequality (LMI) self-scheduling control (25). Research has also been carried out on a number of related ways of assessing the robustness of missile autopilot controller design (26). A good literature survey in robust autopilot design can be found in Ref. 15. The robust control design is formulated to minimize the following effects: • Parameter Variation. Aerodynamic derivatives, moment of inertia, and the center of gravity may have significant variations over the entire missile flight envelope.

MISSILE CONTROL

• Coupling Dynamics. The residual error caused by inexact cancellation in decoupling pitch and roll–yaw dynamics for BTT missiles needs to be addressed. • Unmodeled Dynamics. Most missile autopilot design consider missile rigid-body dynamics only, and the missile flexible modes are regarded as unmodeled dynamics. Robust control design allows the unmodeled dynamics to be taken into account to avoid structural vibration or instability. • Sensor Noises. Autopilot needs to attenuate the effects caused by sensor noises, calibration errors, drifts, and parasitic dynamics. • Tracking Error. A successful missile interception depends on the ability of autopilot to track the guidance commands. The uncertainties and noises in the seeker output and in the prediction of target maneuvers may affect the autopilot tracking performance. Nonlinear Autopilot Design. Nonlinear control techniques used in missile autopilot design include feedback linearization (27), variable structure control (VSC) with a sliding mode (28), and nonlinear H앝 control (29). The motivations of nonlinear autopilot design come from the concerns of the three common kinds of missile nonlinearities: dynamic couplings, nonlinear aerodynamics, and actuator limitations. • Dynamic Couplings. Missile dynamics are coupled kinematically and inertially. The kinematic coupling terms can be isolated by casting the missile dynamic equations in the stability axes, whereas the inertial couplings, such as the roll–yaw coupling into pitch, can be accommodated by the feedback linearization approach because the extent of coupling is measurable. • Nonlinear Aerodynamic. Nonlinear aerodynamics are the result of the nonlinear and uncertain characteristics of the stability coefficients and control coefficients. A nonlinear control scheduling, as a function of Mach number, AOA, dynamic pressure, and so on, can be designed to remove control uncertainties caused by nonlinear aerodynamics and to approximately equalize the control effectiveness. • Actuator Limitations. The missile control surfaces have their limitations in the amounts of deflection and deflection rate. To avoid saturating the control surfaces, a command-limiting mechanism designed by dynamic inversion analysis needs to be implemented. Nonlinear dynamic inversion analysis also leads to an early understanding of design limitations, fundamental feedback paths, and a candidate feedback control structure. References 30 and 31 discuss some techniques used in nonlinear autopilot design. Adaptive Autopilot Design. Adaptive control systems attempt to adjust on-line to accommodate unknown or changing system dynamics as well as unknown exogenous system disturbances. There are two general classes of adaptive control laws: direct and indirect. A relatively simple indirect adaptive control solution for the autopilot design challenge is gain scheduled adaptation (32), where the autopilot is designed offline for a number of operating conditions and the required gains are prestored against related flight conditions. In con-

315

trast, direct adaptive controls such as the self-tuning regulator (33) and model reference adaptive control (34) update the autopilot gains directly on the basis of the history of system inputs and tracking errors. Intelligent Autopilot Design. Missile autopilot design task requires tuning parameters to achieve desirable performance. By augmenting a neural network in the tuning process, the parameter adjustment process can be standardized. This can be done as follows. First, build the desired flying qualities into the performance model. The autopilot structure is prefixed with the parameters undetermined. Then by comparing the actual system performance with the desired flying qualities, the neural network is trained to learn the rules of tuning. Accordingly, the autopilot parameters can be updated to meet the requirements. Application of neural network techniques to missile autopilot design and to future generation flight control system was investigated in Refs. 35 and 36.

BIBLIOGRAPHY 1. C. T. Myers, Guided Missiles—Operations, Design and Theory. New York: McGraw-Hill, 1958. 2. B. D. Richard, Fundamentals of Advanced Missiles. New York: Wiley, 1958. 3. M. R. Mendenhall, Tactical Missile Aerodynamics: Prediction Methodology. Washington DC: Amer. Inst. Aeronautics and Astronautics, 1992. 4. J. N. Nielsen, Missile Aerodynamics. New York: McGraw-Hill, 1960. 5. P. Garnell, Guided Weapon Control Systems, 2nd ed., Oxford: Pergamon, 1980. 6. W. A. Kevin and B. J. David, Agile missile dynamics and control. Proc. AIAA Guidance Navigation Control Conf., San Diego, CA, July 1996. 7. S. S. Chin, Missile Configuration Design. New York: McGrawHill, 1961. 8. A. Arrow, Status and concerns for bank-to-turn control of tactical missiles. AIAA J. Guidance, Control, Dynamics, 8 (2): 267–274, 1985. 9. F. W. Riedel, Bank-to-Turn Control Technology Survey for Homing Missiles, NASA CR-3325, 1980. 10. D. E. Williams, B. Friendland, and A. N. Madiwale, Modern control theory for design of autopilots for bank-to-turn missiles, AIAA J. Guidance, Control, Dynamics, 10 (4): 378–386, 1987. 11. J. H. Blakelock, Automatic Control of Aircraft and Missiles. New York: Wiley, 1991. 12. F. W. Nesline and M. L. Nesline, How autopilot requirements constraint the aerodynamic design of homing missiles. Proc. Amer. Control Conf., 1984, pp. 716–730. 13. M. P. Horton, Autopilots for tactical missiles; an overview. Proc. Inst. Mechanical Eng., Part 1, J. Syst. Control Eng., 209 (2): 127– 139, 1995. 14. C. F. Lin, Advanced Control System Design. Englewood Cliffs, NJ: Prentice-Hall, 1991. 15. H. Buschek, Robust autopilot design for future missile system, Proc. AIAA Guidance, Navigation, and Control Conference, New Orleans, 1997, pp. 1672–1681. 16. B. A. White, Eigenstructure assignment for aerospace applications, in A. J. Chipperfield and P. J. Flemming (eds.), IEE Control

316

17.

18. 19.

20.

21.

22.

23. 24.

25.

26.

27.

28.

29.

30.

31.

32. 33. 34.

35.

36.

MISSILE GUIDANCE Engineering Series, No. 48, London: Peregrinus, 1993, pp. 179–204. K. Sobel and J. R. Clotier, Eigenstructure assignment for the extended medium range missile, AIAA J. Guidance, Control, Dynamics, 13 (2): 529–531, 1992. R. E. Kalman, Contributions to the theory of optimal control, Boletin de la Sociedad Mathematica mexicana, 5: 102–119, 1960. F. W. Nesline, B. H. Wells, and P. Zarchan, A combined optimal/ classical approach to robust missile autopilot design, AIAA J. Guidance, Control, Dynamics, 4 (3): 316–322, 1981. F. W. Nesline and P. Zarchan, Why modern controllers can go unstable in practice, AIAA J. Guidance, Control, Dynamics, 7 (4): 495–500, 1984. D. G. Benshabat and Y. Chait, Application of quantitative feedback theory to class of missiles, AIAA J. Guidance, Control, Dynamics, 16 (1): 47–52, 1993. M. J. Ruth, A classic perspective on application of H앝 control theory to a flexible missile airframe, Proc. AIAA Guidance, Navigation Control Conf., Boston, MA: 1989, pp. 1073–1078. R. T. Reichart, Robust autopilot design using 애-synthesis, Proc. Amer. Control Conf., San Diego, CA, 1990, pp. 2368–2373. S. R. Baguley and B. H. White, A Study of H앝 robust control for missile autopilot design, Royal Military College of Science, Tech. Rep., Shrivenham, UK. P. Apkarian, J. M. Biannic, and P. Gahinet, Self-scheduled H앝 control of missile via linear matrix inequalities, AIAA J. Guidance, Control, Dynamics, 18 (3): 532–538, 1995. K. A. Wise, Comparison of six robustness tests evaluating missile autopilot robustness to uncertain aerodynamics, AIAA J. Guidance, Control, Dynamics, 15 (4): 861–870, 1992. H. J. Gratt and W. L. McCowan, Feedback linearization autopilot design for the advanced kinetic energy missile boost phase, AIAA J. Guidance, Control, Dynamics, 18 (5): 945–950, 1995. R. D. Weil and K. A. Wise, Blended aero & reaction jet missile autopilot design using VSS techniques, Proc. 30th IEEE Conf. Decision Control, Brighton, UK, 1991, pp. 2828–2829. K. A. Wise and J. L. Sedwick, Nonlinear H앝 optimal control for agile missiles, AIAA J. Guidance, Control, Dynamics, 19(1): 157– 165, 1996. P. K. Menon and M. Yousefpor, Design of nonlinear autopilots for high angle of attack missiles. Proc. AIAA Guidance, Navigation, Control Conf., San Diego, CA, 1996. K. A. Wise and J. L. Sedwick, Nonlinear H앝 optimal control for agile missiles. AIAA-95-3317, Proc. AIAA Guidance, Navigation, Control Conf., Baltimore, 1995, pp. 1295–1307. W. J. Rugh, Analytical framework for gain scheduling, Proc. Amer. Control Conf., San Diego, CA, 1990, pp. 1688–1694. C. F. Price and W. D. Koenigsberg, Adaptive control and guidance for tactical missiles, Reading, MA: Analytical Sci. Corporation. N. D. Porter, Further investigations into an adaptive autopilot control system for a tail controlled missile based on a variation of the model reference technique, Royal Aircraft Establishment, Tech. memor. DW8, Farnnborough, UK. M. B. McFarland and A. J. Calise, Neural-adaptive nonlinear autopilot design for an agile anti-air missile. Proc. AIAA Guidance, Navigation, Control Conf., San Diego, CA, 1996. M. L. Steinberg and R. D. DiGirolamo, Applying neural network technology to future generation military flight control systems. Int. Joint Conf. Neural Netw., 1991, pp. 898–903.

CIANN-DONG YANG CHI-CHING YANG HSIN-YUAN CHEN National Cheng Kung University

MISSILE CONTROL. See MISSILE GUIDANCE.

MISSILE GUIDANCE Missile guidance addresses the problem of steering, or guiding, a missile to a target on the basis of a priori known target coordinate information and/or real-time target measurements obtained from onboard and/or external sensors.

launched over 3000 V-2 missiles against Allied targets on the European continent, primarily Antwerp, Belgium, London, and Southern England. During the late 1940s and early 1950s, the U.S. Army, under Project Hermes, launched over 70 V-2s. The V-2 would became the prototype for future U. S. and Soviet rocket and strategic ballistic missile program developments.

A BRIEF HISTORY: FROM 1944 TO THE PRESENT

Lark-Guided Missile

The Missile Age

Because of the lack of success of anti-aircraft artillery in stopping Kamikaze aircraft attacks against naval vessels, the U.S. Navy initiated the development of the Lark guided missile in 1944. The ﬁrst successful intercept of an unmanned aircraft occurred six years later on December 2, 1950. An account of this, as well as the development of other missiles (e.g., Sparrow and Hawk), is provided in Reference 2.

Even before World War I—when powered ﬂight was in its ﬁrst decade—forward-thinking individuals from several countries advocated the use of unmanned vehicles to deliver high-explosive weapons from afar. Although the earliest efforts to develop a practical ﬂying bomb were undertaken in the United States and Great Britain, it was in Germany that a workable concept ﬁnally emerged. After 14 years of intense research, the Germans ushered in the missile age during World War II with their Vengeance weapons: the Luftwaffe-developed V-1 buzz bomb and the Army-developed V-2 rocket (1).

V-1 Buzz Bomb. Powered by a pulse-jet engine, generating 2670 N (600 pounds) of thrust, the V-1 reached a speed of 322 km per hour (200 miles per hour) and had a range of about 241 km (150 miles). Weighing 21,138 N (4750 pounds) with an 8900 N (2000 pound) high-explosive warhead, the V-1 was launched from a long ramp with the aid of a hydrogen peroxide/potassium permanganate-propelled booster motor. A gyroscope, magnetic compass, and a barometric altimeter were used to correct deviations in altitude and direction. Despite its 0.8045 km (0.5 mile) accuracy, the V-1 proved very useful as a terror weapon against large cities. Near impact, the control surfaces would lock and spoilers would be deployed from the tail to induce a steep dive. At this point, the pulsejet usually ceased functioning. The eerie silence that followed warned people below of the impending impact. The V-1 was launched by the thousands against London and the Belgian port of Antwerp during 1944, 1945. Well over 10,000 V-1s were launched against Great Britain, in all kinds of weather, by day and night. Although Royal Air Force pilots had some success in shooting down V-1s, the V-1s proved effective as terror weapons. V-2 Rocket. The V-2, which was developed at the secret Peenemunde ¨ rocket center, was ﬁrst used in combat on September 6, 1944. Fueled by an alcohol–liquid–oxygen propellant generating 244,750 N (55,000 pounds) of thrust for about 1-minute after launch, the V-2 had a range of about 322 km (200 miles). After a 1 minute powered ﬂight and reaching an altitude of about 113 km (70 miles), the V-2 ﬂew an arcing free-falling (ballistic) trajectory at speeds in excess of 1.609 km/s (1 mile per second)—carrying a 7,120 N (1,600 pound) warhead. Between September 1944 and March 1945, from mobile ﬁeld battery positions in France and Holland, German ﬁeld units

The First Ballistic Missiles After World War II, signiﬁcant improvements in inertial guidance system technology led to the Redstone missile— the ﬁrst short-range U.S. ballistic missile with a highly accurate inertial guidance system. Additional progress was made with the medium-range U.S. Jupiter missile (3). ICBMs Additional advancements in the area of nuclear warhead design, inertial guidance system, and booster engine technology led to the development of the intercontinental ballistic missile (ICBM). The ﬁrst U.S. ICBM—the Atlas— was tested in 1959. The Atlas would be used to launch satellites into orbit, launch probes to the moon and other planets, and to launch the Mercury spacecraft into orbit around the Earth. The Atlas was followed by the Titan one year later. Both Atlas and Titan were liquid-fuelled multistage rockets that needed to be fuelled just before launch. In 1961, the Minuteman ICBM was put into service. Located within dispersed hardened silos, the Minuteman used a solid propellant stored within the missile. The LGM-30 Minuteman III was deployed in 1970. This system was designed such that specially conﬁgured EC-135 airborne launch control aircraft could automatically assume command and control of an isolated missile or missiles in the event that command capability is lost between the launch control center and the remote missile launch facilities. In 1986, the LGM-118A Peacekeeper was deployed. This three-stage solid propellant system permits 10 warheads to be carried via multiple reentry independently targeted vehicles (MIRVs). At the peak of the Cold War, the Soviet Union possessed nearly 8,000 nuclear warheads on ICBMs. During the Cold war, the United States built up its strategic defense arsenal, focusing on a nuclear triad consisting of 1) long-range bombers (B-52 bombers and KC-135 tankers) with nuclear air-to-surface missiles, 2) U.S.-based ICBMs, and 3) submarine-launched ballistic missiles (SLBM) launched from nuclear-powered submarines. To complement the ground-based leg of the triad, the U.S. Navy would develop the submarine-launched Polaris, Poseidon, and Trident ICBMs. Trident I and II were

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright © 2007 John Wiley & Sons, Inc.

2

Missile Guidance

deployed in 1979 and 1988, respectively. Both accommodate nuclear MIRVs and are deployed in Ohio-class (Trident) submarines, each carrying 24 missiles with eight 100 kiloton warheads per missile. Trident II missiles weigh roughly 65 tons and are about 44 feet long and 7 feet wide. For comparison sake, it is worth noting that the bomb dropped on Hiroshima on August 6, 1945 (designated “Little Boy”) was a 8,900 lb, 10 feet long, 2.33 feet diameter, 13–16 kiloton uranium-235 based gun-type ﬁssion weapon. Similarly, the bomb dropped on Nagasaki three days later (designated “Fat Man”) was a 10,800 lb, 10.67 feet long, 5 feet diameter, 21 kiloton plutonium-239 based implosiontype ﬁssion weapon. Nuclear Non-Proliferation: SALT, ABM, and MAD The ﬁrst major Nuclear Non-Proliferation Treaty (NNPT) opened for signature on July 1, 1968. In addition to addressing what nations could “rightfully” possess nuclear weapons and relevant nuclear proliferation issues, it addressed disarmament and stockpile reduction as well as the peaceful use of nuclear technology (i.e., energy generation). The treaty is revisited periodically by participating states. Because of the large number of Soviet nuclear warheads during the Cold War, some in the United States felt that U.S. ICBM ﬁelds were threatened. On March 14, 1969, President Nixon announced his decision to deploy a missile defense system (called Safeguard) to protect U.S. ICBM ﬁelds from attack by Soviet missiles. This decision initiated intense strategic arms negotiations between the United States and the Soviet Union. The Strategic Arms Limitation Talks (SALT), between the United States and the Soviet Union, led to a 1971 agreement ﬁxing the number of ICBMs that could be deployed by the two nations. The Antiballistic Missile (ABM) Treaty—signed by the U.S. and the Soviet Union on May 26, 1972—was designed to implement the doctrine of mutually assured destruction (MAD). MAD was intended to discourage the launching of a ﬁrst strike by the certainty of being destroyed by retaliation. The treaty prohibits/limits deployment of certain sea, air, and spacebased missiles and sensors. A key motivation behind these arrangements was to perpetuate the existing balance of power and avoid the economic chaos that would result from a full-scale arms race. In 1976, in view of technical limitations imposed by the ABM treaty, the U.S. Congress ordered the closing of Safeguard only four months after becoming operational. In 2001, the ABM treaty came under attack in the U.S. Congress as the United States and Russia (former Soviet Union) discussed how to differentiate between theater and strategic missile defenses. BMD and SDI In 1983, President Reagan initiated the Ballistic Missile Defense (BMD) program under the Strategic Defense Initiative (SDI). SDI would focus on space-based defense research. Because SDI deployment would contravene the ABM treaty, many critics felt SDI, with its potential offensive use, would escalate the arms race. In 1984, the Strategic Defense Initiative Organization (SDIO) was formed. In 1987, Judge Abraham D. Sofaer, State Department Legal Advisor, concluded that the ABM treaty did not preclude

testing of space-based missile defense systems, including directed energy weapons. SDI research would continue. With the breakup of the Soviet Union in 1991, the need for great nuclear arsenals came into question. In 1993, the SDIO was replaced by the Ballistic Missile Defense Organization (BMDO). The national objectives of SDI were replaced by regional objectives. In 1998, emphassis shifted back to national missile defense. In 2002, BMDO was renamed the Missile Defense Agency (MDA).

Strategic Arms Reduction Treaties In November 1994, the Strategic Arms Reduction Treaty I (START I) became effective, with the United States, Russia, Belarus, Kazakstan, and Ukraine agreeing to reduce nuclear warheads by 25%. In appreciation for the ratiﬁcation, the United States appropriated $1.5 billion for assistance in dismantling nuclear weapons, properly storing weapons grade materials, and turning military factories into civilian buildings. Initiated in 1992, START II called for the removal of MIRVs and of almost three quarters of nuclear warheads over nine years, thereby reducing the U. S. and Russian arsenals to 3000–3500 strategic warheads. The U.S. Senate approved ratiﬁcation on January 26, 1996, but the Russian Duma never ratiﬁed the treaty. Multiple warhead Peacekeepers were to be eliminated by 2003 under START II. On June 13, 2002, the United States withdrew from the 1972 ABM treaty. The Russian Federation followed suit by withdrawing from START II negotiations the next day. The Treaty of Moscow, also referred to as the Strategic Offensive Reductions Treaty (SORT), was signed by Presidents George W. Bush and Vladimir Putin in 2002 and took effect in 2003. SORT promises to reduce the number of operationally deployed warheads from 6000 to 2200 by 2012.

Missile Warning Systems Although the United States has no active ABM defense system in place, an extensive warning system has been in place for many years. Air and space defense is delegated to the North American Aerospace Defense Command (NORAD)—a joint U.S.—Canadian organization that was initially founded May 12, 1958 as the North American Air Defense Command and adopted its current name in 1981. A Ballistic Missile Early Warning System (BMEWS) consisting of warning and tracking radars in Alaska, Greenland, and the United Kingdom can detect missiles 4800 km (∼3000 miles) away and provides a 15 minute warning of an attack on North America. A Perimeter Acquisition Radar Characterization System (PARCS), operating within the U. S. interior, tracks incoming warheads, and determines impact areas. Phased-array radar antennas along the U.S. Atlantic, Paciﬁc, Alaskan, and Gulf coasts provide warning of SLBM launches. With the collapse of the USSR in 1991 and the terrorist attacks on the United States of September 11, 2001, the NORAD mission has shifted considerably to the monitoring of all aircraft ﬂying within the interior of the United States.

Missile Guidance

3

Persian Gulf War In January 1991, the role of air power in modern warfare was dramatically demonstrated during the Persian Gulf War. Initial attacks by the U.S.-led multinational coalition were designed to suppress Iraqi air defenses. These attacks included Tomahawk cruise missiles launched from warships in the Persian Gulf, F-117A Stealth ﬁghter-bombers armed with laser-guided smart bombs, and F-4G Wild Weasel aircraft carrying HARM anti-radar missiles. These attacks permitted F-14, F-15, F-16, and F/A-18 ﬁghter bombers to achieve air superiority and to drop TV- and laser-guided precision bombs. During the ground war, A10 Thunderbolts with armor-piercing, heat-seeking, or optically guided AGM-65 Maverick missiles provided support for ground units. The AH-64 Apache and AH-1 Cobra helicopters ﬁred laser-guided Hellﬁre missiles, guided to tanks by ground observers or scout helicopters. The E-3A Airborne Warning and Control System (AWACS), a ﬂying radar system, provided targeting information to coalition members. Missile Defense Although most weapon systems performed superbly during the Gulf War, little could be done to stop the Iraqi Scuds launched against Saudi Arabia and Israel. However, a Patriot surface-to-air missile (SAM) system was brought in to repel Scud attacks. Although the Patriot system had been used in 1987 to destroy another Patriot during a demonstration ﬂight, the system was originally designed as an anti-aircraft defense system. Thus, its effectiveness against the Scuds was limited, primarily because intercepts often did not take place at sufﬁciently high altitudes. Part of the problem was attributed to the fact that the Patriot relied on proximity detonation rather than a hit-to-kill, which would often cause the incoming Scud to break up, leaving a free-falling warhead to detonate on the civilian population below. The many Patriot–Scud engagements were televised to a world audience and demonstrated the need for a high-altitude air defense system that could intercept (tactical) ballistic missiles far from critical military assets and civilian population centers. For this reason, much research shifted toward the development of hit-to-kill Theater High Altitude Air Defense (THAAD) systems that would focus on incoming targets situated within a 200 km (∼124 miles) range and no higher than 150 km (∼93 miles). In his January 1991 State of the Union Address, President George H. W. Bush formally announced a shift in SDI to a concept of Global Protection Against Limited Strikes (GPALS), and by December, he signed into law the Missile Defense Act of 1991. On January 24, 1997, a Standard Missile 2 (SM2) Block IVA successfully intercepted and destroyed a Lance missile at the White Sands Missile Range in New Mexico. During the test, the SM2 successfully transitioned from radar mid-course guidance to its heat-seeking endgame/terminal guidance system prior to destroying the target with its blast fragmentation warhead. On February 7, 1997, BMDO carried out a test in which a Patriot Advanced Capability-2 (PAC-2) missile successfully intercepted a theater ballistic target missile over the Paciﬁc Ocean. In April 1997, BMDO estab-

Figure 1. Information ﬂow for missile-target engagements.

lished the Joint Program Ofﬁce (JPO) for the National Missile Defense (NMD). On June 24, 1997, the ﬁrst NMD ﬂight test was successfully completed. During this test an Exoatmospheric Kill Vehicle (EKV) sensor was used to identify and track objects in space. In 2007, Lockheed Martin is expected to begin ﬂight testing of a THAAD system at the Paciﬁc Missile Range (Kauai, Hawaii). To appreciate the formidable problems associated with developing a THAAD system, it is necessary to understand the issues associated with the design of missile guidance systems. These issues will be addressed in subsequent sections. MISSILE GUIDANCE, NAVIGATION, AND CONTROL SUBSYSTEMS We begin our technical discussion by describing the subsystems that make up a missile system. In addition to a warhead, a missile contains several key supporting subsystems. These subsystems may include 1) a target-sensing system, 2) a missile-navigation system, 3) a guidance system, 4) an autopilot or control system, and 5) the physical missile (including airframe and actuation subsystem); see Fig. 1. Target-Sensing System The target-sensing system provides target “information” to the missile guidance system, e.g. relative position, velocity, line-of-sight angle, and rate. Target-sensing systems may be based on several sensors, e.g., radar, laser, heat, acoustic, or optical sensors. Optical sensors, for example, may be as simple as a camera for a weapon systems ofﬁcer (WSO) to visualize the target from a remote location. They may be a sophisticated imaging system (see below). For some applications, target coordinates are known a priori (e.g., via satellite or other intelligence) and a target sensor becomes irrelevant. Navigation System A navigation system provides information to the missile guidance system about the missile position in space relative to some inertial frame of reference, e.g., ﬂatEarth constant-gravity model for short-range ﬂights and rotating-Earth variable-gravity model for long-range ﬂights. To do so, it may use information obtained from a variety of sensors, which may include simple sensors such as accelerometers or a radar altimeter. It may include more sophisticated sensors such as a global positioning system (GPS) receiver or an optical terrain sensor that relies on

4

Missile Guidance

comparisons between an image of the terrain below with a stored image and a stored desired trajectory. Optical stellar sensors rely on comparisons between an image of the stars above with a stored image and a stored desired trajectory. Guidance System Target and missile information are used by the guidance system to compute updated guidance commands, which when issued to the missile autopilot should ideally guide (or steer) the missile toward the target (4, 5). When target coordinates are known a priori, missile coordinates provided by the navigation system (e.g., GPS-based) are periodically compared with the (pre-programmed) target coordinates to compute appropriate guidance corrections. In general, the quality of the computed guidance commands depends on the quality of the gathered sensor data and the ﬁdelity of the mathematical models used for the missile and target. Targets may be stationary, mobile, or highly maneuverable (e.g., silo, ship, ﬁghter aircraft). Physically, guidance commands may represent quantities such as desired thrust, desired (pitch/yaw) acceleration, desired speed, desired ﬂight path or roll angle, and desired altitude. Guidance commands issued by the guidance system to the missile autopilot are analogous to the speed commands issued by automobile drivers to the cruise control systems in their cars. In this sense, the missile guidance system is like the automobile driver and the missile autopilot is like the automobile cruise control system. Missile guidance commands are computed in accordance with a guidance algorithm. Guidance algorithms and navigational aids will be discussed below. Autopilot The primary function of the autopilot—sometimes referred to as the ﬂight control system (FCS) or attitude control system (ACS)—is to ensure 1) missile attitude stability and 2) that commands issued by the guidance system are followed as closely as possible (4). The autopilot accomplishes this command-following objective by computing and issuing appropriate control commands to the missile’s actuators. These actuators may include, for example, rocket thrusters, ramjets, scramjets (for hypersonic missiles), or servomotors that move aerodynamic control surfaces. More speciﬁcally, the autopilot compares commands issued by the guidance system with real-time measurements (e.g., acceleration, attitude and attitude rate, and altitude) obtained from onboard sensors (e.g., accelerometers, gyroscopes, and radar altimeters) and/or external tracking systems. This comparison, essentially a subtraction of signals, produces a feedback error signal, which is then used to compute control commands for the missile actuators. This computation, the purpose of the autopilot, may be based on a and is based on the autopilot design and hence its complexity. Autopilot design, however, is based on a very complex mathematical model that captures the following dynamical features: missile airframe, aerodynamics (depending on speed, dynamic pressure, angle-of-attack, slide-slip angle, etc.), actuators, sensors, ﬂexible modes, and uncertainty descriptions, e.g., dynamic uncertainty, parametric uncertainty (6, 7), and disturbance/noise bounds. It should

be noted that commands that are issued by the guidance system to the autopilot cannot always be followed exactly because of the presence of multiple sources of uncertainty. Sources of uncertainty may include disturbances acting on the missile, sensor noise, unmodeled or uncertain missile airframe, actuator, and sensor dynamics. Flight Phases The ﬂight of a missile can be broken into three phases: 1) a launch, separation, or boost phase; 2) a mid-course or cruise phase; and 3) an endgame or terminal phase. During each phase, a missile may use distinct guidance, navigation, and control systems, speciﬁcally designed to accommodate the requirements during that phase of the ﬂight. During each phase, the missile may very well use different sets of sensors, actuators, and power sources. Guidance System Performance Terminology To describe the function and performance of a guidance system, some terminology is essential. The imaginary line that connects a missile center-of-gravity (cg) to the target cg is referred to as the line-of-sight (8). The length of this line is called the range. The associated vector from missile to target is referred to as the range vector. The time derivative of the range vector is called the closing velocity. The most important measure of performance for any missile guidance system is the so-called miss distance. Miss distance is deﬁned to be the missile-target range at that instant when the two are closest to one another (8). The objective of most guidance systems is to minimize the miss distance within an allotted time period. For some applications (hit-to-kill), zero miss distance is essential. For some applications (e.g., to minimize collateral damage), it is essential to impact the target at a speciﬁc angle. Because miss distance is sensitive to many variables and small variations from missile to missile, other quantities are used to measure performance. One of the most common measures used is circular error probability (cep). The cep for a missile attempts to provide an average miss distance for a class of missile-target engagements (i.e., Monte Carlo runs). If a missile has a cep of 10 m, then most of the time, say, 68% of the time, it will detonate within 10 m of the target. CLASSIFICATION OF MISSILES, TARGETS, GUIDANCE SYSTEMS, NAVIGATION METHODS, AND TARGET-SENSING METHODS The guidance system used by a missile depends on the intended use of the missile. Missiles are classiﬁed according to many categories. The most commonly used classiﬁcations are as follows: strategic, tactical, exoatmospheric, endoatmospheric, aerodynamic, ballistic, surface-to-surface, surface-to-air, air-to-surface, air-to-air, inertially guided, terrain guided, stellar guided, satellite guided, passive; active, homing, command guided, radar guided, laser guided, heat seeking, ﬁre-and-forget, line-of-sight guided, radar terrain guided, TV guided, cruise, skid-to-turn (STD), and bank-to-turn (BTT). Each category is now brieﬂy discussed.

Missile Guidance

Strategic Missiles Strategic missiles are used primarily against strategic targets, that is, resources that permit an enemy to conduct large-scale military operations (e.g., battle management/command, control, and communication centers; industrial/weapons manufacturing centers; and so on). Such targets are usually located far behind the battle line. As such, strategic missiles are typically designed for long-range missions. Although such missiles are usually launched from naval vessels or from missile silos situated below ground, they are sometimes launched from aircraft (e.g., strategic bombers). Because such missiles are intended to eliminate the most signiﬁcant military targets, they typically carry nuclear warheads rather than conventional warheads. Strategic missiles typically operate at orbital speeds (∼5 miles per second), outside the atmosphere, and over intercontinental distances. They use rockets/thrusters/fuel and require very precise instrumentation for critical mid-course guidance. GPS has made such systems very accurate. Tactical Missiles Tactical missiles are used primarily against tactical targets, that is, resources that permit an enemy to conduct small-scale military operations (for example, a ship, an airﬁeld, and a munitions bunker). Such targets are usually located near the battle line. As such, tactical missiles are typically designed for short- or medium-range missions. Such missiles have generally carried conventional explosive warheads, the size of which depends on the designated target. Tactical missiles sometimes carry nuclear warheads in an effort to deter the use of tactical nuclear/chemical/biological weapons and to engage the most hardened targets (e.g., enemy nuclear strategic missile silos). Tactical missiles typically operate at lower speeds (< 1 mile per second), inside the atmosphere, and over short-tomedium distances (e.g., 150 miles). They typically use aerodynamic control surfaces (discussed below) and require adequate instrumentation for mid-course and terminal guidance. A target sensor (e.g., radar seeker) permits such missiles to engage mobile and highly maneuverable targets. Exoatmospheric Missiles Exoatmospheric missiles ﬂy their missions mostly outside the Earth’s atmosphere. Such missiles are used against long-range strategic targets. Because they ﬂy outside the atmosphere, thrusters are required to change direction. Such thrusters use onboard fuel. To maximize warhead size, and because missile weight grows exponentially with fuel weight, it is important that guidance and control systems for long-range missiles (e.g., strategic and exoatmospheric) provide for minimum fuel consumption. Endoatmospheric Missiles Endoatmospheric missiles ﬂy their missions inside the Earth’s atmosphere. Such missiles are used against strategic and tactical targets. In contrast to exoatmospheric missiles, endoatmospheric missiles may use movable control surfaces such as ﬁns (called aerodynamic control surfaces),

5

which deﬂect air ﬂow in order to alter the missile ﬂight path. In such a case, the missile is called an aerodynamic missile. Endoatmospheric missiles may, in some cases, rely entirely on rocket power. In such a case, they are not aerodynamic. Exoatmospheric missiles that ﬂy outside the Earth’s atmosphere rely on rocket power and thrusters. These missiles are not aerodynamic. Examples of aerodynamic missiles are the Sidewinder and Patriot. Ballistic Missiles Ballistic missiles assume a free-falling (unpowered) trajectory after an internally guided, self-powered (boost and mid-course) ascent. Such missiles are usually used against long-range strategic targets. ICBMs, for example, are usually exoatmospheric strategic missiles that were developed for use against strategic targets and are typically launched from underground missile silos and submarines. Modern ICBMs contain multiple independently targeted nuclear warheads deployed via MIRVs. Examples of ICBMs are the Atlas, Titan, Minuteman, Polaris, Peacekeeper, and Trident. The Iraqi Scud, used in the Persian Gulf War, is another ballistic missile. Surface-to-Surface Missiles (SSMs) SSMs are typically launched from the ground, beneath the ground (e.g., from a missile silo), or from naval platforms against ground targets (e.g., tank, munitions depot, and missile silo) or naval targets (e.g., battleship and submarine). ICBMs are typically SSMs. SSMs may carry nuclear, biological, chemical, or conventional warheads. Examples of SSMs are the anti-ship Silkworm and the Tomahawk. Surface-to-Air Missiles (SAMs) SAMs are typically launched from the ground, beneath the ground (e.g., from a missile silo), or from naval platforms against aircraft and missiles. SAMs were developed to defend surface targets from air attacks, especially from highaltitude bombers ﬂying well above the range of conventional anti-aircraft artillery (AAA). Most air defense SAMs employ separate radars to acquire (detect) and track enemy air threats. The separate radar is also used to guide the SAM toward the hostile target; endgame guidance may be accomplished by the missile’s onboard guidance system. SSMs are typically heavier and carry larger warheads than SAMs because they are usually intended to penetrate hardened (e.g., armored) targets. Shoulder launched SAMs (e.g., Stinger) have recently become a major concern given increased terrorist activities. Air-to-Surface Missiles (ASMs) ASMs are launched from aircraft against ground targets (e.g., a bridge or airﬁeld) or naval targets. Although ASMs are typically intended for tactical targets, they are used by both strategic and tactical bombers. Equipping strategic bombers with long-range ASMs extends their range, signiﬁcantly reducing the range that they need to travel toward the intended target. Examples of ASMs are the antitank Hawk and Hellﬁre, the anti-radar AGM-88 HARM, the anti-ship Exocet and AGM-84D HARPOON, and the

6

Missile Guidance

anti-armored vehicle AGM-65 Maverick. Other ASM systems include the Advanced Medium-Range Air-to-Air Missile (AIM-120 AMRAAM) and the airborne laser (ABL) system being developed by several defense contractors. The ABL system has been considered for boost-phase intercepts during which the launched missile has the largest thermal signature and is traveling at its slowest speed. Air-to-Air Missiles (AAMs) AAMs are launched from aircraft against aircraft, ballistic missiles, and most recently against tactical missiles. Such missiles are typically light, highly maneuverable, tactical weapons. AAMs are generally smaller, lighter, and faster than ASMs because ASMs are typically directed at hardened, less-mobile, targets. Some SAMs and ASMs are used as AAMs and vice versa. Examples of AAMs are the AIM7 Sparrow, AIM-9 Sidewinder, AIM-54 Phoenix, and the AIM-120A AMRAAM. Guidance Methods: Fixed Targets with Known Fixed Positions A missile may be guided toward a target having a known ﬁxed position using a variety of guidance methods and/or navigational aids, e.g., inertial, terrain, stellar, and satellite guidance and navigation.

Inertially Guided Missiles. Inertially guided missiles use missile spatial navigation information relative to some inertial frame of reference to guide a missile to its designated target. For short-range missions, one may use a ﬂat-Earth constant-gravity inertial frame of reference. This approach is not appropriate for long-range missions, approaching intercontinental distances, for which the Earth may not be treated as ﬂat. For such missions, the sun or stars provide an inertial frame of reference. One can also use an Earth-centered variable-gravity frame. Position information is typically obtained by integrating acceleration information obtained from accelerometers or by pattern-matching algorithms exploiting imaging systems. Because accelerometers are sensitive to gravity, they must be mounted in a ﬁxed position with respect to gravity. Typically, accelerometers are mounted on platforms that are stabilized by gyroscopes or startracking telescopes. Terrain and stellar navigation systems are examples of imaging systems. Satellite navigated missiles use satellites for navigation. Some satellite guided missiles use the Navstar GPS—a constellation of orbiting navigation satellites—to navigate and guide the missile to its target. GPS has increased accuracy (reduced miss distance) signiﬁcantly. Guidance Methods: Mobile Targets with Unknown Positions If the target position is not known a priori, the aforementioned methods and aids may be used in part, but other real-time target acquisition, tracking, navigation, and guidance mechanisms are required. The most com-

monly used classiﬁcations for the guidance system in such cases are as follows: passive, active, and semiactive. Each of these and related topics is now discussed.

Passive Missiles. Passive missiles are missiles that have a target sensor sensitive to target energy emissions (e.g., radar and thermal energy) and a guidance system that uses received target emission signals to guide the missile toward the target. Such missiles are said to have a passive guidance system. Although such systems are, in principle, simple to implement, it should be noted that they rely on a “cooperative target,” i.e., targets that radiate energy at appreciable (detectable) power levels. Such systems are also susceptible to decoys. Active Missiles. Active missiles use an energy-emitting transmitter combined with a reﬂection–detection receiver (e.g., an active seeker) to acquire targets and guide the missile toward the target. Such missiles are said to have an active guidance system. For such systems, great care is taken to ensure that transmitted and received signals are isolated from one another. Stealthy targets are those that absorb or scatter (misdirect) the transmitted energy. Receivers can consist of a gimballed (movable) seeker antenna. Such mechanically directed antennas are slow and have a limited ﬁeld of view. Fixed phase array antennas, operating on interferometric principles; offer rapid electronic scanning capability as well as a broad ﬁeld of view. Semiactive Missiles. Semiactive missiles use a reﬂection-sensitive receiver to guide the missile to the target. The reﬂected energy may be provided by a ground-based, ship-based, or aircraft-based energy emission (e.g., radar or laser) system or by such a system aboard the launching platform. In either case, a human operator (e.g., WSO) illuminates the target with a radar or laser beacon and the missile automatically steers toward the source of the reﬂected energy. Such missiles are said to possess semiactive guidance systems. For such implementations, the illuminating power can be large. Passive systems, of course, are stealthier than semiactive or active systems as they do not intentionally emit energy toward the target. Anti-radar missiles typically use passive guidance systems because radars are constantly emitting energy. As an anti-radar missile approaches the intended radar, radar operators typically shut down the radar, which causes the missile to lose its critical guidance signal. In such a case, an active or semi-active guidance system must take over. It should be noted that active systems require more instrumentation than passive systems and hence are heavier and more expensive. – Homing Missiles. Homing missiles, like homing pigeons, home in on a target by steering toward energy emitted by or reﬂected from the target. If the missile homes in on energy emitted by the target, then it uses a passive guidance system.

Missile Guidance

If the missile transmits a signal and homes in on the reﬂected energy, its guidance system is active. In principle, sensor information and homing improve as the missile gets closer to the target. – Command Guided Missiles. A command guided missile is a remotely controlled semiactive missile. A cooperating (ground, ship, or aircraft-based) control station uses a radar (or two) to acquire the target, track the target, and track the missile. Available computers are used to compute guidance commands (on the basis of ranges, elevations, and bearings) that are transmitted via radio uplink to the missile autopilot. Powerful computers, capable of exploiting complex target models and performance criteria, can provide precision guidance updates in real time. Such systems are limited by the distance from the tracking station to the missile and target. Noise increases, and guidance degrades, as the engagement moves further from the tracking station. Such systems are also more susceptible to electronic countermeasures (ECMs). Although command-guided missiles do not require a seeker, one can be included for terminal guidance to maximize the probability of interception at long distances from the tracking station. The Patriot is a command-guided SAM. To signiﬁcantly increase ECM immunity, some short-range command-guided missiles have a wire that unspools at launch, which keeps the missile connected to the command station, e.g., all-weather optically guided anti-tank Tow missile.

–

–

–

1. Beam Rider Guidance (BRG). BRG is a speciﬁc form of command guidance in which the missile ﬂies along a beam (e.g., radar or laser), which in principle points continuously toward the target. If the missile stays within the beam, an intercept will occur. Guidance commands steer the missile back into the beam when it deviates. BRG causes problems at large ranges because of beam-spreading issues. 2. Command-to-LOS Guidance. Command-to-LOS guidance—used by the Tow missile—is another command guidance method that improves on beam rider guidance by taking beam motion into account. – Energy-Guided Missiles. Radar-guided missiles are guided to the target on the basis of radar energy. Laser-guided missiles are guided on the basis of laser energy. The Hellﬁre is a laserguided anti-tank missile. Heat-seeking missiles are guided on the basis of infrared (IR, heat, or thermal) energy. The AIM-9 Sidewinder is a heatseeking AAM. Most AAMs employ radar homing or heat-seeking devices and have replaced automatic gunﬁre as the main armament for ﬁghter aircraft. The shoulder-operated Stinger is a heatguided ﬁre-and-forget SAM. Such a missile is

–

7

called a ﬁre-and-forget missile because it allows the user to ﬁre, take evasive action, forget, and engage other hostile targets. Degradation of Electromagnetic EnergyBased Sensors. The performance of many electromagnetic energy-based sensors (e.g., millimeter wave radars, electro-optical thermal imagers, and laser radar) degrades under adverse weather conditions such rain, fog, dust, or smoke. This degrading occurs when the size of the weather particles are on the same order as the wavelength of the energy return from the target. Under adverse conditions, microwave radars with wavelengths in centimeters (10 GHz) are not degraded, millimeter radars with millimeter wavelengths (100 GHz) are slightly degraded, and electro-optical systems with micrometer wavelengths (105 GHz) are severely degraded. The AIM-120A AMRAAM is a ﬁghter-launched ﬁre-and-forget AAM that uses IR sensors to acquire (detect) targets at long range. It uses inertial mid-course guidance without the need for the ﬁghter to illuminate the target. A small active seeker is used for endgame homing. LOS Guidance. When a missile is near the target, the guidance system may use line-of-sight (LOS) guidance. The guidance system of an LOSguided missile uses target range and LOS information obtained from the target sensor (e.g., a seeker) to generate guidance commands to the missile autopilot. Radar Terrain Guidance. A radar terrainguided missile uses a radar altimeter, an a priori stored path and terrain proﬁle to navigate and guide the missile over the terrain during the mid-course phase of a ﬂight (typically). The stored path represents a desired path over the terrain. The down-looking radar altimeter is used to measure the altitude with respect to the terrain below, which is used to determine where the missile is with respect to the desired path. Deviations from the path are corrected by adjusting guidance commands to the autopilot. The Tomahawk is an all-weather cruise missile that uses radar terrain guidance called Terrain Contour Matching (TERCOM) (9). TERCOM terrain proﬁles—obtained by reconnaissance satellites and other intelligence sources—become ﬁner as the missile approaches the target. Such navigational/guidance systems permit terrain hugging. Terrain echoes (referred to as clutter) then confuse observing radars. TV Guidance. TV-guided missiles use imaging systems that permit a WSO to see the target and remotely guide the missile to the target.

Cruise Missiles Cruise missiles are typically SSMs that use inertial and terrain following navigation/guidance systems while cruising toward the target. When near the target, endgame guid-

8

Missile Guidance

ance is accomplished by either homing in on 1) target emitted/reﬂected energy, and 2) a target feature by exploiting a forward-looking imaging system and an onboard stored image, or by 3) using a more detailed terrain contour with a more-accurate downward-looking sensor. Cruise missiles offer the ability to destroy heavily defended targets without risking air crew. Because they are small, they are difﬁcult to detect on radar, particularly when they hug the terrain. Examples of cruise missiles are the AGM-86, Tomahawk (9), and Harpoon. The Tomahawk uses a TERCOM guidance during the cruise-phase. For terminal guidance, a conventionally armed Tomahawk uses an electro-optical Digital Scene-Matching Area Correlator (DSMAC) guidance system that compares measured images with stored images. This technique is often referred to as an offset navigation or guidance technique. At no time during the terminal scene-matching process does the missile look at the target. Its sensor always looks down. DSMAC makes Tomahawk one of the most accurate weapon systems in service around the world. Skid-to-Turn and Bank-to-Turn Missiles Skid-to-turn (STT) missiles, like speed boats, skid to turn. Bank-to-turn (BTT) missiles, like airplanes, bank to turn (5, 10–16). BTT airframe designs offer higher maneuverability than conventional STT designs by use of an asymmetrical shape and/or the addition of a wing. BTT missile autopilots are more difﬁcult to design than STT autopilots because of cross-coupling issues. STT missiles achieve velocity vector control by permitting the missile to develop angle-of-attack and side-slip angles (5). The presence of slide-slip imparts a skidding motion to the missile. BTT missiles ideally should have no side-slip. To achieve the desired orientation, a BTT missile is rolled (banked) so that the plane of maximum aerodynamic normal force is oriented to the desired direction. The magnitude of the force is controlled by adjusting the attitude (i.e., angle-ofattack) in that plane. BTT missile control is made more difﬁcult by the high roll rates required for high performance (i.e., short response time) (4). STT missiles typically require pitch-yaw acceleration guidance commands, whereas BTT missiles require pitch-roll acceleration commands. An overview of tactical missile control design issues and approaches is provided in Reference 17. GUIDANCE ALGORITHMS In practice, many guidance algorithms are used (4, 8, 18–20). The purpose of a guidance algorithm is to update missile guidance commands that will be issued to the autopilot. This update is to be performed on the basis of missile and target information. The goal of any guidance algorithm is to steer the missile toward the target, which results in an intercept within an allotted time period (that is, until the fuel runs out or the target is out of range). The most common algorithms are characterized by the following terms: proportional navigation, augmented proportional navigation, and optimal (8, 20). To simplify the mathematical details of the exposition to follow, suppose that the missile-target engagement is restricted to the pitch plane

of the missile. Given this, the engagement dynamics take the following simpliﬁed form (21): ˙ = Vt cos(λ(t) − γt (t)) − Vm cos(λ(t) − γm (t)) R(t) ˙ = λ(t)

1 [−Vt sin(λ(t) − γt (t)) + Vm sin(λ(t) − γm (t))] R(t)

(1) (2)

where R represents range to the target, λ represents LOS angle, and (Vm , Vt ) and (γm , γt ) denote missile-target speeds (assumed constant) and ﬂight path angles. Proportional Navigation Guidance (PNG) For proportional navigation guidance (PNG) (8, 20), the missile is commanded to turn at a rate proportional to the closing velocity Vc (i.e., range rate) and to the angu˙ For a PNG law, the pitch plane lar velocity of the LOS λ. acceleration command ac PNG (t) takes the form ˙ acPNG (t) = NVc (t)λ(t)

(3)

where N is a constant of proportionality referred to as the PNG gain or constant. For tactical radar homing missiles using PNG, an active seeker provides LOS rate while a Doppler radar provides closing velocity. Traditionally, LOS rate has been obtained by ﬁltering the output of a 2 degreeof-freedom rate gyro mounted to the inner gimbal of the seeker (22). More recently, ring laser gyros (RLGs) have been used. Unlike conventional spinning gyros, the RLG has no moving parts, no friction, and hence negligible drift. For IR missiles using PNG, the IR system provides LOS rate information, but Vc must be estimated. PNG Optimality and Performance Issues It can be shown that PNG minimizes the square integral t criterion 0 f ac2 (τ)dτ subject to the following assumptions: 1. Zero miss distance at tf 2. Linearized (small angle) missile-target dynamics 3. constant missile-target speeds (23) where tf denotes the ﬂight time. A missile using PNG is ﬁred not at the target, but at the expected intercept point if the target were to move at constant velocity in a straight line; i.e., the missile is ﬁred so that, at least initially, it is on a collision triangle with the target. The initial angle between the missile velocity vector and the LOS is the missile lead angle. If the missile is not on a collision triangle with the target, then a heading error (HE) exists. It is instructive to understand how PNG missile acceleration requirements vary with 1. Initial heading error when the target is not maneuvering 2. A constant acceleration target maneuver These cases are now brieﬂy discussed assuming linearized (small-angle) two-dimensional (2D) dynamics with constant missile and target speeds (Vm , Vt ) missile autopilot responds instantaneously to guidance acceleration commands (i.e., no lag), and ideal sensor dynamics (8). We note

Missile Guidance

that the Stinger is an example of a ﬁre-and-forget supersonic SAM that uses PNG with passive IR/UV homing. 1. PNG Performance: Non-maneuvering Target, Heading Error. First, consider the impact of a heading error on PNG missile acceleration requirements when the target moves at a constant speed in a straight line. Under the simplifying assumptions given above, the resulting commanded acceleration is as follows:

−Vm HEN t acPNG (t) = 1− tf tf

N−2 (4)

This expression shows that PNG immediately begins removing any heading error and continues doing so throughout the engagement. The acceleration requirement decreases monotonically from acPNGmax = −Vm HEN acPNG (0) = to zero as the ﬂight progresses. A tf larger N results in a larger initial missile acceleration requirement, but a lesser endgame missile acceleration requirement. The larger the N, the faster the heading error is removed. 2. PNG Performance: Target Undergoing Constant Acceleration. Now, consider the impact of a constant target acceleration at on PNG missile acceleration requirements. Under the simplifying assumptions given above, the resulting commanded acceleration is as follows:

N acPNG (t) = 1− N −2

t 1− tf

N−2

at

(5)

In sharp contrast to the heading error case examined above, this expression shows that the PNG missile acceleration requirement for a constant target maneuver increases monotonically throughout the ﬂight. As in the heading error case, a higher N results in a greater initial acceleration requirement and a relaxed acceleration requirement near the end of the ﬂight (acPNGmax = N acPNG (tf ) = [ ]at ≤ at ). N −2 Zero Effort Miss (ZEM) Distance An important concept in guidance law design is that of zero effort miss distance, denoted ZEM(t) and deﬁned as the miss distance that would result if the target would continue at a constant speed in a straight line and the missile made no further corrective maneuvers. Given this, if one deﬁnes the time-to-go as tgo = def tf − t and the zero effort miss distance perpendicular to the LOS as ZEMPLOS (t) then for PNG it can be shown that

acPNG (t) = N

ZEMPLOS (t) 2 tgo

(6)

where ZEMPLOS (t) = y + y˙ tgo , y ≈ Rλ denotes the relative (small angle) vertical displacement between the missile and target, and R ≈ Vc tgo . The concept of zero effort miss distance is used to derive more advanced guidance laws (8). The concept is very powerful since ZEM can be approximated in so many different ways.

9

PNG Miss Distance Performance: Impact of System Dynamics For the two cases considered above, the associated relative displacement y ≈ Rλ satisﬁes

y+

N tf − t

y+

N (tf − t)2

y˙ = at

y(tf ) = 0

(7)

and we have zero miss distance. The preceding discussion on PNG assumes that guidance-control-seeker dynamics are negligible. In practice, this assumption is not satisﬁed and the inherent lag degrades miss distance performance. When a ﬁrst-order lag with time constant τ is assumed for the combined guidance-control-seeker dynamics, one obtains small miss distances so long as τ is much smaller than tf , e.g., tf > 10τ. In practice, of course, high-frequency dynamics impose bandwidth constraints that limit how small τ can be. Despite the above (general) rule-of-thumb, it is essential that high-frequency system dynamics be carefully modeled/analyzed to obtain reliable performance predictions. Such dynamics include those associated with control system, computational delays, A/D and D/A conversion, actuators (e.g., thrusters, canards, and tail ﬁns), missile structure (e.g., ﬂexible and servoelastic modes), guidance system (e.g., lead-lag compensation), and sensors (e.g., seeker radome, accelerometers, gyros). As one might expect, noise and parasitic effects place a practical upper bound on the achievable guidance system bandwidth. In practice, statistical Monte Carlo simulations [exploiting adjoint methods (8)] are used to evaluate performance before ﬂight testing. Such simulations consider the above as well as acceleration/control saturation effects (14, 15), typical target maneuvers, and worst case target maneuvers. TPNG and PPNG In Reference 24, the authors distinguish between true PNG (TPNG) and pure PNG (PPNG). For missile’s using TPNG, acceleration commands are issued perpendicular to the LOS (as above). For PPNG, acceleration commands are issued perpendicular to the missile velocity vector. The advantages of PPNG over traditional TPNG are highlighted in Reference 24. In contrast to PPNG, TPNG requires 1) a forward acceleration and deceleration capability (because acceleration command is perpendicular to LOS; not missile velocity), 2) unnecessarily large acceleration requirements, and 3) restrictions on the initial conditions to ensure intercept. Tactical Missile Maneuverability Tactical radar-guided missiles use a seeker with a radome. The radome causes a refraction or bending of the incoming radar wave, which in turn, gives a false indication of target location. This phenomenon can cause problems if the missile is highly maneuverable. One parameter that measures maneuverability is the so-called missile (pitch) turning rate frequency (or bandwidth) deﬁned by (2) def

ωα =

γ˙ α

(8)

where γ˙ denotes the time rate of change of ﬂight path angle and α denotes angle-of-attack (AOA). ωα measures

10

Missile Guidance

the rate at which the missile rotates (changes ﬂight path) by an equivalent AOA. Assuming that the missile is modeled as a “ﬂying cylinder” (8) with length L and diameter Splan D, it has a lift coefﬁcient CL = 2α[1 + 0.75 α], where Sref 2 πD . Noting that am = Vm γ˙ is the missile Splan ≈ LD, Sref = 4 1 acceleration, Q = ρVm2 the dynamic pressure, W = mg the 2 missile weight, and ρ the density of air, it follows that

def

ωα =

αm Vm

γ˙ gQSref CL = = = α α αVm

ρgVm Sref 1 + 0.75

Splan Sref

(9)

From this, it follows that ωα decreases with increasing missile altitude and with decreasing missile speed Vm .

Let ω denote the guidance-control-seeker bandwidth.

Homing Requirement. If ω is too small, homing is poor and large miss distances result. Typically, we desire (10)

that is, the guidance-control-seeker bandwidth should be sufﬁciently large so that the closed-loop system “accommodates” the maneuverability capabilities of the missile, which implies that the guidance-controlseeker bandwidth ω must be large when ωα is large (low altitude and high missile speed Vm ). Robustness Requirement. If ω were too large, however, it is expected that problems can occur. This result in part, is because of radome-aerodynamic feedback ˙ Assuming n-pole of the missile acceleration am into λ. dynamics, it can be shown that the missile acceleration am takes the form

FG λ˙ 1 + FGRA (11)

where G = NVc represents the guidance system, F = ω n ( ) represents the ﬂight control system, R is the s+ω radome slope (can be positive or negative), and A = s + ωα denotes the missile transfer function from am ωα Vm to pitch rate θ˙ For stability robustness, we require the associated open-loop transfer function def

L = FGRA = NVc

ω n s + ω α R

s+ω

ωα V m

(12)

to satisfy an attenuation speciﬁcation such as ω |L( jω)| ≈ NVc |R| <ε for some sufﬁciently small ω α Vm constant ε > 0. This result however, requires that the guidance-control-seeker bandwidth ω satisﬁes

ω<ε

Vm |R|NVc

ωα

Vm |R|NVc

ωα

(14)

The lower inequality should be satisﬁed for good homing. The upper inequality should be satisﬁed for good robustness with respect to radome effects.

When ωα is small (e.g., at high altitudes or low speeds),

Radome Effects: Homing-Robustness Trade-offs

˙ = FG[λ˙ − RAam ] = am = FG[λ˙ − Rθ]

From the above, it follows that we require the guidancecontrol-seeker bandwidth ω to satisfy ωα <ω<ε

W

ωα <ω

for stability robustness which implies that the guidance-control-seeker bandwidth bandwidth ω must be small when Vm is small, (|R|, N, Vc ) are large, or ωα is small (high altitude and low missile speed Vm ).

(13)

designers make the guidance-control-seeker bandwidth ω small but sufﬁciently large to accommodate missile maneuverability (i.e., satisfy the lower inequality). In such a case, radome effects are small and the guidance loop remains stable yielding zero miss distance after a sufﬁciently long ﬂight. One can, typically, improve homing performance by increasing ω and N. If they are increased too much, radome effects become signiﬁcant, miss distance can be high, and guidance loop instability can set in. When ωα is large (e.g., at low altitudes or high speeds), designers would still like to make the guidancecontrol-seeker bandwidth ω sufﬁciently large to accommodate missile maneuverability (i.e., satisfy the lower inequality). This, result generally, can be accomplished provided that radome effects are not too signiﬁcant. Radome effects will be signiﬁcant if Vm is too small, (|R|, N, Vc ) are too large, or ωα is too small (i.e., too high an altitude and/or too low a missile speed Vm ). Given the above, it therefore follows that designers are generally forced to trade off homing performance (bandwidth) for stability robustness properties. Missiles using thrust vectoring (e.g., exoatmospheric missiles) experience similar performance-stability robustness trade-offs. Augmented Proportional Guidance (APNG) Advanced guidance laws reduce acceleration requirements and miss distance but require more information (e.g., timeto-go and missile-target range) (19). In an attempt to take into account a constant target acceleration maneuver at , guidance engineers developed augmented proportional guidance (APNG). For APNG, the commanded acceleration is given by ˙ + acA PNG (t) = NVc λ(t)

1 1 Nat = ac PNG (t) + Nat 2 2

(15)

1 2 ZEM ), where ZEM = y + y˙ tgo + at tgo is 2 tgo 2 the associated zero effort miss distance. Equation (15) shows that APNG is essentially PNG with an extra term to account for the maneuvering target. For this guidance law,

or acA PNG (t) = N(

Missile Guidance

it can be shown (under the simplifying assumptions given earlier) that

acA PNG (t) =

1 t N 1− 2 tf

N−2

at

(16)

In contrast with PNG, this expression shows that the resulting APNG acceleration requirements decrease with time rather than increase. From the expression, it follows that increasing N increases the initial acceleration requirement but also reduces the time required for the acceleration requirements to decrease to negligible levels. For N = 4, the maximum acceleration requirement 1 for APNG, acA PNGmax = Nat , is equal to that for PNG, 2 N ]at . For large N = 5, APNG requires a acPNGmax = [ N −2 larger maximum acceleration but less acceleration than PNG for t ≥ 0.2632tf . As a result, APNG is more fuel efﬁcient for exoatmospheric applications than t PNG. Finally, it should be noted that APNG minimizes 0 f ac2 (τ)dτ subject to zero miss distance, linear dynamics, and constant target acceleration (8). PNG Command Guidance Implementation To implement PNG in a command guidance setting (i.e., no seeker), a differentiating ﬁlter must be used to estimate the LOS rate. As a result, command guidance is more susceptible to noise than homing guidance. This issue is exacerbated as the engagement takes place further from the tracking station, noise increases, and guidance degrades. Within Reference 25, the authors address command-guided SAMs by spreading the acceleration requirements over tgo . The method requires estimates for target position, velocity, acceleration, and tgo but takes into account nonlinear engagement geometry. Advanced Guidance Algorithms Classic PNG and APNG were initially based on intuition. Modern or advanced guidance algorithms exploit optimal control theory, i.e. optimizing a performance measure subject to dynamic constraints. Even simple optimal control formulations of a missile-target engagement (e.g., quadratic acceleration measures) lead to a nonlinear two-point boundary value problem requiring creative solution techniques, e.g., approximate solutions to the associated Hamilton–Jacobi–Bellman equation—a formidable nonlinear partial differential equation (23). Such a formulation remains somewhat intractable given today’s computing power, even for command guidance implementations that can exploit powerful remotely situated computers. As a result, researchers have sought alternative approaches to design advanced (near-optimal) guidance laws. In Reference 20, the authors present a PNG-like control law that optimizes square-integral acceleration subject to zero miss distance in the presence of a one pole guidance-controlseeker system. Even for advanced guidance algorithms (e.g., optimal guidance methods), the effects of guidance and control system parasitics must be carefully evaluated to ensure nominal performance and robustness (20). Advanced (optimal)

11

guidance methods typically require additional information such as time-to-go, target acceleration, target model parameters (e.g., ballistic coefﬁcient). As a result, Kalman ﬁlter and extended Kalman ﬁlter (EKF) techniques are often used to estimate the required information. For optimal guidance (OG) algorithms to work well, the estimates must be reliable (20). An overview of guidance and control techniques, including a comprehensive set of references, is given in Reference 18. Other approaches to guidance law design are discussed below.

Variants of PNG Within Reference 20, the authors compare PNG, APNG, and optimal guidance (OG). The zero miss distance (stability) properties of PPNG are discussed within Reference 24. A nonlinear PPNG formulation for maneuvering targets is provided in Reference 27. Closed form expressions for PPNG are presented in Reference 28. A more complex version of PNG that is “quasi-optimal” for large maneuvers (but requires tgo estimates) is discussed in Reference 29. Two-dimensional miss distance analysis is conducted in Reference 21 for a guidance law that combines PNG and pursuit guidance. Within Reference 30, the authors extend PNG by using an outer LOS rate loop to control the terminal geometry of the engagement (e.g., approach angle). Generalized PNG, in which acceleration commands are issued normal to the LOS with a bias angle, is addressed in Reference 31. Three-dimensional (3D) generalized PNG is addressed within Reference 32 using a spherical coordinate system ﬁxed to the missile to better accommodate the spherical nature of seeker measurements. Analytical solutions are presented without linearization. Generalized guidance schemes are presented in Reference 33, which result in missile acceleration commands rotating the missile perpendicular to a chosen (generalized) direction. When this direction is appropriately selected, standard laws result. Time-energy performance criteria are also examined. Capturability issues for variants of PNG are addressed in Reference 34 and the references therein. Within Reference 35, the authors present a 2-D framework that shows that many developed guidance laws are special cases of a general law. The 3-D case, using polar coordinates, is considered in Reference 36.

Optimal Guidance (OG) Laws Weaving targets can cause large miss distances when classicl and “standard” OG laws are used. Tactical ballistic missiles, for example, can spiral or weave into resonances as they enter the atmosphere because of mass or conﬁgurational asymmetries. An OG law, based on weaving (variable amplitude) sinusoidal target maneuvers, is developed in Reference 37). An EKF is used to estimate the target maneuver weave frequency. Methods for intercepting spiraling weaving tactical ballistic targets are also presented in Reference 38, which includes an optimal weave guidance law incorporating an EKF to estimate relative position, relative velocity, target acceleration, target jerk information, and weave frequency information.

12

Missile Guidance

Differential Game Guidance Differential game-theoretic concepts are addressed within Reference 23. In such formulations, a disturbance (e.g., target maneuver) “competes” with a control (e.g., missile acceleration command). The disturbance attempts to maximize a performance index (e.g., miss distance), where as the control attempts to minimize the index. Within Reference 39), the authors provide an analytical study using a zero-sum pursuit-evasion differential game formulation to develop endgame guidance laws assuming that the interceptor has two controls. Linear bi-proper transfer functions are used to represent the missile’s control systems—a minimum phase transfer function for the canard system and a non-minimum phase (NMP) transfer function for the tail control system. A ﬁrst-order strictly proper transfer function is used for the target dynamics. Bounds are assumed for each of the above transfer function inputs (i.e., reference commands). The optimal strategy is bang-bang in portions of the game space. A switching time exists before interception because of the NMP nature of the tail control system. This feature requires good estimates of tgo .H∞ theory (7) provides a natural differential game theoretic framework for developing guidance laws as well as for control laws. Lyapunov-Based Guidance Laws Lyapunov methods have been very useful for deriving stabilizing control laws for nonlinear systems (40). Such methods have been used to obtain guidance laws that require target aspect angle (relative to LOS) rather than LOS rate (41) and that address maneuvering targets in 3-D (42). Other Guidance Laws Circular navigation guidance (CNG) steers the missile along a circular arc toward the target (43). Traditionally, the guidance and control systems are designed separately. Although this approach has worked well for years, increasing performance requirements afﬁrm the value of an integrated guidance and control system design methodology. Integrated guidance and control issues are addressed within a polar coordinate framework in Reference 44. New advanced guidance laws may beneﬁt from linear parameter varying (LPV) (17) and state dependent Riccati equation (SDRE) (45) concepts. Nonlinear State Estimation: Extended Kalman Filter As discussed OG laws often require missile-target model state/parameter estimates, e.g., relative position, velocity of target, acceleration of target, tgo. An extended Kalman ﬁlter (EKF) is often used to obtain the required estimates, which involves using quasi-linearized dynamics to solve the associated matrix Riccati differential equation for a covariance matrix that is used with a model based estimator—mimicking the original nonlinear dynamics— to generate quasi-optimal estimates. It is well known that poor estimates for tgo, for example, can result in large miss distances and signiﬁcant capture region reduction (20). Estimating tgo as R/Vc is valid only if Vc is nearly constant. A recursive (noniterative) algorithm for tgo estimates, that can be used with OG laws, is provided within Reference

46. To develop useful estimation techniques, much attention has been placed on modeling the target. Initially, researchers used simple uncorrelated target acceleration models. This process however, yielded misleading results, which led to the use of simple dynamical models—point mass and more complex. Both Cartesian and spherical coordinate (47) formulations have been investigated—the latter better reﬂecting the radial nature of an engagement. Single and multimodeled EKFs have been used (48) to address the fact that no single model captures the dynamics that may arise. Low-observability LOS measurements make the problem particularly challenging (48). Target observability is explored in Reference 49 under PNG and noise-free angle-only measurements in 2-D. A method for obtaining required estimates for APNG (e.g., y, y, at , tgo ) is presented in Reference 50. As no single (tractable) model and statistics can be used to accurately capture the large set of possible maneuvers by today’s modern tactical ﬁghters, adaptive ﬁltering techniques have been employed. Such ﬁlters attempt to adjust the ﬁlter bandwidth to reﬂect the target maneuver. Some researchers have used classic Neyman–Pearson hypothesis testing to detect bias in the innovations to appropriately reinitialize the ﬁlter. Threshold levels must be judiciously selected to avoid false detections that result in switching to an inappropriate estimator. Long-Range Exoatmospheric Missions: Weight Considerations For long-range exoatmospheric missions approaching intercontinental ranges, orbital speeds are required (e.g., ∼ 20,000 ft/sscond or 13,600 miles/hour or 4 miles/second). To study such interceptors, two new concepts are essential. Fuel-speciﬁc impulse, denoted Isp , is deﬁned as the ratio of thrust to the time rate of change of total missile weight. It corresponds to the time required to generate a weight equivalent amount of thrust. Fuel-efﬁcient missiles have higher fuel-speciﬁc impulses. Typical tactical missile fuel-speciﬁc impulses lie in the range of 200 to 300 seconds. Fuel-mass fraction, denoted mf, is deﬁned as the ratio of propellant weight Wprop to total weight WT = Wprop + Wstructure + Wpayload . SAMs, for example, have a larger fuel-mass fraction than AAMs because SAMs must travel through the denser air at lower altitudes. For fuel-speciﬁc impulses less than 300 seconds, large fuelmass fractions (approaching 0.9) are required for exoatmospheric applications. A consequence is that it takes considerable total booster weight to propel even small payloads to near-orbital speeds. More precisely, it can be shown (8) that the weight of the propellant required for a single-stage booster to impart a speed change AV to a payload weighing Wpayload is given by

⎡

Wprop = Wpayload mf ⎣

exp

V gIs p

−1

1 − (1 − mf )exp

V gIs p

⎤

⎦

(17)

where g denotes the acceleration from gravity near the Wprop def denotes an surface of the Earth and mf = Wprop + Wstructure (approximate) fuel-mass fraction that neglects the weight

Missile Guidance

of the payload Wpayload . Staging can be used to reduce total booster weight for a given fuel-speciﬁc impulse Isp and (approximate) fuel-mass fraction mf . Efﬁcient propellant expenditure for exoatmospheric intercepts is addressed within Reference 51. Three-dimensional mid-course guidance for SAMs intercepting nonmaneuvering high-altitude ballistic targets is addressed within Reference 52. Neural networks are used to approximate (store) optimal vertical guidance commands and estimate tgo. Feedback linearization (39) is used for lateral guidance commands. Acceleration Limitations Endoatmospheric missile acceleration is limited by altitude, speed, structural, stall AOA, and drag constraints— stall AOA at high altitudes and structural limitations at low altitudes (see Eq. 9). Exoatmospheric interceptor acceleration is limited by thrust-to-weight ratios and ﬂight time—the latter is because, when the fuel is exhausted, exoatmospheric missiles cannot maneuver. For the “ﬂying cylinder” considered earlier, the lateral accelerA QSref CL 0.5ρVm2 Sref ation A in gees is given by = = [α + g W W Splan 2 α ] (8). For L = 20 ft, D = 1 ft, W = 100 1bs, Vm = 1.5 Sref 3000 ft/s, and α = 20 deg, at an altitude of 25,000 ft, the resulting acceleration is A ≈ 20g. THAAD Systems Recent research efforts have focused on the development of THAAD systems. Calculations show that high-altitude ballistic intercepts are best made head-on so that there is little target deceleration perpendicular to the LOS (8), because such decelerations appears as a target maneuver to the interceptor. EKF methods have been suggested for estimating target ballistic coefﬁcients and state information to be used in OG laws. Estimating ballistic coefﬁcients W def (β = where CDO is the zero lift drag coefﬁcient) is Sref CDa particularly difﬁcult at high altitudes where there is little 1 drag adrag = ρgVm2 . Also, the high closing velocity of a bal2β listic target engagement signiﬁcantly decreases the maximum permitted guidance system bandwidth for radome slope stability. Noise issues can also signiﬁcantly exacerbate the ballistic intercept problem. FUTURE DEVELOPMENTS Future developments will focus on theater-class ballistic missiles, guided projectiles, miniature kill vehicles, space-based sensors for missile defense and boost-phase interceptors. The Age of Air-Breathing Hypersonic Flight During the Gulf Wars, it often took considerable time to get a missile on a critical target (e.g., Iraqi leadership), which gave further impetus for a prompt global strike (PGS) capability—one that permits accurate strikes across thousands of miles in minutes. Many have suggested the

13

retroﬁtting of Trident missiles with conventional warheads for this purpose. This idea has alarmed many who argue that such an application of Trident ICBMs could mistakingly unleash a world-impacting nuclear war. Others have proposed the development of hypersonic missiles that exploit new scramjet technology. In 2004, NASA’s scramjet powered X-43A vehicle ushered in the age of air-breathing hypersonic ﬂight. Two history-making ﬂights were made—one at Mach 7 and the other at Mach 10. [At sea level, 2116.2 lb/ft2 , 59◦ F (standard atmosphere conditions), the speed of sound is 1116.5 ft/second (761.25 mph).] These historical ﬂights unleashed a hypersonics research revolution—one that has already begun to signiﬁcantly shape the design of future aircraft, missile, and space-access systems. Like the X-43A, hypersonic missiles are expected to exploit rocket power to achieve hypersonic ﬂight (∼Mach 5) at which point the scramjet will take over. Challenges impacting the development of hypersonic vehicles include signiﬁcant operational uncertainty and aero-thermo elastic-propulsion interactions. At very high speeds, the heat generated is so severe that classic aerodynamic principles based on ﬂuid mechanics no longer apply. In such a case, gas theory must be used to predict lift and drag properties. A consequence is signiﬁcant aerodynamic uncertainty. The high temperatures induced also result in severe aero-elastic effects (e.g., servoelastic) that make control difﬁcult. Such issues are currently being addressed by the research community. It is truly amazing how, in just over 100 years since the ﬁrst powered Wright Brothers ﬂight on December 17, 1903, we have ushered in the age of air-breathing hypersonic ﬂight.

ACKNOWLEDEMENTS This research has been supported, in part, by a 1998 White House Presidential Excellence Award from President Clinton, the National Science Foundation (NSF), NASA, Western Alliance to Expand Student Opportunities (WAESO), Center for Research on Education in Science, Mathematics, Engineering and Technology (CRESMET), and a Boeing A.D. Welliver Faculty Fellowship. For additional information, please contact [email protected] or visit http://www.eas.asu.edu/ aar/research/mosart/mosart.html.

BIBLIOGRAPHY 1. Neufeld, M. J., The Rocket and the Reich; Harvard University Press: Cambridge, MA, 1995. 2. Fossier, M. W., The Development of Radar Homing Missiles. J. Guidance Cont. Dynamics, 1984, 7,pp 641–651. 3. Haeussermann, W., Developments in the Field of Automatic Guidance and Control of Rockets. J. Guidance Contr. Dynamics 1981, 4, 225–239. 4. Blakelock, J. H., Automatic Control of Aircraft and Missiles, McGraw-Hill: New York, 1992. 5. Williams, D. E., Friedland, B., Madiwale, A. N., Modern control theory for design of autopilots for Bank-to-Turn Missiles, J. Guidance Contr., 1987, 10,pp 378–386.

14

Missile Guidance

6. Lin, C. F., Modern Navigation, Guidance, and Control Processing; Englewood Cliffs, NJ, Prentice-Hall: 1991. 7. Khou, K., Doyle, J. C., Essentials of Robust Control; PrenticeHall: Upper Saddle River, NJ, 1998. 8. Zarchan, P., Tactical and Strategic Missile Guidance; AIAA Inc.: New York, 1990. 9. Macknight, N., Tomahawk Cruise Missile; Motorbooks International: New York, 1995. 10. Arrow, A., An Analysis of Aerodynamic Requirements for Coordinated Bank-to-Turn Missiles. NASACR 3544, 1982. 11. Feeley, J. J., Wallis, M. E., Bank-to-Turn Missile/Target Simulation on a Desk Top Computer, The Society for Computer Simulation International, 79–84, 1989. 12. Reidel, F. W., Bank-to-Turn Control Technology Survey for Homing Missiles. NASAR 3325, 1980. 13. Kovach, M. J., Stevens, T. R., Arrow, A., A Bank-to-Turn Autopilot Design for an Advanced Air-to-Air Interceptor; Proc. of the AIAA GNC Conference; Monterey, CA, August 1987. 14. Rodriguez, A. A., Cloutier, J. R., Performance Enhancement for a Missile in the Presence of Saturating Actuators, AIAA J. Guidance Cont. Dynamics 1996, 19:pp 38–46. 15. Rodriguez, A. A., Yang, Y., Performance Enhancement for Unstable Bank-to-Turn (BTT) Missiles with Saturating Actuators. Int. J. Control. 1996, 63pp 641–678. 16. Rodriguez, A. A., Monne, M., Evaluation of Missile Guidance and Control Systems on a Personal Computer. J. Soc. Comput. Simul. 1997, 68pp. 363–376. 17. Ridgely, D. B.; McFarland, M. B., Tailoring Theory to Practice in Tactical Missile Control. IEEE Contr. Syste. Mag., 1999 19,pp 49–55. 18. Cloutier, J. R., Evers, J. H., Feeley, J. J., Assessment of Airto-Air Missile Guidance and Control Technology, IEEE Contr. Syst. Mag., 1989, pp 27–34. 19. Riggs, T. L., Vergaz, P. L., Advanced Air-to-Air Missile Guidance Using Optimal Control and Estimation.AFATL-TR-8156, Air Force Armament Laboratory, Eglin AFB, FL. 20. Nesline, F. W., Zarchan, P., A New Look at Classical versus Modern Homing Guidance. J. Guidance Contr. 1981, 4,pp 78–85. 21. Jark, J., Kabamba, P. T., Miss Distance Analysis in a New Guidance Law; Proc. of the American Control Conference;June 1999, pp 2945–2949. 22. Waldmann, J., Line-of-Sight Rate Estimation and Linearizing Control of an Imaging Seeker in a Tactical Missile Guided by Proportional Navigation. IEEE Trans. on Contr. Syst. Technol., 2002, 10,pp 556–567. 23. Bryson, A. E., Ho, Y. C., Applied Optimal Control: Optimzation, Estimation, and Control; HPC: New York, 1975. 24. Shukla, U. S., Mahapatra, P. R., The Proportional Navigation Dilemma—Pure or True? IEEE Trans. Aerosp Electron. Syst., 1990, 26,pp 382–392. 25. Ghose, D., Dam, B., Prasad, U. R., A Spreader Acceleration Guidance Scheme for Command Guided Surface-to-Air Missiles; Proc. IEEE National Aerospace and Electronics Conference; 1989, pp 202–208. 26. Oh, J. H., Solving a Nonlinear Output Regulation Problem: Zero Miss Distance of Pure PNG. IEEE Trans. Automat. Contr. 2002, 47,pp 169–173. 27. Yang, C. D., Yang, C. C., Optimal Pure Proportional Navigation for Maneuvering Targets. IEEE Trans. Aerosp. Electron. Sys. 1997 33,pp 949–957.

28. Kecker, K., Closed form Solution of Pure Proportional Navigation, IEEE Trans. Aerosp. Electron. Syst. 1990, 26,pp 526– 533. 29. Axelband, E., Hardy, F., Quasi-Optimum Proportional Navigation. IEEE Transa. Automat. Contr. 1970, 15,pp 620– 626. 30. White, B. A.; Rbikowski, R.; Tsourdos, A., Aim Point Guidance: An Extension of Proportional Navigation to the Control of Terminal Guidance; Proc. American Control Conference;June 2003, pp 384–389. 31. Yuan, P. J., Hsu, S. C., Solutions of Generalized Proportional Navigation with Maneuvering and Nonmaneuvering Targets, IEEE Trans. Aerosp. Electron. Syst., 1995, 31,pp 469–474. 32. Yang, C. D., Cang, C. C., Analytical Solution of Generalized 3D Proportional Navigation, Proc. 34th IEEE Conference on Decision and Control;December 1995, pp 3974–3979. 33. Yang, C. D., Hsiao, F. B., Yeh, F. B., Generalized Guidance Law for Homing Missiles. IEEE Trans. Aerosp. Electron. Syst., 1989, 25,pp 197–212. 34. Chakravarthy, A., Dhose, D., Capturability of Realistic Generalized True Proportional Navigation. IEEE Trans. Aerosp. Electron. Syst., 1996, 32,pp 407–418. 35. Yang, C. D., Yang, C. C., A Uniﬁed Approach to Proportional Navigation. IEEE Trans. Aerosp. Electron. Syst., 1997, 33,pp 557–567. 36. Tyan, F., An Uniﬁed Approach to Missile Guidance Laws: A 3D Extension, Proc. American Control Conference; 2002, pp 1711–1716. 37. Aggarwal, R. K., Optimal Missile Guidance for Weaving Targets. Proc. 35th IEEE Decision and Control; 1996, 3, pp 2775–2779. 38. Zarchan, P., Tracking and Intercepting Spiraling Ballistic Missiles; IEEE Position Location and Navigation Symposium;March 2000, pp 277–284. 39. Shima,T., Golan, O. M., Bounded Differential Games Guidance Law for a Dual Controlled Missile; Proc. American Control Conference;June 2003, pp 390–395. 40. Khalil, H., Nonlinear Systems, 2nd ed.; Prentice Hall: Englewood Cliffs, NJ, 1996. 41. Vincent, T. L., Worgan, R. W., Guidance against Maneuvering Targets using Lyapunov Optimization Feedback Control, Proc. American Control Conference;May 2002, pp 215–220. 42. Zouan, Z.; Yunan, H.; Wenjin, G., Lyapunov Stability Based Three-Dimensional Guidance for Missiles Against Maneuvering Targets; Proc. 4th World Congress on Intelligent Control and Automation;June 2002, pp 2836–2840. 43. Manchester, I. R.; Savkin, A. V., Circular Navigation Guidance Law for Precision Missile/Target Engagements; Proc. 41st IEEE Conference on Decision and Control;December 2002, pp 1287–1292. 44. Balakrishnan, S. N.; Stansbery, D. T.; Evers, J. H.; Cloutier, J. R., Analytical Guidance Laws and Integrated Guidance/Autopilot for Homing Missiles. Second IEEE Conference on Control Applications;September 1993, pp 27–32. 45. Shamma, J. S., Cloutier, J. R., Existence of SDRE Stabilizing Feedback. IEEE Trans. Automat. Contr. 2003, 48,pp 513– 517. 46. Tahk, M. J.; Ryoo, C. K.; Cho, H., Recursive Time-to-Go Estimation for Homing Guidance Missiles. IEEE Trans. Aerosp. Electron. Syst. 2002, 38,pp 13–24. 47. D’Souza, C. N.; McClure, M. A.; Cloutier, J. R., Spherical Target State Estimators; Proc. American Control Conference;June 29–July 1, 1994, 1675–1679.

Missile Guidance 48. Rago, C.; Mehra, R. K., Robust Adaptive Target State Estimation for Missile Guidance Using the Interacting Multiple Model Kalman Filter; IEEE 2000 Position Location and Navigation Symposium;March 2000, pp 355–362. 49. Jahk, M. J.; Ryu, H.; Song, E. J., Observability Characteristics of Angle-Only Measurement Under Proportional Navigation; Proc. 34th SICE Annual Conference; International Session Papers,July 1995, pp 1509–1514. 50. Williams, D. E.; Briedland, B., Target Maneuver Detection and Estimation; Proc. 27th IEEE Conference on Decision and Control;December 1988, pp 851–855. 51. Brainin, S.; McGhee, R., Optimal Biased Proportional Navigation. IEEE Trans. Automat. Contr., 1968, 13,pp 440–442. 52. Song, E. J.; Tahk, M. J., Three-Dimensional Midcourse Guidance Using Neural Networks for Interception of Ballistic Targets, IEEE Trans. Aerosp. Electron. Syst. 2002 38,pp 404–414.

ARMANDO A. RODRIGUEZ Arizona State University

15

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering c 1999 John Wiley & Sons, Inc. Copyright

PHASED ARRAY ANTENNAS A receiving antenna is a spatial sampler of the incident electric field. It provides an estimate of the incident field by performing a spatial averaging or integration of the impinging electric fields along the physical length of the antenna. The purpose of a complex antenna element or an antenna array is to carry out spatial filtering, that is, to select the incident fields arriving from some preferred directions or having some spatiotemporal properties and to zero out the response along other directions. Two of the most prevalent applications of spatial filtering are radio direction finding and adaptive processing. In radio direction finding the objective is to estimate the angle of arrival and intensities of the various signals by measuring a set of induced voltages in the elements of an antenna array. So from the given physical dimensions and geometry of the array and the measured noise-contaminated voltages induced in the antenna elements, the goal is to estimate both the directions of arrival and strengths of the various signals. In Fig. 1 an idealized array of omnidirectional isotropic point sources is shown. This assumption will be relaxed later in the presentation. The goal in radio direction finding is, given the induced complex voltages V i at each of the antenna elements, to estimate their complex signal strengths and directions of arrival. In adaptive processing, the objective is to extract a desired signal of interest buried in various multipaths or reflections of the signal, strong interference, and thermal noise. Generally, it is assumed that the direction of arrival of the signal is known whereas the angles of arrival of the undesired signals are not known. It is desired to extract a signal of interest (SOI) from the measured complex voltages V i at the antenna elements, which are the sum of the signal of interest and other undesired signals, including noise. Always, however, in this process something about the signal is assumed to be known (1). This is the prerequisite for any adaptive system. The a priori information may be available in any of the following forms: (1) the direction of arrival of the signal, (2) some special temporal characteristics like cyclostationarity or constant modulus, or (3) some statistical information, such as that the desired signal and the undesired signals are statistically independent.

Stochastic Versus Direct-Data-Domain Approach Historically, almost all algorithms in radio direction finding and adaptive processing have used algorithms that are based on stochastic methodology so that the algorithm performs in an optimum sense on the average. With the advent of the digital revolution, classical algorithms applied to analog systems are still being used in digital form. The application of a direct-data-domain method is therefore very attractive, as it allows one to obtain an optimum solution for the data on hand and not on the average for all the data sets. In addition, the application of a direct-data-domain approach is computationally very efficient, as no covariance matrix of the data is needed. In most algorithms it is assumed that the covariance matrix of the data is given; however, in reality it is only the data themselves that are available. Hence one tries to approximate the covariance matrix from the data sets. Not only is that a very computation- intensive process, but also, it is difficult to estimate the error that is incurred in this approximation. In addition, a direct-data-domain least-squares approach has a lower Cramer–Rao bound for the parameters of interest than a stochastic methodology (2). 1

2

PHASED ARRAY ANTENNAS

Fig. 1. An array of omnidirectional isotropic point sources.

A case is made here for applying the methodology directly to the data set at hand, which provides a direct least-squares solution based on the available data without making any assumptions of the underlying statistics (3). In this review we will introduce the computationally efficient and numerically robust direct-data-domain methodology for both radio direction finding and adaptive processing. Another ramification of this approach is that it is quite straightforward to allow for mutual coupling between the sensors collecting the data. This issue is addressed later on.

Radio Direction Finding In radio direction finding the technique is to estimate the complex intensities of the various signals incident on the antenna elements from complex signal voltages measured at the terminals of the antenna. There are several assumptions that are involved in the processing. First, the sources of the signals are assumed to be located in the far field. Far field is defined as a distance greater than 2d2 /λ, where d is the largest physical dimension of the antenna array and λ is the wavelength of the signal. The far-field assumption implies that the wavefront of the signal arriving at the array is essentially planar. Another assumption that is generally used is that the antennas are omnidirectional point radiators so that there is no mutual coupling between the elements. However, this is a highly idealized situation. It is never true in real life, and hence, unless the electromagnetic effects (from solution of Maxwell’s equations) are taken into account, use of pure signal-processing algorithms will not provide meaningful results in a real environment. First, we present the methodology for estimating the direction of arrival for idealized omnidirectional point sources. Then, the effects of mutual coupling, when using realistic antennas, can be allowed for by electromagnetic preprocessing of the data as outlined in the section on adaptive processing below. Case A: Uniform Linear Array (Antenna Elements Are Uniformly Spaced). When the omnidirectional point-source antenna elements are uniformly spaced, the complex voltage V n induced in an antenna element n is given by

where dn is the location of the nth antenna element, φi is the direction of arrival of the signal from the end-fire direction as shown in Fig. 1, and Ai is its complex amplitude. P is the total number of signals incident on the array and needs to be determined. For a uniformly spaced array dn = nd, where d = χ is the interelement spacing (as per Fig. 1). Here we have used a single snapshot, i.e., the phasors V n are measured across the entire

PHASED ARRAY ANTENNAS

3

array at a single time instant. It is further stipulated that all P signals are narrowband and the wavelength of transmission is λ. So the goal here is to estimate the 2P unknowns of Ai and φi from the measured voltages V n . As long as there are 2P antenna elements, the problem can be solved by fitting a sum of complex exponentials to the voltages V n . This is computed through the matrix-pencil approach (4,5,6), which is very robust when applied to noisy data. Of course, in a real situation there is noise in the data and hence we need more than 2P antenna elements. The conventionally used methodology of ESPIRIT (7,8) requires the formation of a covariance matrix, which is computationally more intensive than the Matrix-Pencil Technique. From a statistical point of view both the methods have similar variances for the estimates in presence of noise. However, it is important to note that additional processing is required, as in ROOT-MUSIC, where actual directions of arrival are obtained, which involves factoring a high-order polynomial to estimate the strengths of the various signals.

Case B:

Nonuniformly Spaced Elements (Antenna Elements Are Nonuniformly Spaced).

When the antenna elements are spaced nonuniformly, then clearly the above approach based on a single snapshot is not applicable. The processing is to be done in the time domain. In that case one uses the model

where dn is the location of the nth antenna element, f 0 is the frequency of transmission, and φi is the phase associated with the ith incident field. Therefore, Ai is considered to be real. It is important to note in this scenario that if there are coherent multipaths (i.e., the signal and an undesired multipath component of the signal are in phase), a nonuniformly spaced array cannot separate them from a single snapshot without any additional processing. Temporal information is also necessary (9). The various components can be extracted using MUSIC (7) and many of its derivatives, and also ESPIRIT (7) may be used for a certain class of array geometries.

Adaptive Processing Using The Spatial-Domain Least-Squares Approach In the conventional adaptive beam-forming methodology each antenna element is weighted. The processing of information is done over time as the correlation matrix R of the data needs to be formed (10). In the current development we deal with a single frame or a single snapshot. A single snapshot is defined as the set of complex voltages V n measured at each one of the N + 1 antenna elements at a particular instant of time. These measured voltages V n , n = 0, 1, . . ., N, contain the desired signal plus jammer, clutter, and thermal noise components. Hence, in this development one can allow for blinking jammers, time-varying clutter, and coherent multipath components. The price one pays for dealing with a snapshot (frame) is that the number of degrees of freedom is limited to N/2, as opposed to N + 1 in the covariance-matrix-based approaches. However, this limitation is alleviated by doubling the available data, as illustrated later (11,12,13,14). In this model we double the number of data by not only considering them in the forward direction but also conjugating them and reversing the direction of increment of the independent variable; we call this the backward method [1]. This type of processing can be done as long as the series to be approximated can be fitted by exponential functions of purely imaginary argument. This is always true for the adaptive-array case. So by considering the data set x(k) and x∗(−k) we have essentially doubled the amount of data without any penalty, as these two data sets for our problem are linearly independent. Next we use these data to find the adaptive weights, which are related to the directions of arrival of the jammers. Often, due to uncertainties in the direction of arrival of the SOI, there may be signal cancellation in the adaptive process. The expected signal (target return) may not arrive from a single predetermined angle, but, due to refractions of the atmosphere, arrive over a finite angle extent. In addition, there is always a mismatch

4

PHASED ARRAY ANTENNAS

between the look-direction constraint and the true direction of arrival of the desired signal. Correction for this uncertainty is accomplished in the least-squares procedures by establishing look-direction constraints at multiple angles of the adaptive receiver pattern within the transmitter main-beam extent. The multiple constraints are established by using a uniformly weighted array pattern for the same-size array as the adaptive array under consideration. Multiple points of constraints in the received adaptive beam pattern to be formed are chosen for the nonadapted array pattern, and a row corresponding to each constraint is implemented in the matrix equations presented below:

Here Zi represent the various constraints along specific look directions of: the receiver beam pattern and are defined by [Zi = exp (j2τd/λ) cos φi ]. Z0 corresponds to the SOI. Here X i are the actual voltages measured at the ith antenna element due to SOI, jammer, clutter, and thermal noise. W i are the adaptive weights, and Ci are numerical prefixed values of the constraints imposed on the adapted beam to be formed. Let L be the number of look-direction constraints, and M + 1 be the number of weights to be calculated. Therefore M −L + 1 is the number of jammers that can be nulled. The first L + 1 equations in Eq. (3) define the main beam constraints of the adapted receiver pattern. The remaining equations use data from the N + 1 elements, and each entry computes the difference between neighboring elements, thereby canceling the SOI and hence containing only undesired signal components. The number of equations must equal the number of weights, and therefore M = L + N − M. This leads to the relationship N = 2M − L between the number of weights, number of constraints, and number of elements. Using the forward–backward data from a single snapshot, the maximum number of weights or the degrees of freedom that can be achieved for a direct-data-domain approach is approximately N/1.5 + 1, as opposed to N

PHASED ARRAY ANTENNAS

5

+ 1 for the conventionally used statistical method. So there is a slight loss in the degrees of freedom. However, we gain the ability to deal with a highly nonstationary environment where the signal environment may change even from a snapshot to snapshot and thereby allow for blinking jammers. In a phased array, the angle extent of the received beam is established by the main beam of the transmitted wave (usually between the 3 dB points of the transmitted field pattern). Target returns within the angle extent must be coherently processed, but with the appropriate steering vector. In that case, the excitation function Y [right-hand side of Eq. (3)] would have several nonzero elements, depending on the number of constraints used for the main beam. This is called a multiple-constraint receive beam pattern, as opposed to constraining it at a single point based on the assumed direction of arrival of the signal of interest. The advantage of dealing with multiple constraints as opposed to a single constraint in the main beam is illustrated next. Consider a 21-element array with N = 20 and M = 11. The beam is considered to be pointed broadside (θ = 90◦ ), and target returns can be expected over the main beam out to the 3 dB points (±5◦ ). For the broadsidepointed array, consider a target located in the main beam at θ = 94◦ instead of θ = 90◦ . The target signal-to-noise ratio at each element is 20 dB, and we assume no jammers or clutter present. Figure 2 shows the main-beam region of the antenna pattern after adaptation. Since the target is not at the look-direction constraint point (i.e., θ = 90◦ ), the adaptive process considers it as an interfering source and attempts to null it. Because the target is relatively near to the look-direction constraint, the process is not able to form a perfect null. Figure 3 shows the complex array gain along the target direction for 10 random samples of the noise. The point × represents the nonadapted array gain in the target direction. Note that the gain in the target direction is reduced in each case. In addition, there is a wide variation in the array gain from one random sample to the other. Now, if one were to process the returns from different pulses in a pulse burst that was to be coherently integrated, then this variation in the received signal would have significant influence on that integration. We now illustrate how to overcome it. We establish a multiple constraint on the receive pattern as shown in Fig. 4 at 85◦ , 87.5◦ , 90◦ , 92.5◦ , and ◦ 95 . So the receive signal would not be nulled if it were located anywhere within the 10◦ beam width. For this particular case, the excitation vector Y would be of the form Y T = [13, 7.72 + j8.32, 7.72 − j8.32, − 0.816 + j7.149, − 0.816 − j07.149, 0, 0, 0, 0, 0, 0, 0, 0]. The corresponding receive beam pattern with the five constraints is shown in Fig. 4. We now consider the same example as before. However, as seen from Fig. 5 (using the same data from 10 random samples of noise), there is no reduction of the array gain along the direction of the target, and for all the ten runs the array gain vectors are very nearly aligned. The five-constraint approach permits effective radar processing across the main beam’s extent with no effect of the loss of gain in the target direction. The adaptive process has been prevented from nulling the target. In summary, the main-beam constraint allows the look-direction constraint to be established over a finite beam width while maintaining the ability to adaptively null jammers in the side-lobe region. Although the main-beam gain can become degraded if the signal becomes very strong, this does not appear to be a serious limitation for practical radar-processing.

Incorporation Of Mutual Coupling In Adaptive Antennas To illustrate the importance of mutual coupling between antenna elements in a phased array, we consider signal recovery by a linear array of equispaced, thin, half-wavelength dipoles as shown in Fig. 6. However, in this analysis the antennas can be any complex composite element. The method of moments (MOM) is used to perform the electromagnetic analysis of the antenna array. Using a Galerkin formulation, the entries of the MOM impedance matrix measure the interaction between the basis functions, i.e., they quantize the mutual coupling (15,16). The electromagnetic analysis needs to be coupled with the signal-processing algorithms in order to generate accurate and reliable results (17,18). These ideas are explained through an array of N e uniformly spaced isotropic point sensors as shown earlier in Fig. 1. The array receives a signal (called S) from a

6

PHASED ARRAY ANTENNAS

Fig. 2. Main-beam array gain with a strong target at 94◦ .

known direction φ0 and some interference sources (called J i ) from unknown directions. In the absence of mutual coupling, each individual source presents a linear phase progression across the face of the array. Therefore, the voltage at the i-th element due to the incident fields is

where, um = cos φm , S is the complex intensity of the signal incident from direction φ0 , J m is the intensity of the mth interference source arriving from direction φm , and ni is the additive noise at each element. Let β = exp(jk x u0 ) represent the phase progression of the signal between one element and the next. Hence, the term V i − β − 1 V i+1 has no signal component. This is illustrated through the last K equations of Eq. (3), where, K = (N e + 1)/2. The last K − 1 rows of the matrix contain only interference and noise terms. Setting the product of these terms with the weights to zero nulls the interference in a least-squares sense. The equation represented by the first few rows constrains the gain of the array along the direction of the signal. It can be shown that if M + 1 ≤ K, the signal can be recovered and

PHASED ARRAY ANTENNAS

7

Fig. 3. Complex array gain: one constraint.

Fig. 4. Uniformly weighted array pattern with the location of the five constraints.

It is important to point out that there may be signal cancellation if the actual direction of arrival of the signal of interest is slightly different from the assumed direction of arrival. However, this can be avoided by selecting the first row of the matrix and replacing it by placing an a priori 3 dB constraint on the receive beam width of the adaptive pattern as the optimization process progresses. This prevents signal cancellation when there is uncertainty in the direction of arrival. Let us consider a signal corrupted by three jammers that are incident on the array. To focus only on the effects of mutual coupling, it is first assumed that there is no mutual coupling between the antenna elements

8

PHASED ARRAY ANTENNAS

Fig. 5. Complex array gain: five constraints.

Fig. 6. A realistic adaptive antenna array.

and the voltages at the ports of the array is given by Eq. (4). These voltages are then passed to the signal recovery subroutine to find the adaptive weight using Eq. (3), and the signal is estimated using Eq. (5). Next, we consider a realistic antenna array as shown in Fig. 6, where each wire antenna element is centrally loaded with an impedance. The details of the chosen array are presented in Table 1 and illustrated in Fig. 5. The receiving algorithm tries to maintain the gain of the array in the direction of φ0 = 45◦ while automatically placing nulls at in the interference directions. All signals and jammers arrive from the elevation = 90◦ . The base signal and jammer

PHASED ARRAY ANTENNAS

9

intensities and directions of arrival φi are given in Table 2. In all simulations the jammer intensities, the directions of arrival of the jammers, and the signal intensity are used only to find the voltages input to the receiving algorithm. The receiving algorithm itself uses only the direction of arrival of the signal; that is, only the look direction is considered to be known. The signal is kept constant at 1.0 V/m as given in Table 2. The intensity of the first jammer, arriving from φ = 75◦ , is varied from 1.0 V/m (0 dB with respect to the signal) to 1000.0 V/m (60 dB) in steps of 5 V/m. If the jammers are properly nulled, we expect the reconstructed signal to have no residual jammer component. Therefore, as the jammer strength is increased, we expect the reconstructed signal to remain constant. Figure 7(a) presents the results on the magnitude, and Fig. 7(b) on the phase, for the adapted signal using the receiving algorithm when mutual coupling is absent and the antenna array is considered to be an ideal one as shown in Fig. 1. The magnitude of the reconstructed signal is indistinguishable from the expected value of 1.0 V/m. This figure demonstrates that, in the absence of mutual coupling, the receiving algorithm is highly accurate and can still null a strong jammer. Figure 8(a,b) show the results for the magnitude and phase, respectively, of the received signal when using the measured voltages that are affected by mutual coupling. Here, the array consists of seven wires. The magnitude of the reconstructed signal varies approximately linearly with respect to the intensity of the jammer. This is because the strong jamming is not nulled and the residual jammer component completely overwhelms the signal. The reason the signal cannot be recovered when mutual coupling is taken into account can be understood visually by comparing the adapted beam patterns in the ideal case of no mutual coupling with the case where mutual coupling is present. In Fig. 9(a) we see the beam pattern for the ideal case. The pattern clearly displays the three deep nulls at the directions of the interference. The high side lobes are in the region where there is no interference. Because of the deep nulls, the strong interference can be completely nulled and the signal recovered correctly. Figure 9(b) shows the beam pattern when the mutual coupling is taken into account. As is

10

PHASED ARRAY ANTENNAS

Fig. 7. Signal recovery for an idealized array without mutual coupling.

clear, the gain of the antenna along the signal direction is considerably reduced. The pattern nulls are shallow and are displaced from the desired locations. The shallow nulls result in inadequate nulling of the interference; hence the signal cannot be recovered. The receiving antenna is next assumed to be a linear array of N e elements as illustrated in Figure 6. The elements are parallel, thin equispaced dipoles. Each element of the array is identically point loaded at the center. The dipoles are x-directed, of length L and radius a, and are placed along the x axis, separated by distance x. The array lies in the X–Z plane.

PHASED ARRAY ANTENNAS

11

Fig. 8. Signal recovery for a realistic array in the presence of mutual coupling.

We begin by analyzing the response of the antenna array to an incident field Einc . Since the array is composed of thin wires, the following simplifying assumptions are valid (15,16): (1) The current flows only in the direction of the wire axes (here the z direction). (2) The current and charge densities on the wire are approximated by filaments of current and charge on the wire axes (which lie in the y = 0 plane). (3) Surface boundary conditions can be applied to the relevant axial component of the wire axes. The integral equation that characterizes the current on the wires and describes the behavior of the array is (15,16)

We solve this equation using the method of moments to obtain the MOM impedance matrix. The basis functions used are piecewise sinusoids as described in Ref. 15 and shown in Fig. 10. P (chosen odd) basis functions are used per element. Using these basis functions and a Galerkin formulation, Eq. (6) is reduced to

12

PHASED ARRAY ANTENNAS

Fig. 9. Antenna beam pattern, (a) for an idealized array without taking into account mutual coupling in the analysis, and (b) for a realistic array.

the matrix equation

where I is the MOM current vector containing the coefficients of the expansion of the current in the sinusoidal basis, V is the MOM voltage vector representing the inner product of the weighting functions and the incident field, and Z and Y are the MOM impedance and admittance matrices respectively. Both matrices are of order N × N, where N = N e P is the total number of unknowns used in the MOM formulation.

PHASED ARRAY ANTENNAS

13

Assuming that the incident field is linearly polarized and arrives from direction (θ, φ), it can be written in the functional form as

where k = −k( cos φ sin θ + sin φ sin θ + cos φ) is the wave vector associated with the direction of arrival of the incident signal. Using P(odd) basis functions on each antenna, the current on the structure can be written as

where f p,n (z) is the p-th basis function on the nth element whose functional form is given by

where z = L/(P + 1) and zp,n = z0,n + pz. z0,n is the z-coordinate of the bottom of the n-th antenna as shown in Fig. 10. Substituting Eq. (9) in (6) and using testing functions f q,m (z), the entries of [V] are given by

where xm is the x-coordinate of the axis of the m-th antenna. For the impedance matrix [Z] the elements are given by

14

PHASED ARRAY ANTENNAS

Fig. 10. Basis functions assumed in the electromagnetic analysis using the MOM analysis.

with

For the case m = n, i.e., both subsections i and l are on the same antenna element, the term (xm − xn ) is set to a, the radius of the wire (15). An analytic expression for the entries of the MOM impedance matrix is derived in Ref. 15. Because of the choice of a piecewise sinusoid basis and the choice of an odd number of basis functions per antenna element, only one basis function is nonzero at the port. This is illustrated in Fig. 10, where the basis function ZL is the only one contributing to the current at the port. Therefore, the measured voltage at the port of the nth antenna is given by

i.e., the measured voltage at a port of the array is directly proportional to the coefficient of the basis function corresponding to the port. The MOM analysis results in a matrix equation that relates the coefficients of the current expansion to the MOM voltages through the admittance matrix. Since the MOM impedance and admittance matrices are independent of the incident fields, they can be evaluated a priori. The measured

PHASED ARRAY ANTENNAS

15

voltages at the ports of the antenna are related to the current coefficients by Eq. (10). Using this equation and Eq. (7), the N e -dimensional vector of measured voltages can be written as

where ZL is the N e × N e diagonal matrix with the load impedances as its entries, Y port is the matrix with the rows of Y that correspond to the ports of the array, [V] is the MOM voltage vector of order N, i.e., the number of unknowns in the MOM analysis, and Y port is a rectangular matrix of order N e × N with N > N e . Since Y port is a rectangular matrix with more columns than rows, Eq. (9) represents an underdetermined system of equations. Our goal is to estimate some part of V given V meas . Therefore, we need a method to collapse the N e × N matrix Y port to an N e × N e matrix. The proposed method is most easily understood when illustrated with an example. If P unknowns are used per wire element, N = N e P. Consider the case with N e = 2 and P = 3. Then N = 6, and basis function 2 corresponds to the port on the first element, while basis function 5 corresponds, to the port on the second element. In this case, Eq. (11) can be written as

If the signal and all the jammers are incident from approximately the same elevation θ, the entries in V are not all independent of each other. From Eq. (7), if weighting functions i and i + 1 belong to the same array element,

Letting α = ejkzcosθ , we have

16

PHASED ARRAY ANTENNAS

Therefore, Eq. (12) can be reduced to

where V is the vector of length N e whose entries are the MOM voltages that correspond to the ports, and B is the N e × N e matrix that relates the measured voltages to V . Equation (18) is a relation between the measured voltages and the MOM voltages that correspond to the ports of the array. In a practical application, the measured voltages are the given quantities and are affected by mutual coupling. The MOM voltages on the right-hand side of Eq. (18) are the voltages that are directly related to the incident fields and so are free from the effects of mutual coupling. Both vectors are of order N e , the number of ports. Therefore, this equation can easily be solved for the MOM voltages corresponding to the ports of the antenna. Furthermore, if the elevation angle of interest (θ) is fixed, the matrix B can be evaluated a priori. Hence the computational cost of eliminating the mutual coupling is limited to the solution of a small matrix equation. The open-circuit voltages are the voltages that would be measured at the ports of the array if the ports were open-circuited. In Ref. 17 the authors assume that these voltages are free of the effects of mutual coupling. However, the open-circuit voltage at a particular element is the voltage measured in the presence of the other open-circuited elements. Therefore the effect of mutual coupling has been reduced but not eliminated. Mutual coupling can be assumed to have been eliminated only when there is nothing impeding the path of the incident fields—not even the array itself. We proceed with the same example presented earlier, where the intensity of the incident signal is held constant at 1.0 V/m. The intensity of the first jammer is varied from 1.0 V/m to 1000 V/m (60 dB above the signal) in steps of 5 V/m. For each value of the jammer intensity, the MOM voltage vector is calculated and the measured voltages are calculated. In the first scenario the measured voltages are used to find the open-circuit voltages. The open-circuit voltages are passed to the direct-data-domain algorithm of Ref. 18. In the second scenario Eq. (17) is used to find the voltage vector V . These voltages are used to recover the signal and null the jammers using the same algorithm. If the jammers are properly nulled, the reconstructed signal magnitude should remain constant as a function of jammer strength. Figure 11 presents the results when the open-circuit voltages are used to recover the signal. As can be seen, the recovered signal shows a near-linear relation to jammer strength. This indicates that the jammer has not been adequately nulled and the residual jammer strength has overwhelmed the signal. The results of compensating for the mutual coupling using the technique presented in this paper are shown in Fig. 12(a) for the magnitude and 12(b) for the phase. The magnitude of the reconstructed signal varies between 0.996 V/m and 1.004 V/m, that is, the error in the signal recovery is very small. This figure shows that the strong jammer has been effectively nulled and the signal can be reconstructed. The reason that the use of the open-circuit voltages is inadequate to compensate for the mutual coupling, while the technique presented here is adequate, is illustrated using the adapted beam patterns for the two cases. The adapted beam pattern associated with using the open-circuit voltages is shown in Fig. 13(a). The nulls are placed in the correct locations. However, they are shallow, resulting in inadequate nulling of the interference. The beam pattern associated with compensating for the mutual coupling using the technique presented in this paper is shown in Fig. 13(b). The nulls are deep and placed in the correct directions. This demonstrates that the mutual coupling has been suppressed enough to null even a strong jammer.

PHASED ARRAY ANTENNAS

17

Fig. 11. Signal recovery using the open-circuit voltages.

Figures 11 and 13 allow us to conclude that using the open-circuit voltages does reduce the effect of mutual coupling somewhat. However, the reduction is inadequate to suppress strong interference. This is because the open-circuit voltage at an array element is the voltage in the presence of the other open-circuited elements. The direct-data-domain technique along with the MOM presented proves to be far superior in compensating for mutual coupling. This is because by using multiple basis functions per antenna element, the mutual-coupling information has been represented accurately.

Effect Of Noise To illustrate the effect of thermal noise on the adaptive signal corrupted by three jammers as given in Table 3, we consider an array of z-directed dipoles that is centrally terminated by a 50 resistance. Seven unknowns per wire are used in the MOM analysis, leading to a total of 91 unknowns. The signal-to-noise ratio was set at 13 dB. Note that jammer 1 is a strong jammer (66 dB with respect to the signal). For each of the 13 antenna channels, a complex Gaussian random variable is added to the measured voltages due to the signal and jammers. This set of voltages, affected by noise, is passed to the signal recovery routine described earlier. The computational procedure is repeated 500 times with different noise samples. These 500 samples are used to evaluate the mean and the variance of the parameter of interest. The output signal to interference plus noise ratio (SINR) in decibels is defined as

The results of the above simulation are presented in Table 4. When the measured voltages are used directly to recover the signal, then—mainly due to the high bias in the estimate of the signal—the output SINR is only 6.35526 dB. The high bias can be directly attributed to the inadequate nulling of the strong jammer.

18

PHASED ARRAY ANTENNAS

Fig. 12. Signal recovery in a realistic array after taking mutual coupling into account: (a) magnitude, (b) phase.

However, when the mutual coupling is eliminated using the technique presented in this paper, the jammers are completely nulled, yielding accurate estimates of the signal. The total interference power is suppressed to nearly 20 dB below the signal. The examples presented here illustrate how one can effectively deal with the effects of mutual coupling between the sensors. Using the MOM with multiple basis functions per element allows us to reduce the mutual coupling to an extent where it becomes inconsequential. Hence, the effects of mutual coupling in the analysis have not been eliminated but rather taken into account.

PHASED ARRAY ANTENNAS

19

Fig. 13. Antenna beam pattern (a) using open-circuit voltages and (b) after allowing for the presence of mutual coupling.

Epilogue For the deployment of any realistic phased arrays, the electromagnetic nature of the array must be taken into account. We have shown that the mutual coupling between the elements of the array causes adaptive algorithms to fail. This problem is associated with both covariance-matrix approaches (as stated earlier in Ref. 17) and the direct-data-domain approach (investigated here). To properly characterize the antenna, the MOM is used. The use of multiple basis functions per element in a practical manner is a major advance and provides a pragmatic approach to the design of phased-array antennas. Recognizing that the MOM voltage vector is free from mutual coupling eliminates the mutual coupling from consideration. By using a relationship between the entries of the MOM voltage vector, a squarematrix equation is developed between the given measured voltages and the relevant entries of the MOM voltage vector. It is shown that this method works very well in the presence of strong interfering sources.

20

PHASED ARRAY ANTENNAS

Through a successful coupling of the electromagnetic analysis with the signal-processing algorithms used in radio direction finding and adaptive antennas, it is possible to make wide use of realistic phased-array antennas.

BIBLIOGRAPHY 1. T. K. Sarkar et al. A pragmatic approach to adaptive antennas, IEEE Antennas Propag. Mag., 42 (2): 39–55, 2000. 2. Y. Hua T. K. Sarkar A note on the Cramer–Rao bound for 2-D direction finding based on 2-D array, IEEE Trans. Signal Process. 39: 1215–1218, 1991. 3. S. Choi D. Shim T. K. Sarkar A comparison of tracking-beam arrays and switching-beam arrays operating in a CDMA mobile communication channel, IEEE Antennas Propag. Mag., 41 (6): 10–22, 1999. 4. Y. Hua T. K. Sarkar Matrix pencil method for estimating parameters of exponentially damped/undamped sinusoids in noise, IEEE Trans. Acoust. Speech Signal Process., 38: 814–824, 1990. 5. T. K. Sarkar O. Pererira Using the matrix pencil method to estimate the parameters of a sum of complex exponentials, IEEE Antennas Propag. Mag., 37 (1): 48–55, 1995. 6. F. Del Rio T. K. Sarkar Comparison between the matrix pencil method and the Fourier transform technique for high resolution spectral estimation, Digital Signal Process. Rev. J. 6 (2): 108–125, 1996. 7. P. Stoica R. Moses Introduction to Spectral Analysis, Englewood Cliffs, NJ: Prentice-Hall, 1997.

PHASED ARRAY ANTENNAS

21

8. A. Medouri et al. Estimating one- and two-dimensional direction of arrival in an incoherent/coherent source environment, IEICE Trans. Commun., E80-B (11): 1728–1740,1997. 9. T. K. Sarkar S. Nagaraja M. C. Wicks A deterministic direct data domain approach to signal estimation utilizing uniform 2D arrays, Digital Signal Process. Rev. J., 8 (2): 114–125, 1998. 10. R. A. Monzingo T. W. Miller Introduction to Adaptive Arrays, New York: Wiley, 1980. 11. T. K. Sarkar N. Sangruji An adaptive nulling system for a narrow-band signal with a look-direction constraint utilizing the conjugate gradient method, IEEE Trans. Antennas Propag., 37: 940–944, 1989. 12. T. K. Sarkar et al. A deterministic least squares approach to adaptive antennas, Digital Signal Process. Rev. J., 6 (3): 185–194, 1996. 13. S. Park T. K. Sarkar Prevention of signal cancellation in adaptive nulling problem, Digital Signal Process. Rev. J., 8 (2): 95–102, 1998. 14. S. Park T. K. Sarkar A deterministic eigenvalue approach to space time adaptive processing, Proc. IEEE Antennas and Propagation Soc. Int. Symp., 1996, pp. 1168–1171. 15. T. K. Sarkar B. J. Strait D. C. Kuo Special programs for analysis of radiation by wire antennas, Technical Report AFCRL-TR-73-0399, Syracuse University, June 1973. 16. A. R. Djordjevic et. al. Analysis of Wire Antennas and Scatterers: Software and User’s Manual, Norwood MA: Artech House, 1995. 17. I. J. Gupta A. A. Ksienski Effect of mutual coupling on the performance of adaptive arrays, IEEE Trans. Antennas Propag., 31: 785–791, 1983. 18. R. S. Adve T. K. Sarkar Estimation of the effects of mutual coupling in an adaptive nulling system with a look direction constraint, IEEE Trans. Antennas Propag., 48: 2000. 19. T. K. Sarkar B. J. Strait Optimization methods for arbitrarily oriented arrays of antennas in any environment, Radio Sci., 11 (12): 959–967, 1976.

TAPAN K. SARKAR RAVIRAJ ADVE University of Toronto MAGDALENA SALAZAR PALMA Polytechnic University of Madrid

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering c 1999 John Wiley & Sons, Inc. Copyright

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA The numerous applications of radars for the detection and identification of targets and in navigation and meteorology are well known. Its uses in remotely sensing the earth environment and planetary surfaces have received much attention more recently. The problem of interpreting the radar signals from complex structures makes it necessary to develop reliable analytical solutions to a wide class of realistic models of electromagnetic scattering problems in irregular stratified media. This article deals with the evaluation of the radar crosssections for irregular stratified models of radar targets. To this end, scattered electromagnetic fields are analyzed using a full wave approach. To obtain full wave solutions for the electromagnetic fields scattered by irregular stratified media with rough interfaces, iterative analytical and numerical approaches are used to solve generalized telegraphists’ equations developed for unbounded layered structures. Generalized telegraphists’ equations for the coupled waveguide mode amplitudes were first derived for irregular waveguides with finite cross-sections by S. A. Schelkunoff. For unbounded layered structures, the complete field expansions consist of the radiation term, as well as lateral waves and guided (surface) waves of the layered medium. Exact boundary conditions are imposed at each of the rough interfaces of the irregular layered structures. The full wave solutions are applicable to layered structures with inhomogeneous medium parameters and irregular boundaries with a broad range of roughness scales. They account for specular point scattering as well as diffuse (Bragg) scattering in a unified self-consistent manner. They have been used to evaluate the like- to cross-polarized bistatic radar scatter cross-section as well as the Mueller matrix elements that relate the incident to the scattered Stokes vectors. They have also been used to derive the observed like- and cross-polarized backscatter enhancements that are attributed to double scatter. The full wave approach can be used to evaluate the near-fields as well as the far-fields excited by small-scale (including subwavelength) and large-scale fluctuations in the surface height, complex permittivity, and permeability of the irregular stratified media. It is well established that the perturbation theory of Rice (1) can be used to correctly predict the copolarized and cross-polarized electromagnetic fields scattered by slightly rough two-dimensionally slightly rough surfaces that separate two semi-infinite media characterized by distinct electrical and magnetic properties. Perturbation theory is valid provided that the mean-square heights and slopes of the rough surfaces are of the same order of smallness; namely, β = 4k2 0 h2 1, σ2 s 1, and k0 h ≈ σs (where k0 is the free space electromagnetic wavenumber of the medium above the rough interface, h is the root-mean-square height, and σs is the root-mean-square slope of the rough surface). These analytical results, which have been extended to “tilted” slightly rough surfaces (2), have been validated experimentally and numerically (3) for both vertically and horizontally polarized excitations for arbitrary angles of incidence and scatter (general bistatic conditions). It is also generally assumed that the physical optics approach (4) (which is based on the Kirchhoff approximations of the surface electromagnetic fields) can be used to correctly predict the scattered electromagnetic fields from two-dimensionally random rough surfaces provided that the radii of curvature of the rough surface are very large compared to wavelength (namely, k2 0 ρ2 1, where ρ2 is the mean-square radius of curvature) and conditions for deep phase modulation between the scattered field contributions from the stationary phase (specular) points exist (5). Implicit in the physical optics approach is the requirement that the major 1

2

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

contributions to the scattered fields come from regions of the rough surface that are in the vicinity of the (stationary phase) specular points on the rough surface. For this reason, for example, the single scatter physical optics approach cannot be used to correctly predict the cross-polarized fields scattered in the plane of incidence. The physical optics approach also fails to correctly predict the polarization (vertical and horizontal) dependence of the backscatter cross sections when the surface slopes are very small even when the large radius of curvature criterion is satisfied. Thus, for perfectly conducting surfaces with very large radii of curvature, the Kirchhoff approximations for the surface current Js (Js = 2nxH i , where n is the unit vector normal to the rough surface and H i is the magnetic field of the incident electromagnetic waves) are correctly used to predict the physical optics co-polarized fields provided that the specular points contribute significantly to the scattered fields. If, however, the mean-square slopes are very small, such that for backscatter at oblique incidence no specular points exist, the physical optics solutions fail no matter how large the radii of curvature of the rough surface. Because of these limitations that are inherent in the two most familiar scattering theories, researchers have attempted to develop more rigorous scattering theories that can bridge the broad range of scattering problems not covered by either the perturbation theory or the physical optics approach. When the small slope and height criteria as well as the large radii of curvature criteria and conditions for deep phase modulation and specular point scattering are satisfied, the perturbation solutions and the physical optics solutions are in agreement with each other. When neither the perturbation nor the familiar physical optics solutions are individually applicable to the random rough surfaces considered (as in the case of microwave backscatter from the sea surface and the enhanced backscatter observed in controlled laboratory experiments), both the perturbation solutions and the physical optics solutions fail (6). This has provided the motivation to develop several versions of hybrid-perturbed physical optics approaches that combine the salient features of both of these theories (2,5,7,8,9). It has also been shown that the enhanced backscatter that has been observed from very rough surfaces is due to multiple scattering (10,11,12,13). One problem with these hybrid solutions based on a two-scale surface model is that the results critically depend upon wavenumber kd where spectral splitting is assumed to occur (8). In general, using these hybrid approaches, one cannot choose kd such that the large-scale surface hl and the small-scale surface hs (that rides on the large scale surface) simultaneously satisfy the physical optics and the perturbation restrictions, respectively. It has also been shown that even when a hybrid solution is used to approximately determine the copolarized cross section through a “suitable” choice of kd , it cannot be used to determine the cross-polarized cross section (14,15). The full-wave solutions are not restricted to electromagnetic scattering by layered media with irregular interfaces. Scattering due to inhomogeneities in the complex electrical permittivities and magnetic permeabilities in each layer can also be accounted for in the analysis. The full-wave solutions can also be used to determine the coupling between the radiation fields, the lateral waves, and the guided (surface) waves of the layered structures. They can be used to determine the scattered near fields as well as far fields. Both large-scale and small-scale (including subwavelength) fluctuations of the rough surface and medium parameters are accounted for in the analysis.

Schelkunoff’s Generalized Telegraphists’ Equations for Bounded Irregular Waveguides and the Use of Local Mode Expansions Generalized telegraphists’ equations, which are based on the use of complete expansions of the electromagnetic waves (into vertically and horizontally polarized radiation fields, lateral waves, and surface waves) as well as on the imposition of exact boundary conditions at the rough interfaces of irregular stratified media, have been derived (16,17,18) for electromagnetic fields scattered by irregular stratified media with rough interfaces. The

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

3

Fig. 1. Schematic of procedures for deriving the generalized telegraphists’ equations.

analytical procedures used to derive these generalized telegraphists’ equations are similar to those advanced by Schelkunoff (19), to solve problems of mode coupling in irregular waveguides with finite cross sections and impedance boundary conditions. Since the field expansions do not converge uniformly on the irregular boundaries, Schelkunoff (19,20) employed precise mathematical procedures to avoid term-by-term differentiation of the field expansions (infinite sets of TE and TM modes for cylindrical waveguides). The method used to convert Maxwell’s equations into sets of generalized telegraphists’ equations for the reflected and transmitted wave amplitudes in irregular layered structures is shown schematically in Fig. 1. The intrinsic properties duality, reciprocity, realizability, and invariance to coordinate transformations are also listed in Fig. 1. For open structures consisting of half-spaces (such as the irregular-layered media considered in this work), the complete field expansions are associated with integrals along two branch cuts (the radiation and the lateral wave terms) and residues at pole singularities (waves guided by the stratified structure) (16,17,18). Schelkunoff’s method has also been used to solve problems of mode coupling in a wide class of irregular waveguides such as waveguide tapers and waveguide bends, as well as in waveguides with nonperfectly conducting surfaces that are characterized by impedance boundary conditions (19,20). In all these bounded waveguide systems, the field expansions are expressed in terms of infinite, discrete sets of propagating and evanescent waveguide modes associated with the characteristic equations for cylindrical waveguides with ideal, perfectly conducting boundaries. In waveguides of arbitrarily varying cross sections with finitely conducting boundaries, the modes of the ideal cylindrical waveguides, while complete, do not individually satisfy the correct boundary conditions, and the mode expansions do not uniformly converge on the irregular boundaries. To keep

4

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

his analysis rigorous, Schelkunoff (19,20), employed rather tedious, but necessary, mathematical procedures on imposing exact boundary conditions. Thus, for example, orders of integration and differentiation are not interchanged in order to account for the nonuniform convergence of the field expansions and the fact that the range of the cross section variables (limits of the corresponding integrals) are not constant. The coupling between the waveguide modes is due to the nonideal boundary conditions. In an attempt to reduce the number of significant coupled, spurious modes that need to be accounted for in multimode waveguides with irregular cross sections, new generalized telegraphists’ equations were derived based on field expansions in terms of a complete set of waveguide modes that individually satisfy the boundary conditions locally. Thus, for example, in waveguides with abrupt or gradual tapers, waveguide modes in uniform tapers (with constant flare angles) were used in the local field expansions (21,22,23). In waveguide bends with arbitrarily varying curvatures, the fields were expressed in terms of local annular waveguide modes (24,25) and in waveguides with varying impedance boundaries, modes that locally satisfy the impedance boundary conditions were used (26,27). The modal equations for the local waveguide modes were usually more difficult to solve than those for ideal cylindrical waveguides. However, the generalized telegraphists’ equations can be solved numerically (using the Runge–Kutta Method) more readily when the local modal expansions are used, since coupling into the spurious local modes is bunched more tightly around the incident mode. This is because the local modes individually satisfy the local boundary conditions in the waveguide. These analytical and numerical results were validated experimentally in a series of controlled laboratory experiments used to synthesize waveguide transition sections (28). These controlled laboratory studies were first conceived by Wait (29,30,31), to study VLF radio wave propagation in scaled laboratory models of the earthionosphere waveguide. In these models, the ionosphere effective boundary was simulated by an absorbing foam material with a specified complex dielectric coefficient and thickness (manufactured by Emerson Cumming) (32, 33).The earth’s curvature was also simulated in these laboratory models using a nondissipative inhomogeneous dielectric material to load the interior of the straight model waveguide (34,36). This experimental procedure to stimulate curvature was carried out in the scaled model at microwave frequencies (scaling factor 106 ). It is analogous to the mathematical earth-flattening technique developed by Kerr (37). The dominant mode in the (simulated) curved model waveguide had the same characteristics of the earth-detached mode in the earth-ionosphere waveguide. They can be expressed in terms of Airy integral functions (instead of sinusoidal functions in empty, rectangular waveguides).

Generalized Telegraphists’ Equations for Irregular Stratiﬁed Media with one or Two Half-Spaces Following the extensive analytical, numerical, and experimental work on electromagnetic wave propagation in bounded irregular waveguide structures, propagation in irregular stratified structures with one or two infinite half-spaces were analyzed using the full-wave method. Approximate impedance boundary conditions (38,39) were replaced by exact boundary conditions at the rough interface between two media characterized by different complex permittivities and permeabilities (16,18). Furthermore, scattering due to laterally inhomogeneous permittivities and permeabilities in each layer of the irregular stratified media are also accounted for in the analysis. The initial impetus for this work was the complex and intriguing sloping beach problem considered by Wait and Schlak (40) in which the sea was modeled (two-dimensionally) as a small-angle wedge region adjacent to horizontal dry land. Exact modal expansions of the fields in the four wedge-shaped regions (sea water, wet land under the sea, dry land, and free space) involve Kontorowich–Lebedev transforms (41). The relationships between the Fourier, Watson, and Kontorowich–Lebedev transforms have been obtained through the use of a generalized Bessel transforms (42). The analytical solution based on the Kontorowich–Lebedev transforms involve integration over the order of the Bessel functions. Schlak and Wait (40) employed a geometric optics approach which give exact results for parallel stratified media. However, these results were shown by them

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

5

Fig. 2. Electromagnetic radiation in irregular layered structures.

to be nonreciprocal even for the small wedge angles they considered. King and Husting (43), who conducted a series of controlled experiments on laboratory models, showed that the results were more accurate when the direction of propagation was toward the apex of the wedge (rather than away from it).

Transformation of Maxwell’s Equations In this section the full-wave procedures used to convert Maxwell’s equations into generalized Telegraphists’ equations are outlined for propagation of electromagnetic waves in irregular layered structures with twodimensionally rough interfaces and transversely varying electromagnetic parameters. The interested reader will find the details of the analysis in the published literature (16,17,18). The procedures are shown schematically in Fig. 1, and the irregular layered structure is illustrated in Fig. 2. The vertical axis is y and the interface between medium i and i + 1 is given by f (x, y, z) = y − hi,i+1 (x, z) = 0. The complex permeativity and permeability µ in each layer of the structure are assumed to be functions of the lateral variables x and z.

6

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

Maxwell’s equation for the transverse (y, z) components (denoted by subscript T) of the electric E and magnetic H fields can be expressed as follows:

and

in which the operator ∇ T is

and the transverse vectors are

The electric and (dual) magnetic current densities are J (A/m2 ) and M (V/m2 ). The exact boundary conditions imposed at each of the interfaces of the irregular layered structure are the continuity of the tangential components of the electric and magnetic fields

in which n is the unit vector normal to the interfaces

The full-wave, complete expansions for the vertically (V) and horizontally (H) polarized electric and magnetic fields are given in terms of the transverse basis functions

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

7

in which the y-dependent scalar basis functions ψP for the vertically (P = V) and horizontally polarized waves associated with the radiation fields, the lateral waves, and the surface waves of the layered structure are (18)

and

In the above equations, R and T are associated with the Fresnel reflection and transmission coefficients, vr is the y component of the wave vector in medium r kr (u, vr , w), vq − 1,q = vq − 1 − vq , and the z-dependent scalar

8

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

function is

The wave impedances and admittances for the vertically and horizontally polarized waves are

The transverse components of the electric and magnetic fields are expressed completely as follows:

and

in which the symbol v denotes summation (integration) over the complete wave vector spectrum consisting of the radiation term and lateral waves (associated with branch cut integrals) and the waveguide modes (or bound surface waves) of the layered structure (associated with the residues at the poles of the reflection coefficients). In Eqs. (17) and (18) the scalar field transforms for the vertically (P = V) and horizontally (P = H) polarized electric and magnetic fields are

and

where the complementary (reciprocal) basis functions are

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

9

and

in which φc (w, z) = (1/2π) exp(iwz) and use has been made of the biorthogonal relationships

In Eq. (25) the Kronecker delta δP,Q implies that the vertically (P, Q = V) and horizontally (P, Q = H) polarized basis functions are orthogonal. Furthermore, the Dirac delta function δ(w − w ) appearing in Eq. (25) is a result of the Fourier transform completeness and orthogonality relationships:

The corresponding completeness and orthogonality relationships satisfied by the scalar basis functions ψP are

and

10

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

In Eqs. (28) and (29),

in which ZV and Y H are the wave impedances and admittances for the vertically and horizontally polarized waves, respectively. Furthermore, the symbol (v, v ) in Eqs. (25) and (29) is the product of the Kronecker delta δq,r and the Dirac delta function δ(v, v ) for the radiation and lateral wave terms or the Kronecker delta δv,v for the bound guided (surface) waves of the layered structure. Thus the radiation fields, the lateral waves, and the guided waves of the full-wave spectrum are mutually orthogonal (16,17). The radiation fields and the lateral waves are associated with branch cut integrals in the complex wave number plane [with branch points at k = k0 (uppermost medium) and k = km (lowermost medium)]. The guided waves of the layered structure are associated with the residues at the poles of the composite reflection coefficient seen from above or below the layered structure. In this work, it is convenient to express the vertically and horizontally polarized scalar field transforms [Eqs. (19) and (20)] in terms of the vertically and horizontally polarized forward wave amplitude aP and backward wave amplitude bP as follows:

Upon substituting the complete field transforms into the transverse components of Maxwell’s equation [Eqs. (1)–(4)], making use of the biorthogonal relationships [Eq. (28)], and imposing the exact boundary conditions at each interface of the irregular layered structure [ Eq. (5)], the following generalized telegraphists’ equations are derived (see Fig. 1):

where AP and BP are associated with the source terms J and M in Eqs. (1)–(4). Furthermore, SBA PQ and SAB PQ are transmission scattering coefficients, while SAA PQ and SBB PQ are reflection scattering coefficients. These scattering coefficients vanish when the layered medium is horizontally stratified with homogeneous medium in each layer. In this case, the forward and backward wave amplitudes for the vertically and horizontally polarized waves are decoupled and analytical closed form solutions are readily obtained. However, if the rough surface height or the complex permittivities and permeabilities are functions of x and z, the wave amplitudes are coupled. In the general case, the basic functions ψP do not individually satisfy the irregular boundary conditions and the complete field expansion do not uniformally converge at the boundaries. Thus, on following precise mathematical procedures (16,17,18,19,20), the orders of integration (summation) and differentiation cannot be interchanged. As a result, the rigorous derivations of the generalized telegraphists’ equations [Eqs. (32) and (33)] are rather tedious.

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

11

The intrinsic properties of the full-wave solutions are (see Fig. 1) duality, reciprocity, realizability, and invariance to coordinate transformation. All the above properties follow directly from Maxwell’s equations. [(1)–(4)], and they are not a result of any additional constraints imposed on the results. A two-dimensional scalarized version of this problem has also been analyzed (44,45,46,47). When the lowermost and/or uppermost half-space is perfectly conducting or a good conducting medium, the two boundary conditions [ Eq. (5)] at the lowermost (and/or uppermost) interface can be replaced by a single surface impedance boundary condition:

In Eq. (34) the unit vector n is normal to the interface and points into the conducting half-space. For an isotropic conducting half-space, the surface impedance Zs is a scalar. In general, the surface impedance can be represented by a dyad. The impedance boundary condition has been used to simplify the analysis of irregular layered structures (48,49,50,51). Contributions from integrals associated with one (or two) branch cuts are eliminated when impedance boundary conditions are used. The generalized telegraphists’s equations have also been derived for irregular multilayered cylindrical structures (52,53,54,55) and irregular spherical structures (56,57,58). For the irregular cylindrical and spherical layered structures, the complete expansions of the fields in terms of cylindrical/spherical harmonies are related to the Watson transformations (59).When the innermost regions of the cylindrical/spherical structures are highly conducting and impedance boundary conditions are used, the contribution from the continuous portion of the wave spectrum can be ignored and the solutions are expressed in terms of discrete waveguide modes (42). A set of generalized telegraphists’ equations similar to Eqs. (32) and (33) have been derived for the irregular cylindrical/spherical structures (52,53,54,55,56,57,58). However, it should be noted that for the spherical case, the wave admittances/impedances and propagation coefficients for the forward and backward propagating wave amplitudes are not the same. For the spherical/cylindrical cases, solutions of the model (characteristic) equations are far more complicated, and numerical techniques have been developed to trace the loci of the complex roots of the characteristic equations (60). These procedures have been used to solve problems of electromagnetic wave propagation in naturally occurring or man-made perturbed models of the earth-ionosphere waveguide. Experiments in controlled laboratory models (based on the pioneering work by Wait) have been conducted to validate the analytical results (29,31,34).

Iterative Solutions to the Generalized Telegraphists’ Equations and Their Relationships to the Small Perturbation Solution and the Physical/Geometrical Solutions to Rough Surface Scattering Iterative analytical procedures, as well as numerical techniques, are used to solve the generalized telegraphists’ equations [Eqs. (32) and (33)] for the forward and backward wave amplitude scattered by two-dimensionally rough surfaces (see Fig. 3). An overview of the results are shown schematically in Fig. 4. The analytical procedures are dealt with first in this section. To obtain the single scatter approximations for the wave amplitudes, the expressions for the primary fields impressed upon the rough interface due to the sources are first derived from Eqs. (32) and (33) upon neglecting all the coupling terms manifested by the scattering coefficients SBA PQ . When the sources are in the far field, the primary, incident fields impressed upon the rough surface are vertically and horizontally polarized plane waves propagating in the direction of the (free space) wave vector . Thus, the primary electric ki 0 = ki 0x ax + ki 0y ay + ki 0z az = k0 ni , where ni is a unit vector and k0 = ω

12

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

fields impressed upon the rough surfaces are

where RP 0 is the P = V, H polarized Fresnel reflection coefficient for waves incident from medium 0 (free space) upon medium 1 (see Fig. 3), and aP is the unit vector in (P = V) or perpendicular (P = H) to the plane of incidence. The primary fields are proportional to the local basis function ψP 0 (v, y) given by Eq. (11). The corresponding vertically or horizontally polarized field transforms and wave amplitudes are obtained using Eqs. (19), (20), and (31). In view of the biorthogonality relationships [(25)], the primary wave amplitudes are proportional to the delta functions corresponding to the polarization (Q = V, H) and direction ki 0 (wi 0 , vi 0 , wi 0 ) of the incident waves. When these expressions for the primary wave amplitudes are substituted for aQ and bQ on the right-hand side of Eqs. (32) and (33) (with the source terms AP and BP suppressed) the (iterative) differential equations for single scattered wave amplitudes are obtained. The solutions for these single scatter wave amplitudes are substituted into the expressions for the field transforms [Eqs. (17) and (18)] to obtain the single scattered fields. Since both vertically and horizontally polarized incident waves are considered and both like- and cross-polarized scattered waves result from two-dimensionally rough surfaces, the results for the diffuse scattered fields are presented here in matrix notation.

In Eq. (36)

where EP s and EP i are the vertically (P = V) and horizontally (P = H) polarized components of the scattered fields and the incident waves (at the origin), respectively. The 2 × 2 scattering matrix S is given by

The elements of the matrix R are

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

13

Fig. 3. Relationships between the incident and scatter wave normals ni and nf , respectively, local tangent planes (r − rs ) · np = 0, and planes parallel to the stationary phase planes (r · rs ) · ns = 0 for rough surface scattering.

The wave vectors in the scatter and incident directions are

and

14

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

Fig. 4. Principal properties of the original and unified full-wave solutions.

where θ0 is the elevation angle (measured from the y axis and φ is the azimuth angle measured from the x axis. Furthermore, Ci 0 = cos θi 0 , C 0 = cos θ 0 , Si 0 = sin θi 0 , S 0 = sin θ 0 . The corresponding quantities associated with medium 1 are denoted by the subscript 1 and θ1 , is related to θ0 by Snell’s law. The vector v is

and rs and r are position vectors from the origin to points on the rough surfaces and to the observation point, respectively:

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

15

Furthermore,

In Eq. (36) the integrations are over the rough surface (transverse) variables xs and zs as well as the wave vector variables k y and k z . The first term Gf contains the exponent exp(ivy h) while the second term Gf d does not. On integrating the second term with respect to xs and zs , the delta functions are obtained:

Thus the second term Gf d can be readily shown to be the specularly reflected wave from a flat surface at y = 0, since RVV and RHH reduce to the Fresnel reflection coefficients for the vertically and horizontally polarized waves and RVH → 0, RHV → 0 for the specular case k → ks = ki + 2k0 cos θi 0 ay , v → 2k0 cos θi 0 ay . Note that the results in Eq. (36) are in complete agreement with the earlier work in which it is assumed that the vector n normal to the rough surface is restricted to the xy plane (hz = 0) (25). This is because the restriction does not constrain the unit vector n to lie in the scatter plane (normal to k xki ). In the recent work by Collin (61), the author uses a different full-wave approach to the problem of scattering of plane waves from perfectly conducting surfaces: He uses a pair of odd and even scalar basis functions for the Dirichlet and Neumann boundary conditions. These basis functions and the corresponding reciprocal basis functions (chosen to be their complex conjugates) are explicit functions of y and implicit functions of x and z [through the expression for h(x, z), the rough surface height]. The resulting source free wave equation is further (Fourier) transformed in x and z to obtain an equation with a dyadic operator for the vector field transform (and equivalent, slope-dependent sources that account for scattering) rather than generalized telegraphists’ equations for the scalar wave amplitude. Upon inverting the dyadic operator, evaluating the residue at k0 , and integrating by parts, Collin’s results are also shown to be in complete agreement with the full-wave results for the perfectly conducting case (|r | → ∞, µr = 1). Collin referred to the results for the diffuse scattered fields (25) as the original full-wave solutions (see Fig. 4). The above first-order iterative solutions for the single scattered fields [Eq. (36)] are restricted to rough surfaces with small mean-square slopes σs < 0.1 (3). This is because the scattering coefficients Sαβ PQ (α, β = A, B) appearing in the generalized telegraphists’ equations [Eqs. (32) and (33)] are explicitly dependent on the slopes of the rough surface. Alternatively, in Collin’s work, the equivalent source terms are slope-dependent. However, unlike the small perturbation solution, the full-wave solutions are not restricted to rough surfaces with small mean-square heights. Furthermore, the full-wave solutions [Eq. (36)] can be used to evaluate the near fields, the far fields, and the fields in the intermediate region. Thus, this work can be applied to probing subwavelength structures, an area that has attracted much interest in near field optics. In addition, the firstorder scattering results can be extended to multiple scattering. In particular, the full-wave approach has been used to account for double scatter that is associated with observed backscatter enhancement (12). When the observation point is at a very large distance from the rough surfaces (k1 r k0 L 1 and k0 r k0 l 1), the integration with respect to the scatter wave vector variables (k 0y , k 0z ) can be performed

16

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

analytically using the stationary phase method. Thus, if the observation point is in the direction

the diffuse far fields scattered from the rough surface are

The expression for S(kf , ki ) in Eq. (52) is the same as the expression for S(k , ki ) in Eq. (36) except that the scatter wave vector k is replaced by kf , where kf 0 = k0 nf [Eq. (51)] and kf 1 the wave vector for y < h(xs , zs ) is related to kf 0 through Snell’s law. Furthermore,

and

In Eq. (54), vy = k0 (Ci 0 + Cf 0 ) = k0 (cos θi 0 + cos θf 0 ). When the integrations with respect to xs and zs are performed, the term Gf D is shown to be the flat-surface quasi-specular (zero-order) scattered field which is proportional to (4Ll/vx Lvz l) sin vx L sin vz l. The expression for the quasi-specular scatter term Gf D is the same as the expression for the total field Gf except that rs in Gf is replaced by rt in Gf D [Eq. (52)]. Thus, for h(xs , zs ) = 0, they are identical and Gf s = 0. It is readily shown that the full wave-solution [Eq. (52)] reduces to the small-height–small-slope perturbation solution of Rice provided that it is assumed that k0 h 1. Thus on retaining the first two terms of the Taylor series expansion of exp(ivy h) it follows that

In this small-height–small-slope limit, the full-wave solution is indistinguishable from the small perturbation solution for the far fields scattered by slightly rough surfaces (see Fig. 3). These limiting forms of the full-wave solutions are, however, no longer invariant to coordinate transformations since h(x, z) does not appear in the exponent. Furthermore, it is shown that they are valid only if the height and slopes are of the same order of smallness. Turning now to the high-frequency limit, it is assumed that the radii of curvature of the rough surfaces are very large compared to wavelength. The unit vector np normal to these large-scale patches of rough surfaces is assumed to have arbitrary orientation. Thus the planes of incidence and scatter with respect to the reference coordinate system (normal to ni xay and nf xay , respectively) are not the same as the local planes of incidence and scatter with respect to the local coordinate system (normal to ni xnp and nf xnp , respectively). Furthermore, the sines and cosines of the angles of incidence and scatter appearing in the scattering coefficients [Eqs. (39)– (42)] are not the same as the sines and cosines of the local angles of incidence and scatter. In order to account for the arbitrary slope of the large-scale surface, the surface element scattering matrix S(kf , ki ) in Eq. (36) is

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

17

replaced by

In Eq. (56) the matrix operator T i decomposes the waves which are vertically and horizontally polarized with respect to the reference plane of incidence (normal to ni xay ) into vertically and horizontally polarized waves with respect to the local plane of incidence (normal to ni xn). Similarly, the matrix operator T f decomposes the waves which are vertically and horizontally polarized with respect to the local plane of scatter (normal to nf xn) back into vertically and horizontally polarized waves with respect to the reference plane of scatter (normal to nf xay ). Thus if ψi and ψf are the angles between the reference and local planes of incidence and the reference and local planes of scatter, respectively, then

in which for j = i or f we have

where

Furthermore, the cosines of the local angles of incidence and scatter appearing in Sn (kf , ki ) [Eq. (56)] are given by

18

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

while Sin 0 and Sfn 0 are the sines of the local angles of incidence and scatter. The corresponding quantities associated with medium 1 are denoted by the subscript 1. The local angles of incidence and scatter in medium 1 are related to the local angles of incidence and scatter in medium 0 through Snell law. Implicit in Eq. (56) are the self-shadow functions U(nf · np ) and U(−ni · np ) (where U is the unit step function) since the local angles of incidence and scatter are less than 90◦ . Furthermore, cos(φf − φi ) and sin(φf − φi ) appearing in Eq. (38) are replaced by (62)

The above changes represented by Eq. (56) constitutes the transformation into the large-scale (patch) coordinate system (see Fig. 5). It is readily shown that at high frequencies, the major contributions come from the vicinity of the stationary-phase, specular points on the rough surface where np is along the bisector between nf and −ni (see Fig. 3). Pursuant to the transformation Eq. (56) it can be shown that at these stationary-phase points, RVV and RHH reduce to the familiar Fresnel reflection coefficients while the cross-polarized terms RVH and RHV vanish at the specular points. Thus in these limits, the full-wave solution [Eq. (52)] reduces to the physical optics solution for the diffuse scattered fields (4). If, in addition, Eq. (52) is evaluated analytically using stationary-phase approximations, the full-wave solution reduces to the geometric optics solution (see Fig. 4). However, in order to account for multiple scatter at the same rough surface, it is necessary to return to the original form [Eq. (36)] even at high frequencies (13).

Full-Wave Solutions for the Radar Cross Sections for Multiple-Scale Rough Surfaces The normalized bistatic radar cross sections σPQ for two-dimensionally rough surfaces are dependent on the polarizations of the scattered (first superscript P = V, H) and incident (second superscript Q = V, H) waves. It is defined as the following dimensionless quantity that depends on the incident and scatter wave-vector directions:

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

19

Fig. 5. Arbitrarily oriented patch of a rough surface.

In Eq. (64) the area Ay is the radar footprint, r is the distance from the rough surface to the far-field observation point. When the rough-surface statistical characteristics are homogeneous though not necessarily isotropic, the (ensemble average) full-wave radar cross section based on the original (denoted by subscript 0) full-wave analysis [Eq. (52)] is expressed as follows:

where SPQ (nf , ni ) is the surface element scattering coefficient for incident waves in the direction ni and polarization Q = V (vertical), H (horizontal), and scattered waves in the direction nf and polarization P = V, H. It should be noted that the scattering coefficients SPQ (nf , ni ) are not functions of slope. In Eq. (65), Q(nf , ni ) is expressed in terms of the surface height joint characteristic function χ2 and characteristic function χ as follows:

20

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

where k0 is the free-space wavenumber and rdt is the projection of rS1 − rS2 (where rS1 and rS2 are position vectors to two points on the rough surface) on the mean plane (y = 0) of the rough surface y = h(x, z) (see Fig. 3):

and drdt = dxd dzd . The vector v is given by Eq. (44). For homogeneous isotropic surfaces with Gaussian joint surface height probability density functions

where vy is the y component of v [Eq. (44)], h2 is the mean-square height and R(rd ) is the normalized surface height autocorrelation function [related to the Fourier transform of the surface height spectral density function W(k)]. When the surface is isotropic and homogeneous, R is only a function of the distance

and Q(nf , ni ) [Eq. (66)] can be expressed as follows for L, l lc (the autocorrelation length):

where J 0 is the Bessel function of order zero and

Note that Q(nf , ni ) remains finite as vy → 0. The above expressions based on the original full-wave solutions are in total agreement with solutions based on Rice’s small perturbation solutions when the height and slopes are of the same order of smallness (3). For these cases

and the integrals in Eq. (66) can be expressed in closed form in terms of the rough-surface height spectral density function [the Fourier transform of the surface height autocorrelation function hh = h2 R(rd )]. However, the solution [Eq. (66)] based on the original full-wave solution is not restricted to surfaces with small mean-square heights. Since it is based on the first-order single-scatter iterative solution, it is nevertheless restricted to surfaces with small slopes (σ2 s < 0.1). When slopes of the rough surface are not small and the scales of roughness are very large compared to wavelength, solutions based on the transformation [Eq. (56)] can be used. Thus, the diffuse scatter cross section

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

21

is expressed as follows:

where the symbol ∗ denotes the complex conjugate and

The (statistical) mean scattering cross section for random rough surfaces is obtained by averaging Eq. (74) over the surface heights and slopes at points rs1 and rs2 . The coherent component of Eq. (75) is defined as

and the incoherent scattering cross section is defined as

The above expression for the radar cross sections for two-dimensional random rough surfaces involves integrals over the random rough-surface heights and slopes and the surface variables xs1 , xs2 , zs1 , zs2 . This expression can be simplified significantly if the radii of curvature of the large-scale (patch) surface are assumed to be very large compared with the free-space wavelength. In this case, the slope at point 2 may be approximated by the value of the slope at point 1 (hx2 ∼ hx1 , hz2 ∼ hz 1). If, in addition, the rough surfaces are assumed to be statistically homogeneous, the cross section is expressed as follows:

in which the analytical expressions for the conditional joint characteristic functions are

22

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA Furthermore,

where P2 (nf , ni |ns ) is Sancer’s (63) shadow function and ns is the value of np at the specular points. For random rough surfaces characterized by a four-dimensional Gaussian surface high/slope coherence matrix we obtain

and

In Eq. (85), C is the Gaussian surface height autocorrelation function and lcx , lcz are correlation lengths in the x and z directions, respectively. When the surface is isotropic (lcx = lcz = lc and σ2 x = σ2 z = σ2 s ), Eqs. (81), (83), and (84) reduce to

where

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

23

and

In this case, the total mean square slope is expressed as

For the assumed isotropic surface with Gaussian statistics, the four-dimensional integral [Eq. (28) with Eqs. (86) and (88)] can be expressed as a three-dimensional integral using a Bessel function identity (4). The resulting full-wave incoherent diffuse scatter cross section that accounts for surface height/slope correlations is expressed as

where

Furthermore,

and

24

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

In Eq. (93),

and

In Eq. (90), p(hx , hz ) is the probability density function for the slopes (assumed here to be Gaussian). It is shown that the above results in which the correlations between the surface heights and slopes have been accounted for in the analysis reduce to the small perturbation results when the heights and slopes are of the same order of smallness and reduce to the physical/geometrical results in the high-frequency limit (64). These full-wave results have also been compared with numerical and experimental results for one-dimensionally (3) and two-dimensionally (64) rough surfaces. When the rough surface consists of multiple scales of roughness as in the case of sea surfaces, two scale models have been introduced to obtain the scatter cross sections. Thus, the surface is assumed to consist of a small-scale surface that is modulated by the slopes of the large-scale surface and the cross section is expressed as a sum of the cross sections for the large- and small-scale surfaces. However, Brown (8) has shown that the hybrid-perturbation–physical-optics results critically depend upon the choice of the spatial wavenumber kd that separates the large-scale surface from the small-scale surface. To apply this hybrid-perturbation–physicaloptics approach, the Raleigh rough-surface parameter β = 4k2 0 h2 s must be chosen to be very small compared to unity. This places a very strict restriction on the choice of kd . As a result, scattering from the remaining surface consisting of the larger-scale spectral components with kl < kd may not be adequately analyzed using physical optics (see Fig. 4). The above Raleigh rough-surface parameter β does not place any restriction on the choice of kd when the full-wave analysis is used. Furthermore, it is shown (65) that the full-wave solution for these multiple-scale rough surfaces are expressed as weighted sums of two cross sections:

where σPQ l is the cross section associated with the surface consisting of the larger spectral components (kl < kd ), while σPQ s is the cross section associated with the surface consisting of the smaller spectral components ks > kd . Scattering by the small-scale surface is modulated by the slopes of the large-scale surface, while scattering by the small-scale surface is diminished by a coefficient (less than unity) that is equal to the magnitude of the small-scale characteristic function squared [Eq. (69)]. Thus, using the full-wave approach, extensive use is made here of the full-wave scattering cross-section modulation for arbitrarily oriented composite rough surfaces. Thus, the incoherent diffuse radar cross sections of the composite (multiple scale) rough surface is obtained by regarding the composite rough surface as an ensemble of individual patches (several correlation lengths in the lateral dimension) of arbitrary orientation (see Fig. 5). The cross section per unit area of the composite rough surface is obtained by averaging the cross sections of the individual arbitrarily oriented pixels. It is shown that the (unified full wave) cross section of the composite rough surface is relatively stationary over a broad range of patch sizes. In this broad range of values of patch sizes, the norm of the relative error is minimum.

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

25

A patch is assumed to be oriented normal to the vector (see Fig. 5)

where a x , a y , and a z are the unit vectors in the fixed (reference) coordinate system associated with the mean plane y = h0 = 0 and hx = ∂h/∂x, hz = ∂h/∂z. The unit vectors ax and az are tangent to the mean plane of the patch. The angles and τ are the tilt angles in and perpendicular to the fixed plane of incidence (the x , y plane). The cosines of the angles of incidence and scatter in the patch coordinate system can be expressed in terms of the cosines of the angles of incidence and scatter in the fixed reference coordinate system (primed quantities; see Fig. 5) as follows:

and

The surface element scattering coefficient for the tilted pixel is expressed as follows (66):

in which SPQ p the elements of the 2 × 2 scattering matrix Sp are obtained from SPQ on replacing the angles θi 0 and θf 0 by θi 0 and θf 0 , respectively. Furthermore, cos(φf − φi ) and sin(φf − φi ) are replaced by the cosine and the sine of the angle (φf − φi ) between the planes of scatter and incidence (with respect to the pixel coordinate system (see Fig. 5) (62). The matrices T f p and T i p relate to the vertically and horizontally polarized waves in the reference coordinate system to the vertically and horizontally polarized waves in the local (patch) coordinate system (66). Thus

where

26

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

and

The angles and τ can be expressed in terms of the derivatives of h(x, z) as follows:

The radar cross section (per unit area) for the tilted patch can be expressed as follows:

and

where

in which

Thus, in Eq. (108) both |DPQ p |2 and Qp are functions of the slopes hx and hz of the tilted patch mean plane (see Fig. 5). For a deterministic composite rough surface, the slopes (that modulate the orientation of the

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

27

patch) are known. The radar scatter cross section for this composite surface is given by summing the fields of the individual patches. However, if the composite surface height is random, the tilted pixel cross section (per unit area) [Eq. (108)] for the rough surface is also a random function of the pixel orientation. Thus, in order to determine the cross section per unit area of the composite random rough surface, it is necessary to evaluate the statistical average of σPQ p . The cross section of the composite random rough surface is given by

where · denotes the statistical average (over the slope probability density function p(hx , hz ) of the tilted patch). The mean-square slope of the tilted patch is given in terms of the surface height spectral density function

For the case kp → 0, Lp , lp → ∞, σp → 0, the cross section σPQ p reduces to the original full-wave solution σPQ 0

[Eq. (65)]. In Eq. (114), the upper limit kp is the wavenumber associated with the patch of lateral dimension Lp = 2π/kp . In the expression for Qp (nf , ni ) [Eq. (109)], the surface height autocorrelation function for the rough surface associated with the patch is given in terms of the Fourier transform of the surface height spectral density function as follows:

where it is assumed that the surface is homogeneous and isotropic, k = (k2 x + k2 z )1/2 . Illustrative examples of the results obtained for the scatter cross section using the above procedures have been published (66). For purposes of comparisons, the generalized telegraphists’ equations [Eqs. (32) and (33)] have also been solved numerically for one-dimensionally rough surfaces (67). The procedures used are outlined here. On extending the range of the wave vector variable u from −∞ to ∞, the Eqs. (32) and (33) are combined into one coupled integrodifferential equation for the forward and backward scattered wave amplitudes a(x, u) and a(x, −u), respectively. On extracting the rapidly changing part exp(−iux), the wave amplitudes are expressed as

The total wave amplitude is the sum of the source-dependent primary wave amplitude AP p and the diffusely scattered wave amplitude AP s due to the surface roughness:

28

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

The primary wave amplitude is obtained from Eqs. (32) and (33) on ignoring the coupling terms Sαβ PQ . The resulting integrodifferential equation for the diffusely scattered term AP s is converted into an integral equation with mixed boundary conditions. This expression is integrated by parts to get rid of the singularity in the scattering coefficient. The resulting integral equation is solved numerically using the standard moments method. Finally the field transforms [Eqs. (17) and (18)] are used to obtain the results for the electromagnetic fields from the wave amplitudes. For the far fields, these expressions can be integrated analytically (over the wavenumber variable) using stationary-phase techniques. These results show that for surfaces with small to moderate slopes the proceeding analytical results are valid (67). When the observation points are near the surface, it is necessary to account for coupling between the radiation fields, the lateral waves, and the surface waves associated with rough surface scattering (68,69). When the rough surface is assumed to be perfectly conducting, the contribution from the branch cut integral associated with the lateral waves vanishes and there are no residue contributions (associated with surface waves) from the singularities of the reflection coefficients. When the approximate impedance boundary condition is used, the lateral wave contribution is eliminated. The full-wave method can also be used to determine the fields scattered upon transmission across rough surfaces (70). When scattering from more than one rough interface in an irregular stratified media is considered, in general, it becomes necessary to account for scattering upon reflection and transmission across rough surfaces. This topic is reviewed in the next section.

Full-Wave Solutions for Three Media Irregular Stratiﬁed Structures Full-wave solutions are derived for the electromagnetic fields scattered by the two rough surfaces in a realistic physical model of a three media environment. They account for five different scattering mechanisms that the waves undergo, assuming that both the transmitter and receiver are above the uppermost interface of the irregular media. Two scattering mechanisms are associated with reflection from above and below the upper interface, and two are associated with transmission across the upper interface; the fifth is associated with reflection from above the lower interface. In view of the fact that in general the two rough interfaces are characterized by independent random rough surface heights (except where the thickness of the intermediate medium vanishes) the rough surface height joint probability density functions are characterized by a family of probability density functions associated with the gamma functions. Multiple bounces between the two interfaces are accounted for in the analysis. The elements of the incoherent Mueller matrix (that relates the scattered to the incident Stokes vectors) can be obtained from the expressions for the scattered fields. From the simulated data it is possible, for instance, to determine the optimal polarizations and the incident and scatter angles of the waves as well as the wavelength, for purposes of suppressing or enhancing the impact of the clutter from the rough interfaces. This work can be used to provide realistic models of electromagnetic scattering from snow-covered terrain, ice-covered sea surfaces, naturally occurring or man-made oil slicks, and coated rough surfaces. The models can be used to reduce the impact of signal clutter from the rough interfaces, to facilitate the detection of buried objects. The diffusely scattered electromagnetic fields for the two irregular structures illustrated in Fig. 6 were investigated initially. In model 1, the upper interface is flat and the lower interface is rough (71,72,73,74). In model 2, the upper interface is rough and the lower interface is flat (75). In this work, the model of the irregular structure considered consists of two random rough (upper and lower) interfaces. The thickness of the coating material (film) between the two random rough interfaces is assumed to be arbitrary. The physical mechanism for scattering from coated rough surfaces is schematically

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

29

Fig. 6. Two previously investigated models with only one rough interface.

illustrated in Fig. 7. The upper interface for y = h01s (xs , zs ) between medium 0 and medium 1 is

where the mean value of h01s is < h01s >= h01 . The lower interface y = h12s (xs , zs ) between medium 1 and medium 2 is

The thickness of the coating layer, H D , is

The unit vectors normal to the large-scale rough interfaces between medium 0 and 1 and between medium 1 and 2 are

where

Using the full-wave approach (71), the diffuse first-order scattered fields can be expressed as the sum

30

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

Fig. 7. Irregular layered model with two rough interfaces.

where EPQ SU (r) is associated with scattering from the upper interface, and EPQ SD (r) is associated with scattering from the lower interface. For eiwt time harmonic plane wave excitations, the incident electric field of magnitude EiP 0 is

In Eq. (125), aP is parallel (P = V) or perpendicular (P = H) to the reference plane of incidence (normal to ni 0 × ay where ni 0 is the unit vector in the direction of the wave vector ki 0 for the incident waves). For planewave excitations and for observation points above the upper interface y ≥ h01s (xs , zs ) the scattered fields due to the rough upper and lower interfaces are given by (72,75,76,77).

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

31

where F PQ mnU (m, n = 0, 1) and F PQ 11D are scattering coefficients associated with the upper and lower interfaces. The integration is over the rough surface variables xs and zs as well as the wave number variables v0 and w of the scattered wave vector k0 . The superscripts of EPQ s denote P (P = H, V) polarized scattered fields due to Q (Q = H, V) polarized incident fields. The fields expressed by Eqs. (126) and (127) are at the observation point y ≥ h01s :

The position vectors to points on the upper and lower rough interfaces are

The incident and scattered wave vectors are

32

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

Fig. 8. Scattering upon reflections in medium 0 above the upper interface.

In Eqs. (131)–(134) the complex sines and cosines of the incident and scatter angles in medium 1 and 2 are related by Snell’s law:

Equations (126) and (127) contain the expressions for the Fresnel reflection (RP αβ ) and transmission (T P αβ ) coefficients for vertically and horizontally polarized waves, the wave impedance η, and refractive index n (76). The physical interpretations of Eqs. (126) and (127) are illustrated in Figs. 8 to 12 (71,72,75,76). Equation (126) represents scattering due to the upper rough interface, and Eq. (127) represents scattering due to the lower rough interface. The first term on the right-hand side of Eq. (126) associated with the scattering coefficient F PQ 00U accounts for scattering upon reflection from above the rough upper interface (see Fig. 8). The second term in Eq. (126) associated with F PQ 01U accounts for waves that undergo multiple reflections in medium 1 and are scattered upon transmission back to 0 (see Fig. 9). The third term in Eq. (126) associated with F PQ 10U accounts for scattering upon transmission from medium 0 to 1 followed by multiple reflections in medium 1 before wave transmission back to medium 0 (see in Fig. 10). The fourth term in Eq. (126) associated with F PQ 11U accounts for multiple reflections in medium 1 before scattering upon reflection in medium 1 from below the upper interface, followed by multiple reflections in medium 1 before transmission back to medium 0 (see Fig. 11). The single term in Eq. (127) associated with the scattering coefficient F PQ 11D accounts for multiple reflections in medium 1 before scattering upon reflection in medium 1 from above the lower interface, followed by multiple reflections in medium 1 before transmission back to medium 0 (see Fig. 12). It is shown that for uniform layered structures, the full-wave solutions sum up to the classical solutions (71,72,73,74,75). The diffuse scattered fields are evaluated at a point in the far-field region above the upper interface. The stationary phase method is used to evaluate the integrals over the scatter wave vector variables v0 and w in Eqs. (126) and (127). Thus, the scattered far fields at rf (the position vector from origin to the receiver) are

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

Fig. 9. Scattering upon transmission (across upper interface) from medium 1 to medium 0.

Fig. 10. Scattering upon transmission (across upper interface) from medium 0 to medium 1.

expressed as follows:

33

34

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

Fig. 11. Scattering upon reflection (in medium 1) below the upper interface.

Fig. 12. Scattering upon reflection (in medium 1) above the lower interface.

and

where the position vectors to the mean upper and lower surfaces are

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

35

and

The wave vectors associated with scattering in the media are

and the terms associated with multiple bounces in the coating material are

The geometric series expansions appearing in Eqs. (142) and (143) are used whenever H D (xs , zs ) is not constant, in order to perform necessary integrations by parts that explicitly involve the derivative of the rough surface heights (75). The normalization coefficients are

For parallel stratified structures (no roughness), the full-wave solutions reduce to the exact, classical solution. The solutions for the like- and cross-polarized diffuse scattered fields presented here can be applied to scattering from irregular layered media with arbitrarily varying rough interfaces such that the thickness of

36

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

the intermediate layer is also arbitrary when random rough surfaces are considered. The rough surface height probability density functions are characterized by a family of gamma functions rather than the standard Gaussian probability density functions to ensure that H D (xs , zs ) ≥ 0 (78). The polarimetric solutions can be applied to remote sensing of dielectric coating materials on rough surfaces. In particular, it is possible to determine the optimal polarizations of the transmitter and receiver such that the presence of clutter from the rough interfaces can be suppressed, in order to facilitate the detection of buried mines for example.

Acknowledgments The manuscript was prepared by Ronda Vietz and Dr. Dana Poulain in the Center for Electro-Optics.

BIBLIOGRAPHY 1. S. O. Rice, Reflection of electromagnetic waves from a slightly rough surface, Commun. Pure Appl. Math., 4: 351–378, 1951. 2. G. R. Valenzuela, Scattering of electromagnetic waves from a tilted slightly rough surface, Radio Sci., 3 (11): 1051–1066, 1968. 3. E. Bahar, B. S. Lee, Full wave solutions for rough surface bistatic radar cross sections: Comparison with small perturbation, physical optics, numerical, and experimental results, Radio Sci., 29 (2): 407–429, 1994. 4. P. Beckmann, A. Spizzichino, The Scattering of Electromagnetic Waves from Rough Surfaces, New York: Macmillan, 1963. 5. D. E. Barrick, W. H. Peake, A review of scattering from surfaces with different roughness scales, Radio Sci., 3 (8): 865–868, 1968. 6. J. T. Johnson et al., Backscatter enhancement of electromagnetic waves from two dimensional perfectly conducting random rough surfaces: A comparison of Monte Carlo simulations with experimental data, IEEE Trans. Antennas Propag., 44 (5): 748–756, 1996. 7. J. W. Wright, A new model for sea clutter, IEEE Trans. Antennas Propag., AP-16 (2): 217–223, 1968. 8. G. S. Brown, Backscattering from a Gaussian-distributed perfectly conducting rough surface, IEEE Trans. Antennas Propag., AP-28: 943–946, 1978. 9. G. L. Tyler, Wavelength dependence in radio wave scattering and specular-point theory, Radio Sci., 11 (2): 83–91, 1976. 10. E. R. Mendez, K. A. O’Donnell, Observation of depolarization and backscattering enhancement in light scattering from Gaussian random surfaces, Opt. Commun., 61 (2): 91–95, 1987. 11. A. A. Maradudin, E. R. Mendoz, Enhanced backscatter of light from weakly rough random metal surfaces, Appl. Opt., 32 (19): 3335–3343, 1993. 12. E. Bahar, M. El-Shenawee, Vertically and horizontally polarized diffuse multiple scatter cross sections of one dimensional random rough surfaces that exhibit enhanced backscatter-full wave solutions, J. Opt. Soc. Amer. A, 11 (8): 2271–2285, 1994. 13. E. Bahar, M. El-Shenawee, Enhanced backscatter from one-dimensional random rough surfaces: Stationary-phase approximations to full wave solutions, J. Opt. Soc. Amer., 12 (1): 151–161, 1995. 14. J. C. Daley, W. T. Davis, N. R. Mills, Radar sea return in high sea states, Nav. Res. Lab. Rep., 7142: 1970. 15. E. Bahar, M. A. Fitzwater, Like and cross polarized scattering cross sections for random rough surfaces—Theory and experiment, J. Opt. Soc. Amer., Spec. Issue Wave Propag. Scattering Random Media, 2 (12): 2295–2303, 1985. 16. E. Bahar, Depolarization of electromagnetic waves excited by distribution of electric and magnetic sources in inhomogeneous multilayered structures of arbitrarily varying thickness—Generalized field transforms, J. Math. Phys., 14 (11): 1502–1509, 1973. 17. E. Bahar, Depolarization of electromagnetic waves excited by distribution of electric and magnetic sources in inhomogeneous multilayered structures of arbitrarily varying thickness—Full wave solutions, J. Math. Phys., 14 (11): 1510–1515, 1973. 18. E. Bahar, Depolarization in nonuniform multilayered structures—Full wave solutions, J. Math. Phys., 15 (2): 202–208, 1974.

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

37

19. S. A. Schelkunoff, Generalized telegraphists’ equations for waveguides, Bell Syst. Tech. J., 31: 784–801, 1952. 20. S. A. Schelkunoff, Conversion of Maxwell’s equations into generalized telegraphists’ equations, Bell Syst. Tech. J., 34: 995–1045, 1955. 21. E. Bahar, Propagation of radio waves in a model nonuniform terrestrial waveguide, Proc. Inst. Electr. Eng., 113 (11): 1741–1750, 1966. 22. E. Bahar, Generalized scattering matrix equations for waveguide structures of varying surface impedance boundaries, Radio Sci., 2 (3): 287–297, 1967. 23. E. Bahar, Wave propagation in nonuniform waveguides with large flare angles and near cutoff, IEEE Trans. Microw.Theory Tech., MTT-16 (8): 503–510, 1968. 24. E. Bahar, Fields in waveguide bends expressed in terms of coupled local annular waveguide modes, IEEE Trans. Microw. Theory Tech., MTT-17 (4): 210–217, 1969. 25. E. Bahar, G. Govindarajan, Rectangular and annular modal analyses of multimode waveguide bends, IEEE Trans. Microw. Theory Tech., MTT-21 (15): 819–824, 1973. 26. S. W. Maley, E. Bahar, Effects of wall perturbations in multimode waveguides, J. Res. Natl. Bur. Stand., 68D (1): 35–42, 1964. 27. E. Bahar, Computations of mode scattering coefficients due to ionospheric perturbation and comparison with VLF radio measurements, Proc. Inst. Electr. Eng., 117 (4): 735–738, 1970. 28. E. Bahar, G. Crain, Synthesis of multimode waveguide transition sections, Proc. Inst. Electr.Eng., 115 (10): 1395–1397, 1968. 29. E. Bahar, J. R. Wait, Propagation in a model terrestrial waveguide of nonuniform height, theory and experiment, J. Res. Natl. Bur. Stand., 69D (11): 1445–1463, 1965. 30. E. Bahar, Propagation of VLF radio waves in a model earth ionosphere waveguide of arbitrary height and finite surface impedance boundary: Theory and experiment, Radio Sci., 1 (8): 925–938, 1966. 31. E. Bahar, J. R. Wait, Microwave model techniques to study VLF radio propagation in the earth ionosphere waveguide, in J. Fox (ed.), Quasi-Optics, New York: Interscience, 1964, pp. 447–464. 32. E. Bahar, Propagation in a microwave model waveguide of variable surface impedance: Theory and experiment, IEEE Trans. Microw. Theory Tech., MTT-14 (11): 572–578, 1966. 33. E. Bahar, Analysis of mode conversion in waveguide transition section with surface impedance boundaries applied to VLF radio propagation, IEEE Trans. Antennas Propag., AP-16 (6): 673–678, 1968. 34. J. R. Wait, E. Bahar, Simulation of curvature in a straight model waveguide, Electron. Lett., 2 (10): 358, 1966. 35. E. Bahar, Scattering of VLF radio waves in the curved earth ionosphere waveguide, Radio Sci., 3 (2): 145–154, 1968. 36. E. Bahar, Inhomogeneous dielectric filling in a straight model waveguide to simulate curvature of waveguide boundaries, Proc. Inst. Electr. Eng., 116 (1): 84–86, 1969. 37. D. E. Kerr, Propagation of Short Radio Waves, MIT Radiat. Lab. Ser. 13, New York: McGraw-Hill, 1951. 38. E. Bahar, Radio wave propagation over a rough, variable impedance, boundary, Part I. Full wave analysis, IEEE Trans. Antennas Propag., AP-20 (3): 354–362, 1972. 39. E. Bahar, Radio wave propagation over a rough, variable impedance, boundary, Part II. Full wave analysis, IEEE Trans. Antennas Propag., AP-20 (3): 362–368, 1972. 40. G. A. Schlak, J. R. Wait, Electromagnetic wave propagation over a nonparallel stratified conducting medium, Can. J. Phys., 45: 3697–3720, 1967. 41. M. J. Kontorowich, N. M. Lebedev, Kontorowich Lebedev Transforms, Academy of Science USSR, J. Phys., 1: 229–241, 1939. 42. E. Bahar, Generalized Bessel transform and its relationship to the Fourier, Watson and Kontorowich–Lebedev transforms, J. Math. Phys., 12 (2): 179–185, 1971. 43. R. J. King, C. H. Husting, Microwave surface impedance measurements of a dielectric wedge on a perfect conductor, Can. J. Phys., 49: 820–830, 1971. 44. E. Bahar, Generalized Fourier transform for stratified media, Can. J. Phys., 50 (24): 3123–3131, 1972. 45. E. Bahar, Radio wave propagation in stratified media with nonuniform boundaries and varying electromagnetic parameters—Full wave analysis, Can. J. Phys., 50 (24): 3132–3142, 1972. 46. E. Bahar, Electromagnetic wave propagation in inhomogeneous multilayered structures of arbitrary thickness— Generalized field transforms, J. Math. Phys., 14 (8): 1024–1029, 1973.

38

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

47. E. Bahar, Electromagnetic wave propagation in inhomogeneous multilayered structures of arbitrary thickness—Full wave solutions, J. Math. Phys., 14 (8): 1030–1036, 1973. 48. E. Bahar, Generalized WKB method with applications to problems of propagation in nonhomogeneous media, J. Math. Phys., 8 (9): 1735–1746, 1967. 49. E. Bahar, Propagation of radio waves over a nonuniform layered medium, Radio Sci., 5 (7): 1069–1076, 1970. 50. E. Bahar, Radiation from layered structures of variable thickness, Radio Sci., 6 (12): 1109–1116, 1971. 51. E. Bahar, Radiation by a line source over nonuniform stratified earth (with G. Govindarajan), J. Geophys. Res., 78 (2): 393–406, 1973. 52. E. Bahar, Radio wave propagation in nonuniform multilayered cylindrical structures—Generalized field transforms, J. Math. Phys., 15 (11): 1977–1981, 1974. 53. E. Bahar, Radio wave propagation in nonuniform multilayered cylindrical structures—Full wave solutions, J. Math. Phys., 15 (11): 1982–1986, 1974. 54. E. Bahar, Field transforms for multilayered cylindrical and spherical structures of finite conductivity, Can. J. Phys., 53 (11): 1078–1087, 1975. 55. E. Bahar, Propagation in irregular multilayered cylindrical structures of finite conductivity—Full wave solutions, Can. J. Phys., 53 (11): 1088–1096, 1975. 56. E. Bahar, Electromagnetic waves in irregular multilayered spheroidal structures of finite conductivity–Full wave solutions, Radio Sci., 11 (2): 137–147, 1976. 57. E. Bahar, Computations of the transmission and reflection scattering coefficients in an irregular spheroidal model of the earth-ionosphere waveguide, Radio Sci., 15 (5): 987–1000, 1980. 58. E. Bahar, Radio waves in an irregular spheroidal model of the erath ionosphere waveguide (with M. A. Fitzwater), IEEE Trans. Antennas Propag., AP-28 (4): 591–592, 1980. 59. J. R. Wait, Waves in Stratified Media, New York: Macmillan, 1962. 60. E. Bahar, M. A. Fitzwater, Numerical technique to trace the loci of the complex roots of characteristic equations in mathematical physics, SIAM J. Sci. Stat. Comput., 2 (4): 389–403, 1981. 61. R. E. Collin, Electromagnetic scattering from perfectly conducting rough surfaces (a new full wave method), IEEE Trans. Antennas Propag., AP-40 (12): 1416–1477, 1992. 62. E. Bahar, Full wave solutions for the depolarization of the scattered radiation fields by rough surfaces of arbitrary slope, IEEE Trans. Antennas Propag., AP-29 (3): 443–454, 1981. 63. M. L. Sancer, Shadow corrected electromagnetic scattering from randomly rough surface, IEEE Trans. Antennas Propag., AP-17: 577–585, 1969. 64. E. Bahar, B. S. Lee, Radar scatter cross sections for two dimensional random rough surfaces—Full wave solutions and comparisons with experiments, Waves Random Media, 6: 1–23, 1996. 65. E. Bahar, Scattering cross sections for composite random surfaces—Full wave analysis, Radio Sci., 16 (6): 1327–1335, 1981. 66. E. Bahar, Y. Zhang, A new unified full wave approach to evaluate the scatter cross sections of composite random rough surfaces, IEEE Trans. Geosci. Remote Sens., 34 (4): 973–980, 1996. 67. E. Bahar, Y. Zhang, Numerical solutions for the scattered fields from rough surfaces using the full wave generalized telegraphists’ equations, Int. J. Numer. Model., 10: 83–99, 1997. 68. E. Bahar, Excitation of lateral waves and the scattered radiation fields by rough surfaces of arbitrary slope, Radio Sci., 15 (6): 1095–1104, 1980. 69. E. Bahar, Excitation of surface waves and the scattered radiation fields by rough surfaces of arbitrary slope, IEEE Trans. Microw. Theory Tech., MTT-28 (9): 999–1006, 1980. 70. E. Bahar, B. S. Lee, Transmission scatter cross sections across two-dimensional random rough surfaces—Full wave solutions and comparison with numerical results, Waves Random Media, 6: 25–48, 1996. 71. E. Bahar, Physical interpretation of the full wave solutions for the electromagnetic fields scattered from irregular stratified media, Radio Sci., 23 (5): 749–759, 1988. 72. S. M. Haugland, Scattering of electromagnetic waves from coated rough surfaces full wave approach, Thesis, University of Nebraska—Lincoln, 1991. 73. E. Bahar, S. M. Haugland, A. H. Carrieri, Full wave solutions for Mueller matrix elements used to remotely sense irregular stratified structures, Proc. IGARSS ’91 Remote Sens.: Global Monit. Earth Manage., Espoo, Finland, Vol. 1, 1991, pp. 1479–1482.

RADAR REMOTE SENSING OF IRREGULAR STRATIFIED MEDIA

39

74. S. M. Haugland, E. Bahar, A. H. Carrieri, Identification of contaminant coatings over rough surfaces using polarized IR scattering, Appl. Opt., 31 (19): 3847–3852, 1992. 75. E. Bahar, M. Fitzwater, Full wave physical models of nonspecular scattering in irregular stratified media, IEEE Trans. Antennas Propag., AP-S 37 (12): 1609–1616, 1989. 76. R. D. Kubik, E. Bahar, Electromagnetic fields scattered from irregular layered media, J. Opt. Soc. Amer. A, 13 (10): 2050–2059, 1993. 77. E. Bahar, Full wave-co-polarized non specular transmission and reflection scattering matrix elements for rough surfaces, J. Opt. Soc. Amer. A, 5: 1873–1882, 1988. 78. R. D. Kubik, E. Bahar, Radar polarimetry applied to scattering from irregular layered media, J. Opt. Soc. Amer. A, 15: 2060–2071, 1996.

EZEKIEL BAHAR University of Nebraska-Lincoln

RADAR ALTIMETRY ALTIMETRY, RADAR Satellite-based radar altimetry over the world’s oceans is the main theme of this article. Rather than measure the unknown clearance of the radar above potentially hazardous topography (which is one rationale for an aircraft radar altimeter for example), satellite-based altimeters are designed to measure the height of the ocean’s surface relative to an objective reference such as the Earth’s mean ellipsoid. Such sea surface height measurements have become essential for a wide variety of applications in oceanography, geodesy, geophysics, and climatology [1]. A satellitebased altimeter circles the Earth in about 90 minutes, generating surface height measurements along its nadir track. These measurements accumulate, providing unique synoptic data that have revolutionized our knowledge and understanding of both global and local phenomena, from El Nino ˜ to bathymetry. A satellite-based radar altimeter also provides measurements of signiﬁcant wave height and wind speed along its nadir track. Although one might view these altimeters as relatively simple instruments, their phenomenal measurement accuracy and precision requires elegant microwave implementation and innovative signal processing. This article provides an overview of the applications that drive these requirements, and a description of the resulting state-ofthe-art design concepts. A nadir-viewing altimeter in a repeat-track orbit is constrained by a fundamental trade-off between temporal coverage (revisit period D days) and spatial coverage (track separation at the equator W kilometers): DW = constant for a given inclination and altitude. If more than one altimeter is under consideration, either as independent assets or as a pre-planned constellation, then the space/time trade-space is enlarged, and more measurement objectives may be satisﬁed. The limitations imposed by this constrain have motivated “multi-beam” or “wide swath” altimeter concepts, although all such architectures imply a compromise on height measurement accuracy. The leading example of this genre is reviewed at the end of this article. The sea surface height (SSH) measurement objectives of space-based altimeters can be grouped into three broad categories: large-scale dynamic sea surface topography, mesoscale oceanic features, and the cryosphere – nearpolar sea ice and continental ice sheets. Satellite altimeters dedicated to determining the ocean’s large scale dynamic surface topography are characterized by absolute sea surface height measurement accuracy on the order of centimeters along tracks of more than 1000 km, and orbits that retrace their surface tracks every 10 to 20 days. In contrast, mesoscale missions focus on sea surface height signals of less than ∼300 km in length. This application requires measurement precision sufﬁcient to sustain relative height measurements, and for geodetic data, relatively dense track-to-track spacing. Geosat is the leading example of this category, for both geodetic (non-repeat) and mesoscale (exact-repeat) orbits. Observation of oceanic

and polar ice sheets requires that the altimeter have robust range and spatial resolution, accuracy, and precision in response to the non-zero average surface slope in both the along-track and cross-track direction of the continental glaciers. Suitable orbits must have near-polar inclination, and multi-year relative accuracy. CryoSat is reviewed as the ﬁrst example of this class of radar altimeter mission. Radar altimeters must provide accurate and precise SSH measurements from a spacecraft whose roll and pitch attitudes are not known exactly. These requirements can be satisﬁed by the pulse-limited altimeter paradigm, which is characterized by (1) large time-bandwidth pulse modulation, (2) antenna directivity that illuminates a surface area larger than the spatially-resolved footprint, and (3) extensive non-coherent (post-detection) waveform averaging. The design of the TOPEX altimeter is described as an example. Footprint resolution and measurement precision can be improved by combining coherent and increased incoherent processing, exempliﬁed by the delay-Doppler altimeter, which borrows applicable techniques from synthetic aperture radar (SAR). The article closes with an overview of future developments and advanced mission concepts. RADAR ALTIMETER SATELLITES All satellite radar altimeters to date (Table 1) are incoherent pulse-limited instruments, as described in a later passage. Since 1973 height measurement accuracy has improved, due primarily to dedicated effort and increasing skill applied to estimation and correction of systematic errors. Performance also has beneﬁtted from improved onboard hardware and algorithms, and improved orbit determination. The Jason-1 altimeter represents the state-ofthe-art in absolute sea surface height measurement accuracy (as of the year 2006). On-line access to descriptions of most of these radar altimeter missions may be found at [2]. Orbits An altimeter’s SSH accuracy on large scales depends to ﬁrst order on how well the height of the altimeter itself can be determined. Given the state-of-the-art in satellite tracking systems, the dominant error in satellite (radial) position determination is uncertainty in knowledge of the gravity ﬁeld (often expressed in term of geoid height) [3]. At lower orbit altitudes, the higher-frequency components of the gravity ﬁeld are enhanced. The impact can be signiﬁcant. For example, gravity variations of about 400 km wavelength are 100 times larger at an altitude of 500 km than they are at 1000 km. In general, the accuracy of precision orbit determination is better for higher altitudes. Atmospheric drag is approximately ten times larger at 800 km than at 1200 km [4]. For example, over one orbit at 1200 km altitude, drag imposes a 1-cm decay on the orbit radius. At 800 km altitude, the effect is ten times larger, resulting in a 10-cm decay per orbit. Atmospheric drag increases signiﬁcantly during periods of higher solar ﬂare activity, the peaks of which occur approximately every eleven years.

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright © 2007 John Wiley & Sons, Inc.

2

Altimetry, Radar Table 1. Summary of Satellite Radar Altimeters Altimeter

AgencyYearOrbitInclination Repeat (days)(degrees)

Altitude (km)

Equatorial Spacing (km)

Band

Propagation Measurements

Accuracy

Skylab (3) GEOS-3 Seasat Geosat ERS-1 TOPEX1 Poseidon1 ERS-2 GFO RA-2 Jason-1 Jason-2 CryoSat

NASA1973No∼48 NASA1975–8No115 NASA1978∼17,3108 USN1985–9∼3, 17.05108 ESA1991–73, 35, 17698.5 NASA1992 –9.91666 CNES1992 –9.91666 ESA1995 –3598.5 USN1998 –17.05108 ESA2002 –3598.5 CNES2001 –9.91666 CNES(2008)9.91666 ESA(2009)36992

435 845 800 800 785 1336 1336 781 800 800 1336 1336 720

– ∼60 160, 800 ∼4, 160 20 – 800 315 315 80 160 80 315 315 –

Ku Ku Ku Ku Ku C, Ku Ku Ku Ku S, Ku C, Ku C, Ku Ku

None None H2 O None H2 O H2 O, e− H2 O H2 O H2 O H2 O, e− H2 O, e− H2 O, e− None

50 m 50 cm 20 cm 10 cm 7 cm 2 cm 5 cm 7 cm 5 cm 7 cm 1.5 cm (1.5 cm) (5 cm)

Whereas a 10-cm decay in radius per orbit may not seem like much for a satellite at 800 km altitude, these decreases accumulate. Smaller orbit radii induce higher spacecraft velocities. Thus, orbit decay accelerates the satellite, shifting its ground track away from its exact repeat path. An altimeter’s repeat pattern can be maintained only by replacing the energy removed by drag forces. Active intervention is required, usually in the form of thruster ﬁrings of controlled strength, duration, and direction. Orbit maintenance maneuvers are required more frequently when the altimeter’s orbit is subject to larger perturbations. Review of Missions The ﬁrst satellite radar altimeter was the proof-of-concept S-193 instrument (General Electric) that ﬂew on three Skylab missions. The objectives were: to verify predicted waveform response to wind and waves, to measure the radar cross section of the sea at vertical incidence, to measure inter-pulse correlation properties, and to observe the effect of off-nadir antenna orientation. Geos-3 (General Electric) provided the ﬁrst geodetic and geophysical results of signiﬁcance within the National Geodetic Satellite Program, including the ﬁrst maps of sea level variability and the marine geoid. Geos-3 and the S-193 altimeters used conventional pulse compression techniques. The Seasat altimeter (Johns Hopkins University Applied Physics Laboratory) was the ﬁrst to use full deramp pulse compression, which opened the way for the very small range resolution required for many oceanographic applications. The deramp technique (described below) has been adopted by all radar altimeters since then. Seasat was designed to measure global ocean topography and the marine geoid, as well as wave height and surface wind speed. The Geosat (Johns Hopkins University Applied Physics Laboratory) altimeter’s design was patterned closely after that of the Seasat altimeter. Geosat was a U. S. Navy military satellite whose primary mission was to map the Earth’s marine geoid to then-unprecedented accuracy (drifting orbit). Since their declassiﬁcation in 1995, data from the ﬁrst 18 month geodetic mission have become the backbone of the global bathymetric chart that is the industry standard [5, 6]. Geosat’s secondary mission was to observe dynamic oceanographic phenomena, for which it was maneuvered into an exact repeat orbit (period 17.05 days) [7]. The

Geosat Follow-On (GFO) altimeter (E-Systems) is meant to replicate as much as possible the Geosat exact repeat mission, leading towards an operational capability for the U. S. Navy. There has been no dedicated geodetic radar altimeter mission since Geosat, although a new mission known as Abyss-Lite is being actively promoted. In the late 1980s, planning for satellite radar altimeter missions split into two themes, determined by the relative priority of their measurements. If the altimeter is the prime payload instrument, then the orbit and mission design can be optimized accordingly. This theme was initiated by TOPEX/Poseidon (T/P), a joint United States (NASA) and French (CNES) mission. TOPEX (Johns Hopkins University Applied Physics Laboratory) was designed to measure and map the dynamic ocean topography with sufﬁcient accuracy to determine large-scale circulation patterns [1]. TOPEX’ most famous contribution is early observation and near-real-time monitoring of El Nino ˜ events, whose height signature over the equatorial eastern Paciﬁc ocean typically is an increase on the order of 10-20 cm with respect to the mean. Poseidon (Alcatel Espace), contributed by France, is a small proof-of-concept instrument that has a solid-state transmitter. Poseidon is the precursor of the Jason altimeters, and the SIRAL instrument aboard CryoSat. Cryosat (described below) will be the ﬁrst radar altimeter designed to observe polar and continental ice sheets from space. The T/P orbit repeat period (9.916 days) was chosen carefully to satisfy adequate observation of the dominant aliased tidal constituents. All solar tidal constituents would be ambiguous with other height signals if the repeat period were an integral number of days (20). For T/P, the time of day for each subsequent observation slips by about two hours. The T/P repeat pass footprint location accuracy is better than ± 1 km, a requirement that is bounded by the cross-track gradient of the oceanic geoid. The T/P instrument package includes a three-frequency radiometer to measure and compensate for propagation delays due to atmospheric moisture (H2 O). TOPEX is the ﬁrst altimeter to use two frequencies to estimate and compensate for propagation delays imposed by ionospheric electrons (e-). The Jason-1 altimeter (Alcatel Espace) [8] is designed to follow in the footsteps of TOPEX, ﬁguratively and literally. Following the launch of Jason-1 into the T/P orbit, TOPEX

Altimetry, Radar

3

was maneuvered into a “tandem” phasing so that the measurements of the two altimeters could be cross-calibrated. The follow-on mission Jason-2 will be identical to Jason-1, and may also include an experimental wide swath ocean altimeter (outlined in the closing sections of this article). If the altimeter is not the primary payload, then the resulting mission and orbit are likely to be determined by other requirements, which may compromise altimetry. The European Space Agency’s satellite altimeters (Selenia Spazio) on ERS-1 and ERS-2, as well as the advanced radar altimeter RA-2 [9] (Alenia Spazio) on ESA’s Envisat, are of second priority with respect to the other instruments on their respective spacecraft. Their sun-synchronous orbits are less than optimum for precision altimetry. The orbit of ERS-1 was adjusted during its mission to a long repeat period (176 days). That long repeat-period generated a relatively dense surface sampling grid useful for estimating sea ice cover, geodesy, and bathymetry, but is less than optimum for most other applications.

GLOBAL DYNAMIC TOPOGRAPHY-ACCURACY The principal objective of an oceanographic satellite radar altimeter designed to observe the dynamic sea surface topography over very large spatial scales is to measure the absolute height hs of the sea surface (Fig. 1) with respect to the standard reference ellipsoid. The key word here is accuracy: the mean bias error of the measurements with respect to an absolute reference. Height measurement accuracy depends, among other factors, upon the accuracy of accounting for variations in the speed of microwave propagation between the radar and the surface. The absolute SSH measurement problem is challenging because the geophysical signal is small, at most on the order of tens of centimeters, yet that signal has to be derived from a satellite altimeter at altitude of 1400 km or so, whose raw range measurement is subject to corrections as large as several tens of meters, and corrections for the geoid of many tens of meters. The accuracy of estimating the dynamic topographic signal is limited by the corrections for variations in the speed of light (nominally c) and other perturbations, as well as the height accuracies of the orbit, marine geoid, tides, and atmospheric pressure. The implied errors have been reduced over the years, after considerable focused effort. State-of-the-art height accuracy (Jason-1) is better than 3 cm (10 day average) or 1.5 cm (1 month average computed with multi-orbit crossover data), which is a remarkable achievement. The mean sea level, governed primarily by the marine geoid, differs from the reference ellipsoid by ± 50 m or more, approaching +100 m in parts of the Indian Sea. Often, the geophysical signal of interest is the dynamic topography ξ, deﬁned as the distance between the marine geoid hG and the physical sea surface, corrected for systematic offsets due to tides and atmospheric pressure, for example. The dynamic topography would be zero if the sea were at rest relative to the Earth. The dynamic topography reﬂects small surface slopes associated with geostrophic currents, of which the Gulf Stream is a well-known example. Crossstream surface slopes are proportional to the mean current

Figure 1. A satellite-borne radar altimeter measures the round trip time delay of transmitted signals, from which is deduced the altimetric height h between the satellite’s orbit and the reﬂecting surface. For most geophysical interpretations, the altimetric height is converted to the surface’s height, which is described using the standard ellipsoid of the Earth as the reference.

ﬂow rate; the resulting slope signals are indicative of largescale oceanic circulation patterns. The altimeter’s measurements Whereas the objective is determination of the distance between the radar and the sea surface, the altimeter actually measures round-trip delay tT . The altimeter’s relative height h is derived from the measured time delay by h = tT c/2, where c is the speed of light. At the accuracy required of an oceanographic altimeter, this deceptively simple proportionality must take into account the small but signiﬁcant retardation of the radar’s microwaves as they propagate through the atmosphere and the ionosphere. In addition to sea surface height, the satellite radar altimeter’s waveform supports two other oceanographic measurements: signiﬁcant wave height (SWH), and surface wind speed (WS). Over a quasi-ﬂat sea, a pulse-limited altimeter’s idealized mean waveform is a step function, whose rise time is equal to the compressed pulse length, and whose position on the time-delay axis is determined by the altimeter’s height (Fig. 2). If the sea surface is modulated by gravity waves, the altimetric depth of the surface increases, which reduces the slope of the waveform’s leading edge. Hence, SWH is proportional to the waveform rise time. If the sea surface is under stress from wind, the resulting ﬁne-scale roughness decreases the power of the pulse reﬂected back to the altimeter. Hence,WS is inversely related to mean waveform power. In practice, the inﬂections of the idealized ﬂat-surface response function waveform are softened by the pulse weighting, and the waveform plateau is attenuated over time by the weighting of

4

Altimetry, Radar

Figure 2. A pulse-limited altimeter’s radiated signal intersects the wavy surface of the ocean from above (a). The output (averaged) waveform is the altimeter’s response to the surface (b). Waveform (round-trip) time delay tT and leading edge slope indicate height above the surface and the large scale roughness (SWH) of the surface, respectively.

the antenna pattern. To extract SWH and WS from waveform data, ﬁnely tuned algorithms have been developed and validated against in situ buoy measurements. For example, the TOPEX Ku-band altimeter measures SWH to within ± 0.5 m up to more than 5.0 m, and WS within ±1.5 m/s up to more than 15 m/s. These ﬁgures correspond to averages over 1 second, or about 6 km along the sub-satellite path of the altimeter’s footprint, which typically is 3 km–5 km wide, determined by mean sea state. Height Error budget The ultimate accuracy of an altimeter depends critically on estimation and removal of the systematic errors inherent to the measurement. Once recovered from the altimetry data, sea surface height hS is used to derive the signal ξ of the dynamic sea surface topography by ξ = hS − hG − δhS in which the independent variables imply geophysical corrections: geoid determination (hG ) and Earth tides, oceanic tides, inverse barometer corrections, etc (δhS ). An oceanographic altimeter collects radar ranging data that are reduced to sea surface height hS according to hS = hO − h − δhA − δhB − δhh

(1)

where the last three terms on the right hand side of Eq. (2) entail corrections to be derived from electromagnetic (EM) reﬂection and propagation phenomena. Orbit radial height (hO ) is determined through extensive instrumentation and analysis, with a net uncertainty. The magnitude of the un-

corrected height errors, and the TOPEX post-compensation residual height uncertainties that remain in hS , are summarized in Table 2. Instrument corrections δhA Height errors that arise in the altimeter and spacecraft environment may be driven close to zero by careful design and calibration [10, 11]. Range delay has to be adjusted to account for the electronic distance from the antenna phase center, through the transmitter, receiver and processor, to the satellite’s center of mass. The required timing correction is a function of spacecraft attitude, temperature, and age, among other perturbations. The waveform leading edge delay is tracked dynamically on-board, but not perfectly. After calibration and compensation, the combined equivalent distance root-sum-squared (RSS) uncertainty of TOPEX from all on-board sources is about 3.7 cm, less than two Ku-band wavelengths. Surface corrections δhB The radiated wave front impinges on the sea surface, and then is reﬂected. In the presence of waves, the mean EM surface sensed by the reﬂection deviates from the physical sea surface. The resulting height measurement bias occurs because ocean waves tend to reﬂect more strongly from their troughs than from their crests. Ocean waves also tend to be asymmetrical in their height distribution, described by skewness. This causes the area of the reﬂecting surface that lies above the mean sea level to be larger than the area below. If scattering were simply proportional to area, this skewness would bias the height measurement. Both EM bias and skewness bias are reduced through empirically derived algorithms that depend on the local signiﬁcant wave height. Propagation corrections δhh EM propagation is retarded by free electrons in the ionosphere, by the air mass of the troposphere, and by the water content of the troposphere. If uncorrected, the height measurement error from these three sources would be about 2.5 m, intolerable for oceanic radar altimetry. The largest error is due to the dry troposphere, but this contribution varies slowly over the planet, and is not problematic. It can be removed almost completely through application of standard models that depend simply on atmospheric pressure and Earth latitude. Atmospheric water content may vary considerably with location. The resulting path length changes, if uncorrected, could imply the presence of large but false oceanic height signals. The only reliable way to counteract its effect is to measure the water content directly along the altimeter’s propagation path. Atmospheric water content may be estimated rather well with radiometric techniques. TOPEX carries a three-frequency (18 GHz, 21 GHz, and 37 GHz) water vapor radiometer (WVR) whose data are used to reduce the wet troposphere path length error to 1.2 cm [12]. The altimeter’s radiation is delayed also by the total count of free electrons (TEC) found in the layer above about 70 km [13]. Depending on solar cycle, solar illumination, etc, the TEC varies widely, causing an apparent increase in path length up to 25 cm. The optical path length delay due to TEC depends on frequency f as f−2. Hence, simultaneous altimeter heights h1 and h2 obtained at two different

Altimetry, Radar

frequencies f1 > f2 can be combined as h=

f12 f12 − f22

h1 −

f22 f12 − f22

h2

to offset the unwanted ionospheric delay error. The TOPEX altimeter operates at Ku-band (13.6 GHz) and C-band (5.2 GHz), generating height estimates hK and hC respectively. The TOPEX algorithm that corrects for the ionospheric path length delay is h = 1.18hK − 0.18hC . Orbit determination h0 The dominant error in absolute height measurements from a radar altimeter is the uncertainty of the satellite’s instantaneous radial distance from the reference ellipsoid. The perturbations on a satellite in low Earth orbit, in order of importance, include: variations in the gravity ﬁeld; radiation pressures; atmospheric pressure; tides, both oceanic and solid Earth; and the troposphere. Real-time observation of the satellite’s orbital perturbations is subject to errors also, compounded by insufﬁcient knowledge of the Earth’s geoid, and location uncertainties for the tracking systems. The principal methods used by T/P for precision orbit determination (POD) rely on its global positioning (GPS) receivers and the Doppler Orbitography and Radiopositioning Integrated by Satellite data (DORIS) system. DORIS instantaneous navigation is better than 4 m on all axes, a tolerance which reduces to less than 5 cm (radial) after precision processing. T/P carries a set of optical retro-reﬂectors (corner cubes) mounted around the circumference of the altimeter antenna. When within view of ground stations equipped with precision range measurement lasers, the corner cubes can be illuminated. The resulting laser ranging measurements are used to calibrate the on-board orbit determination systems. The radial orbit determination error for TOPEX/Poseidon has been reduced to less than 3.5 cm when averaged (RSS) over its orbital repeat period [14]. The predominant errors have 2 cm to 3 cm peaks, concentrated at a once-per-orbit frequency. Orbit determination residuals can be reduced through analysis of height measurements at locations at which the ascending and descending satellite orbits cross. After removal of all geophysical signals and adjustment for changes in the orbit, the measured heights at each cross-over point should be equal. Comparison of the actual cross-over dif-

5

ferentials helps to reﬁne the orbital models. For TOPEX, cross-over analysis of many orbits over a period of 30 to 60 days reduces the residual to about 2 cm. At this level of precision, inaccuracies in the Earth’s tide and geoid models dominate the remaining error. Improvements in these models are expected to lead to 1 cm radial orbit accuracies for the Jason series of altimeters. Such tight orbit determination is unlikely to be achieved in the near future for the radar altimeter satellites at 800 km altitudes. For example, the working objective for POD on GFO is 5 cm, based primarily on GPS tracking and dynamic modeling (although it was better than 7 cm only rarely between 1998–2002). THEORETICAL FOUNDATIONS Altimeters generally fall into one of two kinds, determined by their beamwidth and range resolution. Radar altimeters illuminate the surface through an antenna pattern of width β, which typically is less than a few degrees. As a part of the on-board processing, the received energy is processed into resolved range shells. Beamwidth determines the width βh of the surface illuminated by the antenna. As it intersects the surface, each range of resolved length ρ also determines a surface area of width 2rP [15]. Beamlimited radar altimeters are those for which βh < 2rP . Conversely, 2rP < βh for pulse-limited altimeters. The altimeters cited in Table 1 are all pulse-limited. Pulse-limited altimeters Figure 3 illustrates the pulse-limited condition. The height accuracy of a pulse-limited altimeter is much less sensitive to (small) angular pointing errors than is the case for a beam-limited altimeter. The pulse-limited radius rP of a quasi-ﬂat surface on the Earth of mean radius RE is rP =

cτh/αR

(2)

where αR = (RE + h)/RE is a consequence of the spherical observation geometry. For typical satellite radar altimeters, the pulse-limited footprint over a quasi-ﬂat surface is on the order of two kilometers in diameter. The pulse-limited

6

Altimetry, Radar

as a function of off-boresight angle, λ is radar wavelength, h is height, and the range pulse processing gain (compression ratio) is CR . In the altimetry literature, the radar cross section usually is interpreted to mean σ = σ 0Aσ where σ 0 (sigma-0) is the normalized scattering coefﬁcient (dimensionless) of the terrain, and Aσ is the area of the resolved footprint. The peak power observed within a conventional radar altimeter, at the instant that the pulse-limited area at nadir is fully illuminated, is given from Eq. (3) and Eq. (4) by PP =

PT G2 λ2 CR πcτσ 0 (4π)3 h3 αR

(5)

The power described by Eq. (7) is proportional to compressed pulse length τ, and to the inverse cube of height h−3 . Flat surface response

Figure 3. Elevation (a) and plan view (b) of a pulse-limited radar altimeter’s illumination geometry. The surface area simultaneously illuminated within the duration of the compressed radar pulse length dominates the height measurement. This pulselimited area expands for larger signiﬁcant wave height (SWH).

area is AP = πrP2 =

πcτh αR

(3)

As the pulse continues to impinge and spread over the surface, the resulting pulse-limited annuli all have areas equal to that of the initial pulse-limited footprint. Hence, the received power tends to maintain the level corresponding to the peak of the initial response. The pulse-limited areas expand in response to increasing large-scale surface roughness, which in the oceanographic context is expressed as signiﬁcant wave height SWH. Radiometric response The classical single-pulse radar equation that describes the post-processing peak power P is PT G2 (θ)λ2 CR σ P= (4π)3 h4

(4)

where σ is the effective radar cross section, PT is the transmitted power, G(θ) is the one-way power gain of the antenna

Under reasonable conditions, the expected output g(t) from any linear sensor is given by the convolution g(t) = p(t)∗ s(t) of the sensor’s impulse response p(t) over the distribution s(t) that describes the input data source. A pulse-limited radar altimeter is an example of a linear system, but its response to input data takes on a special form due to the relatively strange geometry through which it views surface height variations. As a function of time, the radiating pulse ﬁrst strikes the surface, and then spreads out over it. The so-called ﬂat surface response is the altimetric counterpart to the generic impulse response function. It serves as the primary analytical basis for description of a radar altimeter’s waveform from a variety of surface topographies. In the altimetry literature two closely related “ﬂat surface” functions appear, denoted here as pI (t) and pF (t). The (idealized) ﬂat surface response pI (t) was introduced originally [15] as a system function to account for the effects of antenna pattern, illumination geometry, and incoherent surface scattering. As an extension, Brown’s [16] ﬂat surface response pF (t) includes the impact of the compressed pulse shape and signal processing as well as the functional dependencies captured in pI (t). Brown’s model is used most widely. The difference between these two ﬂat surface functions is subtle, but signiﬁcant. The average input data distribution presented to a conventional radar altimeter is wellmodeled by s(t) = pI (t)∗ q(t), which is a convolution of the (idealized) ﬂat surface response pI (t) with the topographic distribution q(t) of the surface. The resulting radar altimeter linear model is the convolution gA (t) = pF (t)∗ q(t), where pF (t) = p(t)∗ pI (t) is Brown’s ﬂat surface response function, and p(t) is the conventional linear system impulse response of the altimeter. The output gA (t) is known as the altimeter waveform, examples of which are sketched in Fig. 2. The ﬂat surface response function (averaged) of a conventional satellite radar altimeter is pF (t) = 0 =

t τ

=1

2h 1 [t − ]≤0 τ c 1 2h 0 < [t − ]≤1 τ c 2h 1 ] 1 < [t − τ c

(6)

Altimetry, Radar

7

under the simplifying condition of a perfectly rectangular compressed unity pulse of length τ. The deﬁning characteristic of this response is that it is essentially a step function: following its linear rise over the duration of the compressed pulse, its maximum value is supported for many subsequent delay intervals. In practice, this waveform is attenuated in time due primarily to weighting of the antenna pattern away from boresight. Also, the waveform itself is more rounded, as a consequence of the weighted shape of the compressed pulse produced by a realistic altimeter.

DERAMP ON RECEIVE A satellite-based radar altimeter needs to measure the distance accurately, but only for an essentially planar surface, oriented orthogonally to the radar’s line-of-sight. Conservative design suggests that all radar resources should be concentrated near the reﬂection from that surface. Hence, ocean-viewing altimeters have a small range window that tracks the delay and strength of the surface reﬂection. The ocean’s surface has a signiﬁcant wave height of less than 20 m or so, and its radar backscatter coefﬁcient spans 3 dB to 20 dB, to cite parameters used in the testing of the TOPEX altimeter. In practice, range gate delay and backscatter tracking are met with two servo-regulator feedback loops (Fig. 4). The ﬁrst loop is a second-order height tracker consisting of range position (alpha tracker) and range rate (beta tracker). The second loop is the receiver gain control (AGC). Altimeter height measurement is given by the setting of the range delay coarse and ﬁne values, corrected by the remaining height error measured from the waveform’s position in the tracker. Surface wind speed and signiﬁcant wave height are derived from the AGC values and the waveform’s shape, respectively. The precision of an individual height measurement is determined by range resolution. If a simple short pulse were transmitted, then the height resolution would equal the pulse length. The principal disadvantage of a short pulse is that it contains little energy. The inherent resolution of a pulse is inversely proportional to its bandwidth. Most radar altimeters use some form of modulation on the transmitted signal to maintain a large bandwidth within a longer pulse, thus increasing the transmitted energy at no loss of resolution. A well-established modulation technique used in many airborne radar altimeters is frequencymodulated continuous wave (FM-CW), from which height is proportional to the frequency difference between the transmitted and received signals. An alternative approach is pulse compression, whereby a relatively long large timebandwidth pulse is transmitted, and then processed (compressed) after reception to a simple short pulse of unity time-bandwidth. Satellite-based radar altimeters use a different and specialized form of modulation and demodulation. The relatively distant and narrow range window typical of an ocean-viewing satellite radar altimeter is ideal for the full deramp (stretch) technique [17] which was ﬁrst applied to altimetry by MacArthur [18] in the Seasat altimeter. The deﬁning feature of the full-deramp technique is that the transmitted pulse length is longer that the depth of the

Figure 4. The functional diagram of a modern satellite altimeter is centered on the waveform tracker, whose outputs are: (1) translated into science data to be returned via telemetry, and (2) transformed into closed loop timing and gain controls for the radar.

range window. The full deramp (dechirp) technique employs a transmitted chirp (linear FM signal) of duration TP , bandwidth BP , chirp rate kP , and center frequency f0 . For a pulse initiated at t = 0, the transmitted frequency is f0 + kP t, t ∈ TP , as shown in Fig. 5. The bandwidth is BP = kP TP , and the associated time-bandwidth product is kP TP2 . Pulse bandwidths and the time-bandwidth products for satellite radar altimeters are large, on the order of 300 MHz and 30,000 respectively. The compressed pulse length τ is given by the inverse bandwidth of the transmitted pulse, or alternatively, by the original pulse length TP divided by the time-bandwidth product. Thus, a full deramped altimeter’s height resolution is τ=

1 seconds or k P TP

ρ=

c meters 2 kP TP

(7)

The altimeter tracking system anticipates the time t0 when the reﬂected signals will arrive back at the radar. To meet these, another chirp is generated, at time tc ≈ t0 and center frequency f0 − fI , where fI is to become the receiver’s intermediate frequency (IF). (The deramp chirp time tC in general is slightly different from t0 , as explained in the tracking discussion below.) The deramp chirp is mixed with the incoming signals, after which their difference frequen-

8

Altimetry, Radar

THE TOPEX DESIGN

Figure 5. A long linearly frequency-modulated (chirp) pulse is transmitted (as in Fig. 4), followed by full deramp demodulation on receive to produce a relatively narrow band set of CW data signals at intermediate frequency f1 .

cies are retained, to produce the set of deramped data signals shown in the ﬁgure. The key to many characteristics unique to a radar altimeter lies in this deramp domain. The deramped signal from the mth individual scatterer at time delay tm is a CW segment of length TP and frequency fm = 2k p (tm − tC ),

tm ∈ TR

(8)

where TR is the time spanned by the received signals. For a range window RH meters deep, TR = 2RH /c. The IF bandwidth Bl is determined by the FM rate and the range window, Bl = 2 kP RH /c. If the range depth of the scene is small, as is the case with radar altimeters meant to operate over the ocean, then the corresponding IF bandwidth is small, typically less than 5 MHz. Clearly, the full deramp technique offers a considerable savings in system bandwidth at all subsequent stages, and at no cost in range resolution. The deramped window duration TD ≥ TP + TR must be larger than the pulse duration to accommodate the extra time induced by the range window time span. The full deramp technique works best when TP > TR , a condition that is very well satisﬁed for oceanographic altimeters for which TR is less than 1% of TP . Alert: there is a conceptual pitfall lurking in the deramped signal domain. Time and frequency reverse their customary roles. In this domain, time delay is no longer a measure of the radar range to reﬂectors. Rather, the signals’ time duration TP determines their compressed pulse resolution according to Eq. (7), and thus “time” behaves as a bandwidth. Conversely, each scatterer’s round-trip time delay, relative to the track point, is proportional after deramp to the CW frequency given by Eq. (8), relative to the IF center frequency. Thus “frequency” behaves as delay time, which is proportional to radar range.

The TOPEX Ku-band altimeter [10,11,18] illustrates the essential features of conventional space-based radar altimeters (Fig. 4). The TOPEX altimeter is controlled by a synchronizer whose inputs are derived from the tracker outputs, slaved to a master radio frequency clock at 80 MHz. The tracker and synchronizer control the altimeter in all seven of its operational modes: test, calibrate, stand-by, coarse track acquisition, ﬁne track acquisition, coarse resolution track, and ﬁne resolution track. Table 3 lists values for selected TOPEX parameters. In the ﬁne resolution track mode, the radar transmits a linear-FM (chirp) pulse of length 102.4 µs and bandwidth 320 MHz. The signal generator consists of a digital section that creates 40 MHz chirps at baseband, followed by RF sections that multiply and mix the signals to meet the ﬁnal bandwidth and center frequency. The chirped pulses at 13.6 GHz are ampliﬁed in a traveling wave tube to 20 W (peak), and transmitted at 4.5 kHz pulse repetition frequency (PRF) through an antenna 1.5 m in diameter. The received pulses are ampliﬁed and then mixed with a delayed chirp signal, centered at 13.1 GHz to produce an ensemble of CW data signals spread over a band of ∼3 MHz about an intermediate frequency of 500 MHz. The deramped data signals are further ampliﬁed, subject to gain control in the AGC attenuator, mixed down to in-phase and quadrature video, low-pass ﬁltered and digitized. At this stage, the low-pass ﬁlter has the effect of removing echoes from ranges that are well outside of the desired range gate window. The signals are magnitude-squared, and summed to produce smoothed height waveforms. Extensive waveform averaging over statistically independent samples is essential to suppress the speckle noise that otherwise would dominate the waveforms. The smoothed waveforms are processed to extract the data of interest, and the tracking outputs are fed back to close the control loops of the radar. The Ku-band and the C-band channels are timemultiplexed, which impacts system timing from the PRF to the processor. Sampling and Waveform Processing The mean amplitude of the deramped signals is normalized through an automatic gain control (AGC) ampliﬁer. The receiver gain is set by the level observed in the waveform processor. Signal level is proportional to the scattering coefﬁcient (sigma-0) of the water’s surface, which in turn is a function of surface wind speed. The normalized signals in the TOPEX altimeters are mixed down to in-phase (I) and quadrature (Q) video channels and passed through 625 kHz low-pass ﬁlters prior to analog-to-digital sampling at a 1.25 MHz rate. This produces a set of 128 complex samples uniformly spaced over the 103.2 µs deramp interval TD . The tracker and synchronizer control the position of the altimeter height settings through two paths: coarse, and ﬁne. The coarse height feedback depends on the deramp trigger tC that is slaved to the 80 MHz clock, which has a period of 12.5 ns. Thus, choice of tC is restricted to a set

Altimetry, Radar

9

can be reduced by summing (averaging) many statistically independent waveforms together. Statistical independence between sequential returns observed by a radar altimeter depends primarily on the radar pulse repetition rate (PRF), the antenna size, the spacecraft velocity, and on the sea surface conditions [19]. The pulse-to-pulse statistical independence threshold for TOPEX is about 2.5 kHz, yet its PRF = 4.5 kHz. The pulse rate above the threshold improves the additive SNR, but does not contribute to speckle reduction. Figure 6. The waveform processor (see Fig. 4) of a pulse-limited radar altimeter performs three basic functions: (1) application of a frequency shift to effect the ﬁne range-delay correction, (2) an inverse FFT applied to each return to transform the CW signals into a power distribution as a function of altimetric height (relative to the track point), and (3) summation over many of these individual power waveforms to form the averaged waveform which is sent to the tracker.

of discrete delays separated by 12.5 ns, corresponding to height intervals of 1.875 meters each. The ﬁne height feedback is exercised in the digitized deramp domain (Fig. 6). In the deramp domain, a small frequency shift is equivalent to a small time delay t. For a ﬁne time shift interval −12.5 ns < t < 12.5 ns, the corresponding frequency shift is kP t, bounded by ± 39.0625 kHz. The 25 ns ﬁne time shift adjustment interval is large enough to accommodate anticipated range rates without having to reset the course height feedback selection within one waveform processing cycle. Following frequency shift, the data are inverse Fourier transformed (IFFT) and magnitude-squared detected. The IFFT converts CW to time shift (relative to the track point), and compresses the data to its individual pulse resolution, 0.469 meters. Each resulting waveform is a distribution of power across the bins within the range window (as suggested in Fig. 2). As is true for most radars, the received waveform produced by an individual pulse is corrupted by speckle noise. Speckle is created by the coherent interference within a given echo between unresolved and competing elementary scatterers, and causes the signal’s standard deviation to be large. (In the limit, the standard deviation equals the signal’s mean value for an individual waveform drawn from a Gaussian ensemble). The standard deviation of the speckle

Tracking The TOPEX Ku-band channel averages 228 pulses over a so-called track interval of about 50 ms to produce the smoothed waveforms delivered to the tracker at a 20 Hz rate. For each waveform, the range window (Fig. 2) is partitioned into 128 sample positions or bins, each of size equal to the radar’s range resolution. Groups of bins are organized into tracking gates of various sizes whose outputs are used to calculate the parameters that control the altimeter’s feedback loops, and to provide the ﬁrst-order science data from the instrument [20]. The tracking algorithm, based on an Intel 80186 microprocessor, iterates at the waveform input rate of 20 Hz. Each tracking gate is normalized so that its gain is inversely proportional to its width, which is the number of samples that it spans. The range width of each gate is a power of two times the intrinsic range resolution of the altimeter. The noise gate estimates the mean noise level from samples 5 through 8, which occur well before the waveform begins to respond to surface reﬂections. The mid-point of the waveform’s leading edge is tracked to keep it centered between samples 32 and 33. The AGC gate spans samples 17 through 48, which are centered on bin 32.5, the so-called track point. The output of the AGC gate is fed back to control the altimeter’s gain loop. TOPEX is required to measure waveform power (proportional to sigma-0) with an accuracy of ± 1 dB and a precision of ± 0.25 dB. In response to the waveform levels observed in the AGC gate, the receiver attenuator is adjusted in 1 dB steps. To meet the accuracy and precision requirements, from pulse to pulse the attenuator setting is dithered between neighboring steps. This has the effect of interpolating the mean AGC setting to an effective accuracy of less than 0.1 dB when averaged

10

Altimetry, Radar

over all 228 input waveforms. Waveform power (sigma-0) returned as science data is equal to the mean AGC level plus the AGC tracking gate level. To summarize, the altimeter’s measurements are: SSH – the position of the track point, plus the track-point offset; SWH – proportional to the width of the waveform leading edge of the waveform; and WS – proportional to a function (empirically determined) of 1/(waveform power), hence derived from the AGC setting.

GRAVITY AND BATHYMETRY-PRECISION Radar altimetric data are the basis for state-of-the-art geodesy expressed through the ocean’s surface, and consequently, global bathymetry. The principal objective of a geodetic satellite radar altimeter [21] is to measure the (along-track) slope of the sea surface caused by gravity deﬂections over spatial scales less than a few hundreds of kilometers (Fig. 7). Sea surface slope is derived by taking the difference between two neighboring height measurements, where the slope tangent equals “rise over run”. The key word for these measurements is precision: the standard deviation (noise) of the sea surface height measurement about its mean value. Height measurement precision is determined by the radar altimeter’s post-processing range resolution, and by the amount of averaging available for each estimate. Note that a precision measurement may still have poor accuracy, if its mean value is biased away from the correct value. When comparing two neighboring height measurements, any constant bias is cancelled by differentiation as long as the error is the same for both measurements. The sea surface slope measurement problem is challenging because the desired slope signals are as small as 6 mm height differential (rise) for each 6 km along-track separation (run). Such a slope corresponds to one microradian of gravity deﬂection, or about a one milligal gravity anomaly. In addition to height precision, geodetic altimetry requires smaller along-track resolution than a conventional altimeter, and a suitable orbit. The altimeter’s footprint resolution should be smaller than about 6 km, which corresponds to the minimum half-wavelength scale of the observable gravity anomaly spectrum. The orbit should not repeat for ∼1.2 years, to yield an average ground track spacing of 6 km, again in respect of the anomaly spectrum. The orbit’s inclination should be near 50◦ –63◦ (or 113◦ –120◦ retrograde) to resolve north and east slopes nearly equally, and to cover the lower latitudes where existing data are inadequate. Note that oceanographic radar altimeter missions (TOPEX/Poseidon, Jason-1, ERS1/2, Envisat, and Geosat ERM/GFO) normally are placed into exact-repeat orbits (10 to 35 days), and as a consequence have widely spaced (80 km to 315 km) ground tracks. Such orbits cannot resolve the short-wavelength two-dimensional surface slopes required for useful bathymetry. Since absolute height accuracy is not required, geodetic radar altimeters can be relatively basic instruments. They do not need to compensate for propagation delays, hence they need only one frequency, and they do not need a water vapor radiometer. Indeed, such an instrument is

Figure 7. The ocean’s bottom topography causes subtle variations in the local gravity ﬁeld, which are expressed as small tilts in the ocean’s surface. These are observable by satellite altimetry.

preferred; it has been shown that efforts to correct for path delays usually add noise to slope estimates [22]. Geodetic measurements provided by the Geosat and ERS-1 (both single-frequency altimeters with no WVR) furnish the best resolution oceanic gravity from space to date. Their resulting bathymetric resolution is limited to about 25 km north-south, and poorer resolution of east-west slope components. These results reﬂect the less-than-optimum resolution, waveform precision, and orbit inclination of those two altimeters. Geodetic resolution at the ocean’s surface can be no ﬁner than about 6 km (half wavelength), a limit that is determined by the average depth of the ocean. Gravity anomalies are caused by topographic relief on an interface between two volumes of differing mass density. In the deep ocean, sediments are thin, and the basaltic sea ﬂoor crust is internally ﬂat-layered, and so gravity anomalies at the surface reﬂect the topography of the ocean ﬂoor. Conversely, at continental margins the sea ﬂoor is nearly ﬂat and sediments are generally thick. Beneath these sediments there may be basins or other geologic structures of interest. In such regions, surface slope signals are due pri-

Altimetry, Radar

marily to topographic variations at the interface between crystalline rocks and their sedimentary overburden. The sediment/basement interface provides essential reconnaissance information for petroleum exploration. The correlation between slope and existing depth soundings readily distinguishes these two environments [23]. The slope signals required to estimate bathymetry are band-limited (12 km to 300 km full-wavelength), as determined by fundamental physical principles. Hence, the height measurements of a geodetic altimeter need to maintain relative accuracy–precision–only over this relatively narrow band. Within this band, precision turns out to be the dominant limiting condition. Sea surface slope measurements are derivatives of the altimeter’s natural measurements, height. Taking derivatives eliminates constant and longwave height errors, but it ampliﬁes noise at short wavelengths. Using a simple model in which height errors are assumed to be a Gaussian white noise process over the geodetic band, the one-sigma slope error is about 1.8 ∼rad if the altimeter’s one-sigma height precision is 1 cm for a one-second averaged height value. Experience teaches that height precision degrades with increasing signiﬁcant wave height (SWH). One of the factors that motivated the development of the delay-Doppler approach to radar altimetry was to improve measurement precision.

Pulse-to-pulse coherence requires that the PRF be above the inter-pulse correlation threshold, rather than below it as is normal for conventional incoherent radar altimeters [19]. To assure correlation, the PRF must be high enough so that at least two pulses are emitted while the satellite’s forward motion equals the along-track aperture size of the altimeter’s antenna. Delay-Doppler domain The objective of delay compensation is to remove the extra delay (Fig. 8) that is induced by the spherical curvature of the radar’s ranging wavefront as it impinges on the ocean’s surface. At each along-track angular offset θ D from nadir, there is an extra range distance h = h(sec θ D − 1) due to range curvature. As it is received, the signal includes returns from scatterers at many different angles. Hence, the problem is multi-valued; in the signal domain compensation for wavefront curvature is impossible. This is the situation for conventional incoherent radar altimeters. In the delay-Doppler altimeter, however, along-track coherent processing gets around this dilemma. Transformation from the signal domain to the frequency (Doppler) domain reduces delay compensation to a single-valued problem: at each Doppler frequency fD the extra range-delay increment is unique, and known. The delay increment, in terms of Doppler frequency fD as its independent variable, is

DELAY-DOPPLER The delay-Doppler technique leads to better measurement precision, a smaller effective footprint at nadir, and increased tolerance of along-track surface gradients typical of continental ice sheets. The central innovation in the delay-Doppler concept [24, 25] is that it combines the beneﬁts of coherent and incoherent signal processing, rather than relying exclusively on incoherent averaging as is the case for all conventional satellite radar altimeters. The coherent processing stages, patterned after well-established methods developed for synthetic aperture imaging radar (SAR), allow much more of the instrument’s radiated power to be converted into height measurement data. One consequence of delay-Doppler signal processing is that less transmitted power is required than with a conventional altimeter. The delay-Doppler technique also enjoys the beneﬁts of the pulse-limited range measurement geometry. The coherent processing transforms groups of data into the Doppler frequency domain, where delay corrections are applied, analogous to SAR range curvature correction [26]. Doppler processing determines the size and location of the along-track footprint, which is (1) smaller than the pulselimited diameter, (2) a constant of the system, and (3) relatively immune to surface topographic variations. Waveforms are incoherently summed corresponding to each surface position as the altimeter progresses along track. One direct result is that each height measurement from a delayDoppler altimeter has more incoherent averaging than is possible from a conventional radar altimeter. The delay-Doppler technique exploits coherence between pulses, in contrast to the pulse-to-pulse incoherence that is the norm for conventional pulse-limited altimeters.

11

h( fD ) ≈ αR

λ2 h 2 f 8V 2 D

(11)

where V is the velocity of the spacecraft along its orbit. Recall that the deramped data in the range direction appear as constant (CW) frequencies. Each range delay increment translates into an equivalent CW frequency shift. These unwanted frequency shifts may be nulliﬁed by multiplying the data ﬁeld by equal and opposite CW signals prior to the range IFFT, analogous to the ﬁne tracking frequency shift of a conventional radar altimeter. The result is evident in Fig. 9, which compares the ﬂat surface response waveform (as it would appear in the delay-Doppler domain) before and after delay compensation. Implementation The delay-Doppler altimeter introduces additional alongtrack processing steps (Fig. 10) after the range deramp and before the range IFFT. The net effect of the extra processing is to transform the signal space from one to two dimensions. A Fourier transform is applied to these data in the alongtrack dimension, implemented in real time on-board as a set of parallel FFTs that span the range window width. Signals in the resulting two-dimensional deramp/Doppler domain are phase shifted to eliminate the unwanted range. The delay correction phase functions are ( fD , t) = exp{+ j2πkP

2 h( fD )t} c

(12)

which are CW signals whose frequency is matched to the delay increment given by Eq. (11). The data at this stage consist of an ensemble of two-dimensional CW signals. Fre-

12

Altimetry, Radar

Figure 8. Elevation (a) and post-compensation plan view (b) of a delay/Doppler radar altimeter’s illumination geometry. The resolved along-track footprint XD is Doppler-limited. The crosstrack footprint is pulse-limited. Although the footprint is smaller than the pulse-limited area, more averaging is available at each Doppler bin position.

quency in the time delay direction is proportional to (minimum) delay relative to the range track point, and frequency in the along-track direction is proportional to the scatterer’s along-track position relative to the zero-Doppler position. The remaining data processing is carried out in parallel, consisting of a range IFFT at each Doppler frequency bin, detection, and assignment of the height estimates to their respective along-track positions. The process is repeated over subsequent blocks of data, from which many looks are accumulated at each along-track position. As the altimeter passes over each scatterer, the corresponding height estimates move in sequence from the highest Doppler ﬁlter to each lower frequency ﬁlter, until the scatterer is out of sight. Thus, the ﬁnal waveform at each along-track position is the average (incoherent sum, normalized) of estimates from all Doppler ﬁlters. If the Doppler ﬁlters are designed to span the along-track antenna beamwidth, then all data along-track contribute to the height estimates. The resulting coverage is shown in Fig. 11, which contrasts the “scanning” beam of a conventional altimeter with

Figure 9. Simulated height waveforms, as they would appear in a compressed pulse (delay) and Doppler data array before (a) and after (b) curvature compensation, illustrate how delay/Doppler processing shifts more of the reﬂected energy into the pulse-limited region, thus improving the height estimates.

the “mini-spotlight staring” Doppler-processed beam of a delay-Doppler altimeter. The ﬁgure shows the unfolding coverage of one resolved along-track cell; all illuminated cells are tracked in parallel through the Doppler ﬁlters, in similar fashion. Footprint Delay-Doppler processing may be interpreted as an operation that ﬂattens the radiating ﬁeld in the along-track direction. In this transformed data space (Fig. 8(b)), the (x,y) cells have constant along-track length, but their crosstrack widths decrease as the square root of delay time. The cross-track footprint is determined by the pulse-limited condition. The along-track impulse response is set up by the Doppler ﬁlters. Along-track impulse position is determined by the zero-Doppler position for each burst of data. Alongtrack position can be adjusted by artiﬁcial Doppler shifts to maintain registration of subsequent Doppler bins, which is the along-track analog of the ﬁne height adjustment in an incoherent radar altimeter.

Altimetry, Radar

13

Figure 10. A delay/Doppler radar altimeter waveform processor (see Fig. 4) must be augmented with a cache memory to store the deramped returns from a sequence of transmitted pulses. FFTs are applied across these data to derive their Doppler frequency spectra, which then are corrected for curvature delay by phase multiplication.

Ideally, the along-track zero-Doppler position is equivalent to the geometric sub-satellite point, nadir. The alongtrack location of the zero-Doppler plane is independent of satellite attitude, and also is independent of terrain slope. Thus, the height measurements at all Doppler frequencies can be located along-track with respect to zero Doppler. In practice, the zero Doppler bin location may not coincide with nadir. A vertical spacecraft velocity component adds a Doppler shift to the signals. Vertical velocity and its implied Doppler error can be estimated. Offsetting Doppler shifts can be applied in response to a spacecraft vertical velocity component to assure registration of the Doppler bins with their corresponding along-track positions deﬁned with respect to nadir.

Figure 11. The delay-Doppler altimeter tracks cells resolved along-track (b) through the on-board processor, in contrast to a conventional altimeter (b) that in effect drags its footprint along the surface.

is required. Focus operations would be required if the Fresnel radius were larger than the along-track cell dimension. If a smaller cell size is desired such as for altimetry over land, or a very high satellite altitude or longer radar wavelength were chosen, then the along-track processor would have to incorporate phase matching to focus the data.

Unfocused condition The foregoing is predicated on a simple isometry between Doppler frequency and along-track spatial position. This equivalence is valid for an along-track resolution that is comparable to or larger than the ﬁrst Fresnel zone. In synthetic aperture radar parlance, this zone is known as the unfocused SAR resolution. Using the classic quarter wavelength criterion, the radius a0 of the ﬁrst Fresnel zone is

a0 =

hλ 2

which for a Ku-band altimeter leads to an along-track (unfocused) dimension of 180 m from an altitude of 800 km (or about 230 m from an altitude of 1334 km). As these quantities are less than the nominal delay-Doppler along-track cell size of 250 m, the processing task is trivial: no focusing

Incoherent Averaging There are two stages in a delay-Doppler altimeter at which incoherent averaging takes place: within each Doppler bin, and across neighboring bins. Detected returns from many pulses are averaged together to build the multi-look waveform within each bin. For a typical satellite altimeter, these waveforms would accumulate wihin each 250-m bin at about a 26 Hz rate. Subsequent averaging (incoherent integration) over adjacent waveforms typically extends over 0.1 s (or 1.0 s), during which time the antenna illumination pattern progresses in the along-track direction by an appreciable distance, approximately 0.6 km (or 6 km). Alert: the relative location of each delay-Doppler-derived height estimate is synchronized to coincide with the forward motion of the instrument, thus eliminating along-track elongation of the footprint as is the case for a conventional al-

14

Altimetry, Radar

Figure 12. DDA processing increases the number of independent samples of the surface return, which reduces the intrinsic noise, thus improving the sea surface height measurement precision.

timeter. The result is that a delay-Doppler altimeter generates signiﬁcantly more incoherent averaging than a conventional altimeter, and at less compromise in along-track footprint size (Fig 12). One immediate benﬁt is better measurement precision. Consider the case of height precision in the context of geodetic requirements. Figure 13 shows a plot of height precision versus SWH for a delay-Doppler altimeter and a conventional radar altimeter (RA). The plot shows that the DDA meets the height precision requirement of 1 cm at 3 m SWH, a result that is consistent with previous analyses [27]. The ﬁgure also shows that the DDA is about half as sensitive as an RA to increasing SWH. This is important for geodetic applications, as measurement precision degraded by larger signiﬁcant wave heights is a major source of noise in Geosat surface slope estimates [6].

The ﬂat surface delay time response (after processing) of the delay-Doppler altimeter has the functional form fD (t)

= 0

1 2h [t − ]≤0 τ c 1 2h (13) 0 < [t − ]≤1 τ c 1 2h 1 < [t − ] τ c where τ is the compressed pulse length (Eq. 7). The curve of Eq. 13represents the (average) strength of the altimeter’s response to illumination of a quasi-ﬂat surface as a function of time delay, just as in the conventional case. Note that the response to a ﬂat surface for all regions t < τ have much less relative power for the delay-Doppler altimeter than for the conventional radar altimeter described by Eq. (6). The cross-track (time-delay) width of fD (t) is approximately equal to τ. t = τ t t = − −1 τ τ

Radiometric response Flat surface response The customary concept of ﬂat surface response applies only to the delay time dimension for a delay-Doppler altimeter. This means that the inherent delay/elevation ambiguity characteristic of pulse-limited altimeters is reduced from two spatial dimensions to only one dimension. The cross-track ambiguity that remains is suggested in Fig. 8, which shows that at any given Doppler frequency, there are two possible sources for reﬂections having a given (relative) time delay. These arise from either side of the minimum delay locus, which nominally is the sub-satellite track. Of course, the point of ﬁrst reﬂection (at zero relative delay time) may be to one side of the sub-satellite track, as would be true in general when there is a non-zero cross-track terrain slope. The cross-track ambiguity and the delay/elevation ambiguity both may be at least partially resolved through application of other means such as the monopulse phase sensing technique.

The delay-Doppler altimeter can take advantage of reﬂections from the entire length of the antenna illumination pattern in the along-track direction to estimate the height of each resolved patch of sub-satellite terrain. This implies that substantially more integration is possible than in a pulse-limited altimeter. Under the assumption that the dominant scattering mechanism is non-specular, the integration gain is linear in power. It follows that the total power arising from each resolved cell is larger for the delayDoppler altimeter than for a conventional pulse-limited altimeter, even though the post-processing footprint size is smaller. Height estimation for each resolved scattering cell beneﬁts from integration as long as that cell is illuminated by the antenna pattern. For each scattering cell, the equivalent along-orbit integration is governed by the length βh of the antenna footprint, expanded by the orbital factor αR . The along-orbit integration may be interpreted in terms of

Altimetry, Radar

an equivalent along-track area AD that contributes to the received signal power for a delay-Doppler altimeter on a single-pulse basis. The cross-track dimension is set by the pulse-limited condition. Thus, √ AD = 2hβ cταR (14) The post-processing power of the delay-Doppler ﬂatsurface response function is PD =

PT G2 (θ)λ2 CR σ 0 √ 2β cταR (4π)3 h5/2

(15)

which has an h−5/2 height dependence, and a square-root dependence on compressed pulse length. The height dependence in Eq. (15) is the geometric mean between (h−3 ) for the pulse-limited case described by Eq. (5) and (h−2 ) for the beam limited case. Reduced sensitivity to compressed pulse length τ in comparison to the pulse-limited case may be helpful in system optimization. From Eqs. (4) and (15), the relative power efﬁciency of the two altimeters is given by the ratio PD AD = αR PP AP in which it is assumed that all other factors (such as average transmitted power and antenna gain) are equal in the two cases. The areas AD and AP are given by Eqs. (3) and (14) respectively. To ﬁrst order, the relative radiometric advantage of the delay-Doppler altimeter over the pulselimited altimeter is given simply by the ratio of the equivalent areas over which the signals are integrated. Example: The delay-Doppler technique would require only about 1/10 of the transmitter power of TOPEX to support the same SNR performance, yet it yields waveforms with reduced speckle due to increased incoherent summation. Cross-track interferometry Pulse-limited radar altimeters work best over relatively mild topographic relief of mean slope zero, such as the ocean’s surface. Over ice sheets or terrestrial surfaces, performance is degraded. Unwanted characteristics include footprint dilation over rougher terrain, height errors in proportion to surface mean slope, and the tendency of the footprint location to hop from one elevated region to another (without the control or knowledge of the data analyst). Beam-limited techniques, of which laser altimeters are extreme examples, circumvent these problems, but may imply their own set of disadvantages. A major potential application of radar altimetry is to monitor the height of extensive ice sheets, as found in Greenland or Antarctica. Approximately 95% of these surfaces have slopes less than ∼3 degrees, which is sufﬁcient to trick a conventional altimeter into very large height errors. For example, an unknown one-degree slope would lead to a 120-m surface height error, which is unacceptable. Although the delay-Doppler technique helps to overcome errors induced by surface slope components in the alongtrack direction, that is not sufﬁcient. Error due to an unknown cross-track slope component can be mitigated if its slope is known. Radar interferometry, as an adjunct to a delay-Doppler altimeter, can be used

15

to measure such cross-track surface slopes. The phasemonopulse technique uses this principle to estimate the angle of arrival of reﬂections from a tilted surface collected through two antennas separated in the cross-track direction of the altimeter (Fig. 13). In a radar altimeter that uses phase-monopulse [28], a scatterer at cross-track distance y away from nadir precipitates a path length difference h, observable through the cross-channel differential phase. The cross-track phase-monopulse technique can measure the presence of small (mean) cross-track surface slopes. Once measured, the slope data can be applied to recover accurate estimates of the height h of (gently) sloping surfaces. The cross-track phase-monopulse technique complements the delay-Doppler technique, which is an alongtrack enhancement. D2P Airborne Testbed The ﬁrst embodiment of the delay-Doppler altimeter combined with a phase-monopulse cross-track receiver is the D2P radar developed at the Johns Hopkins Uniersity Applied Physics Laboratory [29]. The D2P is a coherent airborne radar altimeter that operates from 13.72 to 14.08 GHz (Ku-band). The system transmits a linear FM chirp signal at 5 Watts peak power, with pulse lengths ranging from 0.384 to 3.072 microseconds. The system uses two receiver channels and a pair of antenna arrays, separated by a 14.5 cm baseline, to provide for angle measurements in the cross track direction. The system provides real time display of the delay-Doppler spectrum and cross-track phase of a burst sequence (typically 16 consecutive pulses). The D2P system typically is installed into a P-3 research aircraft. Recent campaigns include ﬂights to Greenland, Svalbard, Antarctica, and over sea ice. FUTURE DIRECTIONS CryoSat CryoSat [30] is the ﬁrst satellite of the European Space Agency’s Living Planet Programme to be realized in the framework of the Earth Explorer Opportunity Missions. The mission concept was selected in 1999 with launch originally anticipated in 2004. Unfortunately, the launch (October 2005) failed. A rebuild of CryoSat-2 was approved by ESA, now scheduled for launch in 2009. The Cryosat orbit will have high-inclination (92◦ ) and a long-repeat period (369 days, with a 30-day sub-cycle), designed to provide dense interlocking coverage over the polar regions. Its aim is to study possible climate variability and trends by determining the variations in thickness of the Earth’s continental ice sheets and marine sea ice cover. The Cryosat altimeter will be the ﬁrst of its kind: SAR/Interferometric Radar ALtimeter (SIRAL), whose advanced modes are patterned after the D2P altimeter [31], and whose ﬂight hardware has extensive Poseidon heritage. Unlike previous radar altimeter missions, Cryosat will downlink all altimetric data. These data will support three modes: conventional, interferometric, and synthetic aperture. The conventional (pulse-limited) mode will be used for open ocean (for calibration and sea surface height

16

Altimetry, Radar

reference purposes) and the central continental ice sheets that are relatively level. The interferometric mode will be used for the more steeply sloping margins of the ice sheets. The synthetic aperture mode will be used primarily over sea ice, where its sharper spatial resolution and better precision will support measurement of the freeboard for ﬂoating sea ice. These measurements can be inverted to estimate ice thickness.

be very appealing, since it would attract more potential users (and sponsors), and would require relatively low-cost space-based assets (single frequency, and no WVR, as long as the resolution and precision requirements imposed by geodesy were satisﬁed). Adoption of this paradigm would require users to think outside of traditional boundaries.

AltiKa WSOA The wide-swath ocean altimeter (WSOA) [32] has been promoted by the Jet Propulsion Laboratory as a means to overcome the dominant time/space coverage dilemma that confronts ocean altimetry. The standard altimeter measurement geometry is strictly nadir-viewing: only one subsatellite height proﬁle is gathered during each pass of the spacecraft. Whereas nadir heights can be very accurate, the surface heights of all regions between nadir tracks remain unobserved, and hence unknown. Many applications would prefer a substantially wider swath of simultaneous height measurements. Several altimeters have been proposed over the years that would scan the surface below with a set of altimetric beams arrayed orthogonally to the sub-satellite path. The goal is reasonable–to generate a wide swath of height measurements, rather than the single sub-satellite line of data points typically available. However, there are problems with this general approach. The dominant difﬁculty is that the measurement is based on triangulation, rather than the much more robust (minimum) range measurement of nadir altimetry. Off-nadir triangulation is extremely sensitive to the satellite’s roll angle error δθ. Height accuracy within a beam-limited paradigm, at an off-nadir measurement angle θ, depends to ﬁrst order on h(tan θ sec2 θ)δθ, which increases rapidly from zero as the off-nadir angle is increased. In contrast, a pulse-limited nadir altimeter’s height measurement accuracy is not degraded in response to small attitude errors at the spacecraft. The height accuracy requirements typical of oceanographic applications of a few cm cannot be met by a single-pass multi-beam or wide swath system given the state-of-the-art of controlling or determining spacecraft (roll) attitude control. The WSOA concept promises to overcome this roadblock by combining swaths from ascending and descending passes. The accurate nadir heights from one pass will be applied to remove systematic cross-track height errors in the intersecting swath. Dual-use altimetry To date, the two themes of dynamic mesoscale ocean topography and geodesy have remained disjoint. Geodesy requires a non-repeating orbit, whereas traditional oceanographic altimetry, including mesoscale observations, relies on exact-repeat orbits. Recent investigations suggest that the two objectives could be satisﬁed by one altimeter in a non-repeating orbit, if adequate near-simultaneous ancillary data were available from a more conventional mission such as Jason. The feasibility of dual-use altimetry is a work in progress [35]. If veriﬁed, such a mission could

AltiKa [36] differs from other ocean-viewing altimeters in this article, due primarily to its use of Ka-band (35.75 GHz) rather than Ku-band. The ﬁrst instrument is being built by France for India’s Oceansat-3. AltiKa is singlefrequency, since at Ka-band the retardation due to the ionosphere is sufﬁciently small that it does not have to be measured and compensated. However, the ∼0.84 cm wavelength is vulnerable to atmospheric moisture; it is predicted that as much as 10% of the data will be compromised by rain. The 33 kg instrument requires an input power of 80 W. The offset-fed reﬂector antenna is 1 m in diameter, which will have a beamwidth less than half that of its Ku-band counterparts. Several advantages are claimed for the smaller beamwidth, including operation closer to land. On the other hand, the narrower beam implies that the waveform will be more sensitive to spacecraft attitude errors. AltiKa’s 500 MHz bandwidth leads to a pulse-limited footprint about 30% smaller than usual. The PRF will be 4 KHz, approximately twice that of most conventional altimeters. The higher PRF implies smaller instrument noise, as long as the returns from adjacent transmissions remain mutually incoherent.

BIBLIOGRAPHY 1. L.-L. Fu and A. Cazanave, “Satellite Altimetry and the Earth Sciences,” Academic Press, 2001, 463 pages. 2. URL/AVISO, “http://www.aviso.oceanobs.com/,” July 2003. 3. V. L. Pisacane,“Satellite techniques for determining the geopotential of sea surface elevations,” Journal of Geophysical Research, vol.91, pp. 2365–2371, 1986. 4. M. E. Parke, R. H. Stewart, D. L. Farless, and D. E. Cartwright, “On the choice of orbits for an altimetric satellite to study ocean circulation and tides,” Journal of Geophysical Research, vol.92, pp. 11693–11707, 1987. 5. R. K. Raney, “On orbit selection for ocean altimetry,” IEEE Transactions Geoscience and Remote Sensing, (to appear), 2003. 6. URL/Geodesy, “http://www.ngdc.noaa.gov/mgg/bathymetry/ predicted/explore.HTML,” (accessed July 2003). 7. D. T. Sandwell and W. H. F. Smith, “Marine gravity anomaly from Geosat and ERS-1 satellite altimetry,” J. Geophys. Res., vol.102, pp. 10039–10054, 1997. 8. “Special Sections: Geosat Science and Altimeter Technology,” in Johns Hopkins APL Technical Digest, Vol10, No. 4, 1989. 9. URL/Jason, “http://www-aviso.cls.fr/html/missions/jason/ welcome uk.html,” (accessed July 2003). 10. URL/RA-2, “http://envisat.esa.int/instruments/tourindex/ra2/,” (accessed July 2003). 11. P. C. Marth, J. R. Jensen, C. C. Kilgus, J. et al.,” Prelaunch performance of the NASA altimeter for the TOPEX/Poseidon

Altimetry, Radar

12.

13.

14.

15.

16.

17.

18.

19.

20. 21.

22.

23.

24.

25.

26. 27.

28.

29.

30.

Project,” IEEE Transactions on Geoscience and Remote Sensing, vol. 31, pp. 315–332, 1993. A. R. Zieger, D. W. Hancock, G. S. Hayne, and C. L. Purdy, “NASA radar altimeter for the TOPEX/Poseidon project,” Proceedings of the IEEE, vol.79, pp. 810–826, 1991. S. J. Keihm, M. A. Janssen, and C. S. Ruf, “TOPEX/Poseidon microwave radiometer (TMR) III: Wet troposhperic range correction and pre-launch error budget,” IEEE Transactions on Geoscience and Remote Sensing, vol.33, pp. 147–161, 1995. S. Musman, A. Drew, and B. Douglas, “Ionospheric effects on Geosat altimeter observations,” J. of Geophysical Research, vol.95, pp. 2965–2967, 1990. D. B. Chelton, J. C. Ries, B. J. Haines, et al., “Satellite Altimetry,” in Satellite Altimetry and Earth Sciences, International Geophysics Series, L.-L. Fu andA. Cazanave, Eds. San Diego: Academic Press, 2001, pp. 1–122. R. K. Moore and C. S. Williams, Jr., “Radar return at near-vertical incidence,” Proceedings of the IRE, vol. 45, pp. 228–238, 1957. G. S. Brown, “The average impulse response of a rough surface and its applications,” IEEE Antennas and Propagation, vol. 25, pp. 67–74, 1977. W. J. J. Caputi, “Stretch: a time-transformation technique,” IEEE Transactions on Aerospace and Electronic Systems, vol. AES-7, pp. 269–278, 1971. J. L. MacArthur, C. C. Kilgus, C. A. Twigg, and P. V. K. Brown, “Evolution of the satellite radar altimeter,” Johns Hopkins APL Technical Digest, vol. 10, pp. 405–413, 1989. E. J. Walsh, “Pulse-to-pulse correlation in satellite radar altimetry,” Radio Science, vol. 17, pp. 786–800, 1982. D. B. Chelton, E. J. Walsh, and J. L. MacArthur, “Pulse compression and sea-level tracking in satellite altimetry,” Journal of Atmospheric and Oceanic Technology, vol. 6, pp. 407–438, 1989. D. T. Sandwell and W. H. F. Smith, “Bathymetric Estimation,” in Satellite Altimetry and Earth Sciences, L.-L. Fu andA. Cazenave, Eds. New York: Academic Press, 2001, pp. 441– 457. M. M. Yale, D. T. Sandwell, and W. H. F. Smith, “Comparison of along-track resolution of stacked Geosat, ERS-1 and TOPEX satellite altimeters,” J. Geophys. Res., vol. 100, pp. 15117–15127, 1995. W. H. F. Smith and D. T. Sandwell, “Global seaﬂoor topography from satellite altimetry and ship depth soundings,” Science, vol. 277, pp. 1956–1961, 1997. R. K. Raney, “The delay Doppler radar altimeter,” IEEE Transactions on Geoscience and Remote Sensing, vol. 36, pp. 1578–1588, 1998. R. K. Raney, “Delay compensated Doppler radar altimeter.” United States Patent 5, 736, 957, 1998. R. K. Raney, “Radar fundamentals: technical perspective,” in Principles and Applications of Imaging Radar, Manual of Remote Sensing, F. Henderson andA. Lewis, Eds., 3 ed. New York: Wiley Interscience, 1998, pp. 9–130. J. R. Jensen and R. K. Raney, “Delay Doppler radar altimeter: Better measurement precision,” in Proceedings IEEE Geoscience and Remote Sensing Symposium IGARSS’98. Seattle, WA, 1998, pp. 2011–2013. J. R. Jensen, “Angle measurement with a phase monopulse radar altimeter,” IEEE Transactions on Antennas and Propagation, vol. 47, pp. 715–724, 1999. URL/D2P, “http://fermi.jhuapl.edu/d2p,” Johns Hopkins University Applied Physics Laboratory, (accessed July 2003).

17

31. URL/CryoSat,“http://www.esa.int/export/esaLP/cryosat.html,” European Space Agency, (accessed July 2003). 32. R. K. Raney and J. R. Jensen, “An Airborne CryoSat Prototype: The D2P Radar Altimeter,” in Proceedings of the International Geoscience and Remote Sensing Symposium IGARSS02. Toronto: IEEE, 2002. 33. URL/WSOA, “http://ibib.grdl.noaa.gov/SAT/pubs/Jason2 paper.doc,” (accessed July 2003). 34. R. K. Raney and D. L. Porter, “WITTEX: An innovative threesatellite radar altimeter concept,” IEEE Transactions on Geoscience and Remote Sensing, vol. 39, pp. 2387–2391, 2001. 35. URL/JHUAPL, “http://fermi.jhuapl.edu/,” (accessed July 2003). 36. W. H. F. Smith and R. Scharoo, “ftp://falcon.grdl.noaa.gov/pub/walter/combi anim.gif,” NOAA, (accessed July 2003).

R. KEITH RANEY Johns Hopkins University, Laurel, MD

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering c 1999 John Wiley & Sons, Inc. Copyright

RADAR APPLICATIONS Radar (radio detection and ranging) systems attempt to infer information about remotely located objects from reflections of deliberately generated electromagnetic waves at radio frequencies. Typically, the information sought is detection of the presence of target objects in the midst of clutter, recognition (classification) of targets, and estimation of target parameters such as range (distance from the radar antenna), bearing (azimuth and elevation), orientation, velocity, acceleration, or backscattering cross section (reflectivity) distribution. Early radar systems could only scan the environment, to detect an aircraft when it appeared in their beam and to measure its range and bearing. The range resolution cell was determined by the length of the unmodulated transmitted pulse and was much larger than the aircraft. Thus, it was reasonable to model the aircraft as a point target and the system interference as white Gaussian thermal noise. Subsequently, the detection problem was reduced to that of detecting a point target in white Gaussian noise. Modern radar systems, however, are expected to perform the much more sophisticated tasks just stated for multiple targets simultaneously, at the finest possible target resolution and with the highest possible accuracy. Additionally, the domain of utilization of radar techniques has expanded beyond the traditional aircraft detection and ranges to applications such as estimation of the parameter (range, velocity, acceleration, and backscattering cross section) distribution of spread targets, aerial imaging, and ground or foliage penetrating radar imaging. To achieve their expanded tasks, modern radar systems combine high-quality hardware with sophisticated signal design and processing algorithm development and implementation based on statistical descriptions of both the target characteristics and the clutter distributions. Applications of modern radar can be found in the military, the civilian, and the scientific regime. Military applications include search and surveillance of enemy targets; navigation, control, and guidance of weapons; battlefield surveillance; and antiaircraft fire control. Among civilian applications, prominent are those in air, water, and land transportation, including aircraft navigation; collision avoidance with both other aircraft and terrain obstacles; detection and avoidance of weather disturbances and clean-air turbulence; altimetry; air traffic control; shore-based ship navigation; collision avoidance for ships and small boats; harbor and waterway traffic control; collision avoidance of land vehicles; tracking of vehicles; and traffic law enforcement; as well as space applications in detection and tracking of satellites and control of rendezvous and docking of space vehicles. Finally, scientific applications include remote sensing of the Earth’s environment from aircraft and satellites for planetary observation; weather radar for study and monitoring of precipitation, clouds, and major weather disturbances; ground mapping; ground-penetrating radar for detection of buried objects; foliage-penetrating radar for detection of hidden targets; and high-resolution imaging of objects and terrain via synthetic aperture imaging radars.

1

2

RADAR APPLICATIONS

Radar Target Measurement Elements Point-Target Measurements. Assume that the radar antenna transmits the narrow-band pulse

where f 0 is the carrier frequency at which the radar operates and u(t) is a pulse of duration T and bandwidth B smaller than the carrier frequency. The pulse illuminates a point target and gets reflected back towards the antenna. Let be the round-trip delay between the time at which the rising edge of the pulse leaves the radar antenna, gets reflected by the target, and is received back at the antenna. Since the target is moving, this delay will be a function (t) of time. Ignoring amplitude attenuation and constant phase shifts due to reflection, the received pulse will be

The information about the target motion is contained in the round-trip delay (t) as a function of time and the distortion it causes on the received pulse sR (t). The delay (t) depends on the target position at the instant of reflection, which for a signal received at time t, occurs at time t − (t)/2. Thus:

where R(t) is the target range, that is, the distance from the radar antenna to the target, as a function of time. If the target moves slowly enough for the delay (t) to be approximately constant within the duration T of the illuminating pulse, then the target can be regarded as stationary. However, it is often the case that the target moves very fast in comparison with the pulse duration and different instants of the pulse are differently delayed. In the general case, the relation among the target motion, the round-trip delay, and the received pulse is too complicated to be tractable. Several simplifications can lead, however, to tractable mathematical relations. Assume that the delay (t) is a smooth enough time function to be expandable into a Taylor series around the time instant t0 = (t = 0) ≡ 0 at which the leading pulse edge is received back at the receiver:

where (k) 0 = dk /dtk (t0 ) is the kth derivative of the round-trip delay evaluated at the time instant t0 . Define now the target parameters of interest:

RADAR APPLICATIONS

3

and use Eqs. (3) and (4) to relate them to the target range and its derivatives. Algebraic manipulation gives the target parameters as the following functions of the range R0 = R(t0 /2) and its derivatives R(k) 0 = (dk /dtk )R(t0 /2) at time t0 /2:

The approximations in Eqs. (6)–(8) are valid simplifications for the practical cases of R(1) 0 c of the exact expressions on p. 59 in Ref. 1. From Eq. (8), it is seen that even the simplified expression for the target hyperacceleration 0 is still a complicated function of target range derivatives. Higher-order terms in Eq. (4) have coefficients that are nonmanageable functions of target range derivatives. Fortunately, practical radar systems need only deal with targets moving sufficiently smoothly for only the delay and Doppler coefficients and, occasionally, the acceleration coefficient (and, rather rarely, the hyperacceleration coefficient) to be significant in the expansion in Eq. (4). Additionally, only the delay term t0 is significant in the complex envelope u[t − D(t)], while the higher-order terms affect only the phase in the exponential in Eq. (2). That is, the pulse received by the radar from a single target illuminated with the pulse of Eq. (1) is

where t0 , v0 , γ 0 , and 0 are the target delay, Doppler (velocity), acceleration, and hyperacceleration parameters. Matched-Filter Response to Received Pulse. The received pulse is processed through a bank of filters, each matched to a different set of values of the target parameters. The filter matched to the set of parameter values (τ, v, γ, ) has impulse response

Assume now that the input to this matched filter is the received pulse in Eq. (9), multiplied by an unknown complex-valued amplitude Aeiδ and corrupted by additive noise, that is, Aeiδ sR (t) + n(t).The output of the filter

4

RADAR APPLICATIONS

will be

where T 0 is the time interval during which the radar is in receive mode (e.g., the time interval between two successive pulse transmissions). Equation (11) consists of two terms: a term due to noise and a signal term containing the ambiguity function

Clearly, the magnitude of the signal term at any time t is maximized if the filter parameters are selected equal to the target parameters, that is, if τ = t0 , v = v0 , γ = γ 0 , and = 0 . Target detection can be performed by monitoring thematched-filter outputs at each instant t and examining whether they exceed a preset threshold or not. If the threshold is exceeded, then target detection is declared. If a target is thus detected, its parameters are subsequently estimated as the parameters of the matched filter that produces maximum output. A simplification to this target detection or estimation rule can be achieved by noticing that the matched-filter delay τ is not significant in that it only corresponds to a shift in the time instant of occurrence of the maximum of the matchedfilter output. Indeed, a change in the round-trip delay t0 only changes the time at which the maximum occurs. Thus, only a bank of matched filters needs to be used, in which the delay τ is fixed to zero and the target range is estimated from the time instant of occurrence of the maximum of the matched-filter output. If, however, the other target parameters are significant, an entire bank of filters needs to be used, with each filter matched to different target parameter values. In summary, the criterion for declaring target detection is

and the set of values (ˆt, vˆ , gˆ , ˆ ) that provides the maximum constitutes the target parameter estimates. The threshold is set so as to keep the probability of a false alarm below a specified maximum tolerance. Since the complex amplitude Aeiδ is unknown and varying, constant false alarm (CFAR) techniques need to be utilized to set the threshold adaptively. Distributed-Target Measurements. The theory of single-target measurements needs to be modified and extended if the radar is to operate at a resolution that is sufficiently high for the spread of one or more target parameters to exceed the corresponding resolution bin. Examples of such targets include the terrain, vegetation foliage, extended manmade objects such as buildings with more than one smooth surface, or even aircraft when the resolution bin is significantly smaller than its typical dimensions. In these cases, the pulse received at the radar antenna can be considered as the superposition of a small or large or even infinite number of reflections from individual scattering centers on the target. Serious difficulties in extending the theory of single- to multiple- (distributed-) target measurements arise if the scattering centers of the target are not stationary during illumination with the radar pulse or new ones emerge or several disappear due to target motion. Additionally, the target may be dispersive, that

RADAR APPLICATIONS

5

is, its significant scattering centers may vary with frequency, making the target behavior rather complex. For the theory of distributed-target measurements to remain tractable, the assumption needs to be made that the target is represented by a possibly infinite, yet fixed, set of scattering centers. Additionally, no dispersion can be allowed, that is, the scattering centers need to be frequency independent. With these assumptions in mind, consider a target consisting of N scattering centers illuminated with the pulse of Eq. (1). The reflected pulse measured at the radar antenna will be

producing the signal part at the output of the matched filter of Eq. (10):

Under the assumptions of stationarity of the target scattering centers during illumination with the radar pulse and their independence, cross terms in the signal part of the magnitude squared of the matched-filter output will be relatively small. Thus

Considering the limit of N → ∞ densely packed scattering centers, the magnitude squared of the matched-filter output becomes

In Eq. (17), |A(t0 , v0 , γ 0 , 0 )|2 is the target backscattering cross-section distribution as a function of the delay, Doppler (velocity), acceleration, and hyperacceleration parameters. Clearly, if the magnitude squared of the ambiguity function consists of a single central spike with very narrow width, that is, if

then

In other words, the matched-filter response represents (and measures) the target backscattering cross section for the particular values of delay, Doppler, acceleration, and hyperacceleration coefficients to which the filter is matched. Consequently, a bank of matched filters, each adjusted to a different delay, Doppler, acceleration, and hyperacceleration parameters, yields the entire target cross-section distribution. A simplification can be

6

RADAR APPLICATIONS

obtained by considering only matched filters corresponding to the delay τ = 0 in the bank and utilizing the entire matched-filter output for parameter distribution estimation.

Search (Surveillance or Acquisition), Tracking, and Navigation Radar Search Radar. A search (also known as surveillance or acquisition) radar uses an efficient scan pattern to cover an angular sector with a narrow pencil beam in order to detect the presence of a suspected target. Typical scan patterns include the helical, the Palmer, the spiral, the raster (or TV), and the nodding patterns. In the helical pattern, the beam is continuously rotated in azimuth while it is simultaneously raised or lowered in elevation. The Palmer pattern consists of a rapid circular scan about the antenna axis, combined with a linear movement of the axis of rotation, and is suited to a search area which is larger in one dimension than the other. The spiral scan covers an angular search volume with circular symmetry. Both the Palmer and the spiral scans need to vary the scanning speed during the scan cycle for all parts of the scan volume to receive the same energy. The raster scan is produced by oscillating the antenna beam fast in azimuth and slowly in elevation, while the nodding scan is produced by oscillating the antenna fast in elevation and slowly in azimuth. Both the raster and the nodding scans cover a rectangular area, but they can also be used to obtain hemispherical coverage. Hemispherical coverage can also be obtained by the helical pattern. Tracking Radar. A tracking radar measures the coordinates of a target found by a search radar and provides data that can be used to determine the target path and predict its future position. All or part of the available data (range, elevation and azimuth angle, Doppler frequency shift, acceleration, and hyperacceleration) may be used in predicting future target position. Correspondingly, the radar may track in range, in angle, in Doppler velocity, in acceleration, in hyperacceleration, or in any combination of those. Tracking radars either supply continuous tracking data on a particular target (continuous tracking radar) or supply sample data on one or more targets (track-while-scan) radar. The target parameters in a continuous tracking radar are tracked by a servocontrol loop activated by an error signal generated at the radar receiver. The information available from a tracking radar can either be displayed on a cathode-ray-tube display for action by a human operator, or may be supplied to a digital computer and automatically processed to determine the target path and predict its probable future course. The latter is usually called automatic detection and track mode or integrated automatic detection and track mode when the outputs from more than one radars are automatically combined. Sequential Lobing Radar. The difference between the target angular position and a reference direction, usually the antenna axis, is the angular error. The tracking radar attempts to position its antenna to make the tracking error zero and thus locate the target along the reference direction. One method to obtain the direction and the magnitude of the angular error in one coordinate is by alternately switching the antenna beam between two positions. This is called lobe switching, sequential switching, or sequential lobing. The difference in amplitude between the voltages in the two switched positions is a measure of the angular displacement of the target from the switching axis. The sign of the difference determines the direction that the antenna must be moved in order to align the switching axis with the direction of the target. Two additional positions are needed to obtain the angular error in the orthogonal coordinate. Thus, a two-dimensional sequentially lobing radar might consist of a cluster of four feed horns illuminating a single antenna, arranged so that the right-left-up-down sectors are covered by successive antenna positions. Both transmission and reception are accomplished at each position. Conical Scan Radar. Conical scan tracking radar uses continuous rotation of an offset antenna beam rather than discontinuous stepping between four discrete positions. The angle between the rotation and the antenna axes is called the squint angle. The echo signal is modulated at the frequency of the beam rotation. The phase of the modulation depends on the angle between the target and the rotation axis and can be used to locate the target and continuously position the rotation axis on it.

RADAR APPLICATIONS

7

Monopulse Tracking Radar. The sequential-lobing and conical-scan tracking radars require a train of echo pulses in order to extract the angular error signal. This echo train must contain no amplitude-modulation components other than the modulation produced by the scanning; otherwise the tracking accuracy will be degraded. On the other hand, pulse-to-pulse amplitude modulations have no effect on tracking accuracy if the angular measurement is based on a single pulse rather than on several. If more than one antenna beam is used simultaneously, it is possible to extract angular error information from a single pulse from the relative phase or the relative amplitude of the echo signal received in each beam. Tracking radars that derive angular error information from a single pulse are known as simultaneous lobing or monopulse radars. An example of a simultaneous lobing technique is the amplitude-comparison monopulse, in which the echos received from two offset antenna beams are combined so that both the sum and the difference signals are obtained simultaneously. The sum signal provides range information, while the difference signal provides angular error information in one angular direction. Track-While-Scan Radar. A search radar can obtain the track of a target by marking the coordinates of the target from scan to scan. Such a radar is called track-while-scan radar and either requires a human monitor to mark the target path manually or uses a digital computer to perform automatic detection and tracking. The automatic detection is achieved by quantization of the range into intervals equal to the range resolution. At each range bin, the detector integrates the number of pulses expected to be returned from a target as the antenna scans past and compares them with a threshold to indicate the presence or absence of a target. When a new detection is received, an attempt is first made to associate it with an existing track. When the detection is declared independent of existing tracks, the radar attempts to make a smooth estimate of the target’s present position and velocity, as well as a predicted position and velocity. One method to achieve this is by using either the so-called α-β tracker or a Kalman filter that utilizes a dynamic model for the trajectory of a maneuvering target and the disturbance or uncertainty of the trajectory. Navigation Radar. Navigation radar is used to provide the necessary data for piloting an aircraft from one position to another without any need for navigation information transmitted to the aircraft from a ground station. A self-contained aircraft navigation system utilizes a continuous-wave Doppler radar to measure the drift angle and true speed of the aircraft relative to the Earth. The drift angle is the angle between the centerline (heading) of the aircraft and the horizontal direction (ground track). A navigation radar requires at least three non-coplanar beams to measure the vector velocity, that is, the speed and its direction, of the aircraft. Such a radar measures the vector velocity relative to the frame of reference of the antenna assembly. This vector velocity can be converted to a horizontal reference on the ground by determining the direction of the vertical and the aircraft heading by some auxiliary means. Usually, the radar uses four beams initially symmetrically disposed about the aircraft axis, with two facing forward and two facing rearward. If the aircraft vector velocity is not in the direction of the aircraft heading, the two forward-facing beams will not read the same Doppler frequency. This Doppler difference can be fed in a servomechanism that will align the axes of the antennas with the ground track of the aircraft. The angular displacement of the antennas from the aircraft heading is the drift angle, and the magnitude of the Doppler frequency is a measure of the speed along the ground track. The use of the two rearward beams is similar, but improves the accuracy considerably by reducing the errors caused by vertical motion of the aircraft and pitching movements of the antennas. High-Resolution Imaging Radar A radar image is a visual representation of the spatial microwave reflectivity distribution of a target illuminated by the electromagnetic radiation emitted by the radar. Equivalently, a radar image represents a collection of reflection coefficients assigned to an array partitioning the target space. Thus, a radar image is generated by the same physical mechanism that generates an optical image observed by a human observer, in which the optical reflectivity distribution is reconstructed. In humans, however, the aperture size of the imaging system

8

RADAR APPLICATIONS

is on the order of 10,000 wavelengths, orders of magnitude (in wavelengths) greater than the aperture size of the corresponding radar imaging systems. Since the resolution of an imaging system, that is, its ability to represent distinctly two closely spaced elements, is inversely related to its aperture size, radar imaging systems would appear primitive when compared to their optical counterparts. Whereas a single optical image is usually sufficient for target recognition, several radar images of the same target, corresponding to various viewing angles, are usually required. However, the usefulness of radar imaging systems is not undermined by their lower-resolution capabilities. Advantages of radar imaging systems over their optical counterparts include their day or night capability, since they supply their own illumination and their all-weather capability, since radio waves propagate through clouds and rain with only limited attenuation. Additionally, larger aperture sizes (and, thus, higher resolution) can be synthesized from the given physical aperture using techniques such as those described later in this article. Direct Imaging Radar. Direct imaging radar systematically scans a three-dimensional volume in angle and range with short pulses emitted from a pencil-beam antenna and range gating and displays the intensity of the received signals as a function of the spatial coordinates interrogated. The spatial resolution is established by the angular (beam width) and range (pulse duration) resolution of the sensor without subsequent processing. If range gating is not used, then range is not resolved and the radar image is a two-dimensional projection of the reflectivity distribution along the radar line-of-sight. Direct imaging is the simplest form of radar imaging, requiring minimal data processing and allowing the target to be stationary. However, it requires very large aperture and subnanosecond pulses for a high degree of spatial resolution, while, due to beam widening, its cross-range resolution degrades as the range increases. Synthetic Imaging Radar. Synthetic imaging radar attempts to overcome the limitations of direct imaging radar and create fine spatial resolution by synthetic means in which results from many observations of the target at different frequencies and illumination angles are coherently combined. The term synthetic here refers to the synthesis of resolution commensurate with short-pulse, large-aperture illumination from a number of elemental measurements of illumination with not-as-short pulses and not-as-large aperture. Range Processing Radar. The first task of imaging radar involves discrimination on the basis of range. High resolution in the determination of range is achieved when the transmitted pulse duration T is narrowed down and the corresponding system bandwidth B is increased, so that the time–bandwidth product (TB) is constant. Maximum sensitivity is accomplished when the time–bandwidth product is set to unity, that is, TB = 1. Thus, the required range resolution can be achieved when target reflections are measured over a band of frequencies. Any radar waveform that supports an extended bandwidth can be used, the specific type of waveform only determining the necessary implementation of the receiver for coherently processing the wide-band signal. In contrast to direct imaging methods, in which all the spectral components of the signal must be present simultaneously, synthetic imaging methods require that the spectral components be present sequentially. In the simplest implementation of high range resolution by synthetic means, several narrow-band measurements are made at discrete frequency increments. Such radars are called stepped-frequency systems and can be either continuous wave, at each frequency emitting an unmodulated sinusoid, or pulsed, amplitude-modulating each frequency sinusoid. Stepped-frequency continuous-wave systems are susceptible to aliased responses and transmitter coupling, shortcomings alleviated by pulsing the transmitter and time-gating the receiver as in a pulsed, stepped-frequency system. Although individual narrow-band responses have insignificant resolution potential, the coherent combination of the responses provides the resolution allowed by the total bandwidth spanned. Alternatively, high range resolution can be accomplished using swept-frequency (linear FM) systems and corresponding wide-band receivers. Range resolution in swept-frequency systems is achieved by measuring the difference in instantaneous frequency between the instant of emission of the radar pulse by the transmitter and the instant of its reception back at the receiver. Synthetic Aperture Processing Radar. High resolution in the cross-range direction can be obtained by scanning a focused beam across the object. If the aperture that forms the scanning beam is focused at the

RADAR APPLICATIONS

9

target plane, the minimum lateral extent of the focused spot is approximately

where λ is the wavelength, R is the observation distance, and D is the aperture dimension. Resolution of two adjacent object points on a plane perpendicular to the line-of-sight of the radar is possible if their distance is greater than the spot dimension. Thus, for a fixed wavelength and observation distance, the resolution is increased by increasing the aperture size. High-resolution direct imaging radars would, therefore, need to have physically large aperture. Synthetic imaging radars synthesize equivalent large aperture for high-resolution cross-range imaging by sequentially stepping a sensor through small incremental distances and storing samples of the amplitudes of the corresponding received signals. The stored signals are coherently summed to produce signals equivalent to those that would be received by the corresponding large physical aperture. In effect, synthetic aperture radars (SARs) coherently process signals scattered from the same target for various viewing angles by utilizing relative motion between the sensor and the target. Depending on the type of relative motion between sensor and target, synthetic aperture radars can be linear, spotlight, or inverse. Linear Synthetic Aperture Radar. In linear SAR, also called stripmap SAR, the radar sensor is moved along a linear path and images stationary targets in its line-of-sight. Linear SAR is widely used for mapping terrain features and ground-based objects from airborne platforms. Spotlight Synthetic Aperture Radar. Spotlight SAR involves observing a target with the radar antenna fixed on it while the viewing angle is changed. Inverse Synthetic Aperture Radar. Inverse SAR involves a stationary radar viewing targets rotating about an axis perpendicular to the line-of-sight. Doppler Processing Radar. Spatial resolution in cross range, that is, along an axis perpendicular to the radar line-of-sight, can be obtained if a target rotates relative to the radar sensor and the target reflections are Doppler processed. This is possible since the Doppler frequency shift in waves reflected by a rotating target is proportional to the lateral offset of the reflector along an axis normal to the axis of rotation and the line-of-sight. Indeed, if d and R0 (R0 d) are the distances of a reflecting point and the radar sensor, respectively, from the center of a target rotating at an angular velocity , then the distance of the reflecting point from the radar sensor at time t is approximately

According to Eq. (6), the Doppler coefficient at time t in the received wave will be

From Eq. (22), it is clear that the Doppler coefficient for every reflecting point in a target rotating with angular velocity is a harmonic function of time, the amplitude of which is proportional to the instantaneous lateral distance of the reflecting point from the center of rotation. Doppler processing of the received signal for crossrange resolution can be done on-line by either a bank of contiguous filters or by first sampling it and then analyzing it with Fourier transform processors of sufficiently high speed. Off-line processing, on the other hand, can be performed by recording the received signal for later processing. In either case, the signal is usually frequency translated to retain only its complex envelope.

10

RADAR APPLICATIONS

Holographic Processing Radar. Optical holography records the spatial distribution of the intensity of the interference of light waves diffracted by an object and a reference beam in a hologram. This overcomes the difficulty associated with lack of optical phase sensitive storage media. Later, the hologram can be used to reconstruct the light waves associated with the original object by illumination with the reference beam used in the recording step. A holographic reconstruction allows a viewer the perception of a virtual image of the original object. Microwave holography follows recording and reconstruction procedures analogous to optical holography. In microwave holography, the field amplitude scattered from an object coherently illuminated from a transmitter is mapped over a prescribed recording aperture by a coherent detector that is scanned over the aperture. The detected bipolar signal, representing the complex envelope of the time-varying field, is added to a bias level sufficient to make the resultant always positive. The resulting signal is used to produce a film transparency with an amplitude transmittance function that is real and positive. The area probed by the detector represents the hologram aperture, the reference signal for the coherent detector represents the reference beam, and the signal scattered from the object is the object beam. A variation, known as scanned holography, of the (conventional) procedure described previously attempts to scan the transmitter and the receiver independently and offers some advantages in resolution.

Weather Observation Radar Radar is a powerful research instrument in meteorology and also for telecommunications at frequencies higher than 1 GHz, since it allows for the gathering of considerable quantities of data on the three-dimensional structure of the atmosphere in a flexible, efficient, and rapid manner. Radar applied to hydrometeor observation provides two types of information: (1) quantitative information on the local distribution of the reflectivity and speed distribution in the scattering medium and (2) qualitative information on the small- and medium-scale structure of the targets, their evolution and movement, and other information heuristically extracted by expert analysts. This information can be used to describe atmospheric phenomena and study radio wave propagation, such as the provision of various statistics on precipitation and attenuation. Precipitation Measurements. The most frequent quantitative application of the radar observation is to distinguish ice and liquid phases of precipitation. This task is particularly challenging in convective storms, where liquid water can exist at temperatures colder than 0 ◦ C and ice can be found at temperatures warmer than 0 ◦ C. Equally important and challenging is the task of quantifying rain-, snow-, and hail-fall rates, where the difficulty lies in the dependence of the rates on detailed knowledge of the drop size distributions. Although radar techniques have practical limitations and their accuracy is highly suspect, they offer important advantages over conventional methods based on pluviometer array measurements: they allow for spatial continuity of the observations and improved access to the observation of the variability of the precipitation; they make it possible to observe the three-dimensional structure of the system generating the precipitation, as well as survey over a wide area from a single measurement point in real time; data acquisition, storage, and processing are simple. Radars capable of measuring multiple parameters (e.g., vertical and horizontal reflectivities and/or a spectrum of terminal velocities) in each resolution cell, in combination with satellites, rain gauges, and other instruments, may give the desired accuracy in measuring rainfall rates and discriminating rain from frozen precipitation. Storm and Wind Observation. A pulsed Doppler radar can estimate the reflectivity and range velocity distribution inside a storm’s shield of clouds. If a single beam is used, a three-dimensional picture of a storm typically requires 2 min to 5 min of data collection time. This delay is imposed not only by antenna rotation limitations, but also by the requirement for collection of a large number of radar echos for reduction of the statistical uncertainty in the reflectivity and velocity estimates. Although the storm can change significantly

RADAR APPLICATIONS

11

during this period, with subsequent distortion of the reflectivity and velocity radar images, the returned estimates are considered highly valuable. Practically, more significant than the reflectivity and velocity distributions are estimates of the rainfall rate and wind velocity. Doppler radar, however, measures the range velocity of hydrometeors rather than air, and often this differs significantly from the range component of wind. Nevertheless, since hydrometeors quickly respond to wind forces, their terminal velocities give negligible bias estimates of the range component of the wind. Turbulence Measurement. The mean velocity and spectrum width measured by Doppler radar are weighted averages of point velocities. Therefore, they are sufficient to depict motion on scales larger than the resolution cell, but cannot infer the details of the flow inside the cell. Nevertheless, Doppler radar offers the possibility of measurement and study of turbulence on scales smaller than the resolution cell if a firm connection between the statistical-physical properties of the atmosphere and Doppler-derived measurements is established. Clean Air Observation. A radar designed to identify and track precipitating storms can also detect echos from scatterers in fair weather. In such cases, the distribution of spatial reflectivity in clean air can be associated with meteorological phenomena such as turbulent layers, waves, and fronts, flying birds and insects, or atmospheric pollutants. Clean air echos not related to any visible scatterers have been conclusively proven to emanate from refractive-index irregularities. Waves reflected by sharp, quasipermanent changes in the dielectric permittivity of the atmosphere form the coherent component in the echo received by the radar. Coherent echos exist if the scattering medium does not time-modulate the amplitude or phase of the transmitted radar pulses, even though spatial variations may exist. Coherent echos appear as peaked and narrow components in the Doppler spectrum. On the other hand, incoherent components are contained in the echo signal if time-varying (turbulent) scatter is present. Incoherent echos demonstrate themselves as broad components in the Doppler spectrum.

Laser Radar Systems It is natural to attempt to extend radar techniques to the optical portion of the electromagnetic spectrum. The fact that optical wavelengths are orders of magnitude smaller than their radio counterparts allows for very fine resolution in the estimation of target parameters, such as angular position, range, and tangential and radial velocity. The first optical radar systems investigated used incoherent light from xenon or other flash lamps. With the invention of the laser (light amplification by stimulated emission of radiation), however, they were replaced by systems that employed coherent laser light. Such systems were initially called lidar (light detection and ranging) and, in more advanced, higher-performance versions, ladar (laser detection and ranging) in complete analogy to radar. The development of practical ladar systems followed a path similar to the development of radar systems: Reflectivity and backscattering cross-section data were collected and tabulated for a number of targets at available laser operating wavelengths; laser pointing could be accomplished by mounting the laser on existing radar platforms; the scan patterns of radar were employed in ladar systems as well; the signal design and processing techniques of radar were also applicable to ladar systems. When compared to radar, ladar systems exhibit advantages and disadvantages, both due to their use of very short wavelengths. On one hand, very short wavelengths result in very high information bandwidths and very fine resolution. Exploitation of the bandwidth can be achieved with today’s advanced signal-processing techniques and hardware. Additionally, unlike the use of solid-state lasers in early ladars, which allowed only for signal-envelope processing, today’s ladars use gas lasers, which also allow for signal-phase processing. On the other hand, however, very short wavelengths result in low power efficiency and high atmospheric propagation losses. As a result, ladar is preferable to microwave and millimeter-wave radar for long-range ground-to-space

12

RADAR APPLICATIONS

and space-to-space applications or short-range atmospheric applications, in which the propagation loss penalty does not outweigh fine resolution. Ladar Information Processing. A ladar measures a target’s range, position, velocity, and motion by modulating its laser beam, detecting the reflected return, and processing the return signal to derive the desired information. Methods have been developed for amplitude, frequency, and phase modulation and for modulation by polarization. Laser radiation can be modulated both by direct action on coherent signals inside the laser during their generation (internal modulation) and through action on the radiated light outside the laser (external modulation). A number of electro-optical, acousto-optical, and mechanical beam modulation devices are available with different inherent modulation rates, yielding amplitude or frequency modulation of the transmitted beam. Solid-state lasers cannot provide the necessary spectral purity to utilize phase processing of ladar signals. Gas lasers, such as helium-neon and carbon dioxide, however, have high spectral purity and can be modulated in amplitude or frequency with bandwidths of up to 525 MHz (yielding a resolution of approximately 30 cm) with relatively low drive powers. Ladar signal-processing techniques are similar to those used in microwave radar. In fact, the same circuits for signal-envelope processing may be employed in many cases. The use of ladar allows the exploitation of highly precise and unique methods for angle estimation and tracking.

Ground- and Foliage-Penetrating Radar In the recent years, an attempt has been made to use radar to detect and map “targets” buried under the Earth’s surface or obscured by foliage. Primarily, interest arises from a number of potential applications, such as detecting and locating unexploded ordnance in a battlefield, manmade objects in landfills, buried hazardous waste, subsurface plumes of refined hydrocarbons, or military equipment hidden in forest vegetation. The radar is either spaceborne, airborne, or ground-towed and possibly operates in SAR mode. The radar needs to utilize ultra-wide-band (UWB) signals, containing both low- (for deeper penetration) and high- (for higher resolution) frequency components. This can be achieved in two ways: (1) by emission of pulses that are very short in duration (and, subsequently, ultra-wide in bandwidth) or (2) by sequential emission of narrow-band signals of carrier frequency increases in steps, covering a wide frequency band. Even though the perspective of ground- or foliage-penetrating radar from initial tests has been encouraging, a number of difficulties have delayed the development of this technology. These include: • • • •

High electromagnetic wave absorption, especially under moist soil conditions Random distributions of soil particles, such as rocks, that tend to scatter the electromagnetic energy, increase propagation losses, and reduce the image contrast High clay content to which water binds, and thus dipolar relaxation loss mechanisms are encouraged Roughness of the air-soil interface that tends to increase backscattering that interferes with the penetrating radar signature

Similar factors affect the development of foliage penetrating radar systems. To date, the success of ground- and foliage-penetrating radar surveys seems to be absolutely site dependent. A thorough understanding of a site’s geology, hydrology, and topography is of paramount importance. Before undertaking a radar survey, it is necessary to obtain as much information as possible about the physical characteristics of the specific site. If boring log or monitor well data are available, they should be analyzed to determine soil stratigraphy and hydrology. If such data are not available, it is prudent to gather representative soil samples. The applications for radar in subsurface target detection seem to fall into two broad categories, depending on the scale of the system, target, terrain structures, and search volumes. The case of large scales is made if the

RADAR APPLICATIONS

13

targets sought are large relatively to the average wavelength and the soil inhomogeneities. In this case, imaging would play a (secondary) role in reducing the number of false alarms of the detection procedure. If small targets, such as mines or weapons, are of interest, they would be hard to distinguish from clutter and the role of imaging would be enhanced. Thus, it is difficult or perhaps pointless to develop a single radar system for the detection of both large or deep and small or shallow targets. The wide-frequency-range requirement imposes stringent requirements in the range of both the electronics and the size of the relevant antennae and contributes to the delay of development of this significant radar application. However, ground- or foliage-penetrating radar technologies are presently an area of significant research investigation.

Current Trends Besides research in ground- and foliage-penetrating radar technologies, significant research is also conducted in the development of, so-called, space-time adaptive processing (STAP) algorithms. STAP refers to multidimensional adaptive filtering algorithms that simultaneously combine the signals from the elements of an array antenna and the multiple pulses of a coherent radar waveform. STAP can improve the detection of low-velocity targets obscured by mainlobe clutter, detection of targets masked by sidelobe clutter, and detection in combined clutter and jamming environments. Significant research is also conducted into the use of signal processing tools other than the traditional Fourier transform–based ones for target detection and recognition. Such tools are, for example, based on the theories of, so-called, wavelet-induced multiresolution analyses (WIMA) of signals. A WIMA allows for the decomposition and simultaneous representation of a signal in time and scale and, therefore, is capable of processing signals at different scales. WIMA-based radar target detection and recognition is being actively researched.

BIBLIOGRAPHY 1. A. W. Rihaczek, Principles of High Resolution Radar, Norwood, MA: Artech House, 1996.

READING LIST C. G. Bachman, Laser Radar Systems and Techniques, Dedham, MA: Artech House, 1979. L. J. Battan, Radar Observation of the Atmosphere, Chicago: University of Chicago Press, 1973. P. Bello, Joint estimation of delay, Doppler, and Doppler rate, IRE Trans. Inf. Theory, IT-6 330–341, June 1960. W. G. Carrara, R. S. Goodman, R. M. Majewski, Spotlight Synthetic Aperture Radar: Signal Processing Algorithms, Norwood, MA: Artech House, 1995. I. Cindrich et al. (eds.), Aerial Surveillance Sensing Including Obscured and Underground Object Detection, Bellingham, WA: SPIE—The Society for Optical Engineering, 1994, Vol. 2217. N. K. Del Grande, I. Cindrich, P. B. Johnson (eds.), Underground and Obscured Object Imaging and Detection, Bellingham, WA: SPIE—The Society for Optical Engineering, 1993, Vol. 1942. A. J. Devaney et al., Automatic target detection and recognition: a wavelet approach, Final Report on ARPA Grant F4962093-1-0490, 1995. R. J. Doviak, D. S. Zrni´c, Doppler Radar and Weather Observations, San Diego, CA: Academic Press, 1992. A. K. Fung, Microwave Scattering and Emission Models and Their Applications, Norwood, MA: Artech House, 1994. E. J. Kelly, The radar measurement of range, velocity, and acceleration, IRE Trans. Mil. Electron., MIL-5 51–57, 1961. E. J. Kelly, R. P. Wishner, Matched-filter theory for high-velocity, accelerating targets, IEEE Trans. Mil. Electron., 56–69, January 1965. R. Meneghini, T. Kozu, Spaceborne Weather Radar, Norwood, MA: Artech House, 1990. D. L. Mensa, High Resolution Radar Cross-Section Imaging, Norwood, MA: Artech House, 1991.

14

RADAR APPLICATIONS

D. R. Rhodes, Introduction to Monopulse, New York: McGraw-Hill, 1959. A. W. Rihaczek, S. J. Heshkowitz, Radar Resolution and Complex-Image Analysis, Norwood, MA: Artech House, 1996. H. R. Raemer, Radar Systems Principles, Boca Raton, FL: CRC Press, 1997. H. Sauvageot, Radar Meteorology, Norwood, MA: Artech House, 1991. S. M. Sherman, Monopulse Principles and Techniques, Norwood, MA: Artech House, 1984. M. I. Skolnik, Introduction to Radar Systems, New York: McGraw-Hill, 1980. F. T. Ulaby, M. G. Dobson, Handbook of Radar Scattering Statistics for Terrain, Norwood, MA: Artech House, 1989. J. Ward, Space-time adaptive processing for airborne radar, MIT–Lincoln Laboratory Technical Report 1015, 1994.

GEORGE A. TSIHRINTZIS Northeastern University

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering c 1999 John Wiley & Sons, Inc. Copyright

RADAR SIGNAL DETECTION Each day, we constantly make decisions. Given certain hypotheses, information is selected on which to base each decision, and under certain conditions, we may need to determine the reliability of the information. Such is the case in radar signal detection: A returned signal is received, and we have to decide whether a target is present or absent. If there were no noise or interference, the decision could be made with complete confidence. However, in reality, the received signal is usually heavily corrupted by environmental noise, interference, and noise from the radar system itself, and so on. In order to make a reliable decision, the noise and unwanted signals have to be suppressed with a so-called matched filter before a decision can be made. Owing to the existence of noise and interference, radar signal detection has to be treated as a statistical problem, regardless of whether the signal under detection is deterministic or not. This statistical formalism of radar signal detection theory can be applied to all types of radar signals without restriction. To understand radar signal detection, noise has to be quantitatively described. A time-limited deterministic signal can be described as a time series, and a periodic deterministic signal can be represented as a Fourier series. In contrast, noise cannot be represented as a deterministic function in the time or frequency domains. In other words, we cannot predict a noise-contaminated radar signal with absolute certainty. However, with available noise information such as the expectation, the power, or even the probability distribution of noise, we can select a criterion on which to base our decision. Concepts of Signal-to-Noise Ratio and Matched Filter The signal-to-noise ratio (SNR) in radar and communication systems is defined as

The maximum output SNR, the most frequently used criterion for radar detection, is defined as the ratio of the maximum instantaneous output signal power to the output noise power. The input SNR is a major limiting factor for radar detection performance. For a fixed input SNR, a linear time-invariant filter whose frequency response function maximizes the output SNR is called a matched filter. Matched filtering transforms the raw radar data into a form that is suitable for (1) generating the optimal decision for detection; (2) estimating the target parameters with a minimal rms error, or (3) obtaining the maximum resolving power for a group of targets. The characteristics of matched filters can be described by either a frequency-domain transfer function or a time-domain impulse response function, each being related to the other by the Fourier transform. In the frequency domain, the matched-filter transfer function H(ω) is the complex conjugate of the spectrum of the signal. Thus, in general terms

1

2

RADAR SIGNAL DETECTION

where S(ω) is the spectrum of the input signal s(t) and T is a delay constant required to make the filter physically realizable. The normalizing factor k and the delay constant are generally ignored in formulating the underlying significant relationship. This simplification yields

Equation (3) reveals that the bandwidth of the receiver must be the same as that of the signal. This is understandable, because if the bandwidth of the receiver is wide compared with that occupied by the signal energy, extraneous noise may be introduced into the excess bandwidth, which lowers the output signal-to-noise ratio. On the other hand, if the receiver bandwidth is narrower than the signal bandwidth, the noise energy is reduced along with part of the signal energy. The result is again a lowered SNR. When the receiver bandwidth is identical to the signal bandwidth as in the case of the matched filter, the output SNR is maximized. The conjugate in Eqs. (2) and (3) allows the phases of S(ω) and H(ω) to cancel each other out, and leaves the output signal spectrum a linear phase, e − jωT , which results in a peak at the time instant T in the output. The corresponding time-domain relationship between the signal to be detected and the matched filter is obtained from the inverse Fourier transform of H(ω). This leads to the result that the impulse response of a matched filter is a replica of the time inverse of the known signal function. Thus, if h(t) represents the matched-filter impulse response, the relationship equivalent to Eq. (2) is given by

As before, k and T can be ignored to yield the basic relationship

Figure 1 illustrates the relationship given by Eqs. (3) and (5), where s(t) is a pulsed linear frequencymodulated (LFM) signal with the form

The phase from H(ω) is the negative of that from S(ω), while h(t) is the time reversal of the s(t). Figure 2(a) shows a received signal, which is the signal s(t) of Eq. (6) corrupted by a 6 dB Gaussian noise; that is, the input SNR is −6 dB. It is difficult to detect the existence of the signal s(t) from this figure. However, after the received signal is processed by the matched filter, the detector output peak in Fig. 2(b) clearly indicates the existence of the signal. The output from the matched filter, as shown in Fig. 3, is the convolution between the received signal and the matched-filter impulse response, that is,

RADAR SIGNAL DETECTION

3

Fig. 1. (a) Signal s(t) and (b) matched-filter h(t) relations. Phase in units of degrees. The phase from H(ω) is the negative of that from S(ω), while h(t) is the time reversal of s(t).

Fig. 2. (a) Signal corrupted by noise and (b) the matched-filter output. The peak in the matched-filter output indicates the existence of the signal.

Sampling y(t) at t = T yields the maximum output signal value, that is,

4

RADAR SIGNAL DETECTION

Fig. 3. Block diagram of a matched filter.

Fig. 4. Block diagram of a cross-correlation, which is another implementation of the matched filter.

where Es represents the signal energy. It can be easily verified that the expectation of y(t)max is Es , because the second term in Eq. (8) represents the noise whose mean is zero. This can be easily seen from Fig. 2(b) in which the maximum signal energy occurs at t = T = 2, and the maximum value is close to the expectation of Es = 0.18 in this experiment. A detailed analysis of the matched filter will be given in the section entitled “Analysis of a Matched Filter.” Equation (7) describes the output of the matched filter as the cross-correlation between the received signal and a replica of the transmitted signal. This implies that the matched filter can be replaced by a cross-correlation that performs the same mathematical operation, as shown in Fig. 4. The received signal is multiplied by a delayed replica of the transmitted signal s(t − t1 ), and the product is passed through a low-pass filter. The cross-correlation tests for the presence of a target at only one time: t1 . Targets at other time delays, or ranges, may be found by varying t1 . However, this requires a longer search time. The search time can be reduced by adding parallel channels, each containing a delay line corresponding to a particular value of t1 , as well as a multiplier and a low-pass filter. Since the cross-correlation and the matched filter are equivalent mathematically, the choice of which one to use in a particular radar application is determined by the practicality of implementation. The matched filter, or an approximation, has been generally preferred in the vast majority of applications. Decision Criteria for Radar Signal Detection The statistical detection problem consists of examining the received radar waveform r(t) in a resolution cell to determine which of the following two hypotheses is true. The first hypothesis H 1 asserts that a target is present, and the received signal contains the target signature and noise. The second hypothesis H 0 states that the target is absent, and only noise is present in the received signal. The problem can be compactly stated as

RADAR SIGNAL DETECTION

5

The conditional probability density function completely describes the received signal statistically in both cases:

For reasons of simplicity, r is assumed to be a single sampled point of the received radar signal. The extension from a single sampled point to multiple sampled points is straightforward. The likelihood ratio is defined as

The likelihood ratio (r) is also called the likelihood statistic. It is a random variable since it is a function of the random variable r. The maximum likelihood (ML) decision criterion, which chooses the hypothesis that most likely causes the observed signal, is

This expression means that H 1 is selected if (r) is greater than 1; otherwise H 0 is selected. It can be seen that the ML criterion is a very simple decision criterion. To describe the detection performance better, the probabilities of detection and false alarm are used in radar detection. The probability of detection refers to the probability of asserting the presence of a target when the target is indeed present

where R0 is the decision boundary. The proper value of the boundary R0 depends upon the criterion of decision. The probability of false alarm is the probability of asserting the presence of a target when the target is actually absent:

A sketch of the two density functions is shown in Fig. 5, where Pd and Pfa are, respectively, shown by the vertically and the horizontally hatched areas. If the observed value r is large, we would be confident in picking H 1 . If r is small, we would pick H 0 , as shown in Fig. 5. Obviously, a decision rule should be selected to maximize Pd while restricting the Pfa . The simplest rule in this class, which is extensively used in radar detection, is the Neyman-Pearson criterion. This criterion specifies a decision boundary that maximizes the probability of detection (Pd ) while maintaining a fixed probability of

6

RADAR SIGNAL DETECTION

Fig. 5. Probability of false alarm Pfa and probability of detection Pd , which are functions of the threshold R0 .

false alarm Pfa . The detection problem under the Neyman–Pearson criterion can be formulated as follows:

The optimum decision region can be found by using the calculus of extrema and forming the objective function

where η is a Lagrange multiplier. This can be written as

The integration interval in Eq. (17) is related to choosing the hypothesis H 1 , as illustrated in Fig. 5. It is clear that J and hence Pd are maximized by choosing the hypothesis H 1 when

and by choosing the hypothesis H 0 when

To this end, the decision rule based on the Neyman–Pearson criterion is

RADAR SIGNAL DETECTION

7

and η is determined by the required false alarm probability α. In radar detection, the choice of α is based upon operational considerations, that is, the need to keep the false alarm rate within acceptable bounds (e.g., a few false alarms per second). A typical value of α for radar detection is 10 − 6 . Other popular criteria are the Bayes criterion and the minimum error probability (MEP) criterion. The Bayes criterion minimizes the average cost of the decision. Symbols denoted by C00 , C01 , C10 , and C11 represent the costs for a correct miss (no target is declared when no target is present), a false dismissal (no target is declared when a target is present), a false alarm, and a correct detection, respectively. Also denoted are the a priori probabilities P(H 0 ) and P(H 1 ) by P0 and P1 , respectively. The Bayes rule makes the likelihood ratio test

where η= [P0 (C10 − C00 )]/[P1 (C01 − C11 )]. If we select the cost of an error to be 1 and the cost of a correct decision to be 0, C01 = C10 = 1, and C00 = C11 = 0. In this case, minimizing the average cost is equivalent to minimizing the probability of error. Therefore, the MEP rule is the same as expression Eq. (21), but with η= P0 /P1 . If the a priori probabilities are equal, that is, P0 = P1 , the MEP rule coincides with the ML rule with η=1. Implementation of Decision Criteria Let us suppose that the observed signal r has the following Gaussian distribution conditional probability density functions,

where µ denotes the mean of the received signal value and σ2 represents the noise variance. The likelihood ratio test is therefore

After taking the logarithm of both sides, the criterion becomes

In Eq. (25) it is seen that the likelihood ratio test, in which (r) of Eq. (24) is compared with a threshold η, is transformed into a comparison of the observable r with the threshold in Eq. (25), which is a function of η. As an example, supposing P(H 0 ) and P(H 1 ) are known, with P(H 0 )/P(H 1 ) = 2, then the decision rules are choose H 1 if

8

RADAR SIGNAL DETECTION

Fig. 6. Decision thresholds for different decision criteria.

Since the a priori probability of H 0 is twice that of H 1 , the MEP rule requires a larger value of R0 for the selection of H 0 than the ML, in which this information is not used. The MEP scheme therefore yields a better decision rule in this case. For a Neyman–Pearson criterion, suppose a value of Pfa = 10 − 4 can be tolerated. The threshold η is determined from

to be η= 3.72σ. So H 1 is chosen if

A typical illustration of these thresholds for this example’s three decision criteria is given in Fig. 6. An important observation is that these criteria employ the likelihood ratio test. In other words, the test is performed by simply processing the received data to yield the likelihood ratio and then comparing it with the threshold, which depends upon the criterion used. Thus, in practical situations where the a priori probabilities and the cost may vary, only the threshold changes, and the computation of the likelihood ratio is not affected. As observed previously, in radar detection it is very hard to define the Bayes cost Cij ; moreover, it is also practically impossible to define or evaluate the a priori probabilities P0 and P1 , that is, the probabilities that, in a given resolution interval, a target is present or absent. These are the main reasons why the Bayes and minimum error probability criteria cannot be used in radar detection. In contrast, for the same reason, the Neyman–Pearson criterion is particularly well suited to radar detection, owing to its concept of the “Pfa threshold” fixed a priori, while Pd is maximized.

RADAR SIGNAL DETECTION

9

Fig. 7. ROC curves for different values of µ.

Receiver Operating Characteristic A graph showing the probability of detection, Pd , versus the probability of false alarm, Pfa , with the threshold as a parameter is referred to as a receiver operating characteristic (ROC) curve. We note that the ROC depends on the conditional density function of the observed signal under each hypothesis, that is, p(r|H i ), i = 0, 1, and not on the assigned costs, or a priori probabilities. Suppose that the observed signal r has the probability density functions of Eqs. (22) and (23). Varying the threshold R0 , Pd of Eq. (13), and Pfa of Eq. (14) produces the corresponding ROC curves for σ = 1 and µ = 1, 2, 3 as shown in Fig. 7. The two extreme points on the ROC for Pfa = Pd = 1 and Pfa = Pd = 0 are easily verified. Pd of Eq. (13) may be rewritten as a function of the likelihood ratio (r) as

where η is the threshold of the likelihood ratio just as R0 is the threshold of the observed signal, and p (λ|H 1 ) in Eq. (30) is the conditional probability density function of the variable . Similarly, Pfa of Eq. (14) is rewritten as

Since is a ratio of two non-negative quantities, it takes on values from 0 to ∞. When the threshold η is 0, the hypothesis H 1 is always true and thus Pfa = Pd = 1. When the threshold η is ∞, the hypothesis H 0 is always true and thus Pfa = Pd = 0. These are clearly depicted in Fig. 7. Of course, ROC curves may be drawn for any hypothesis test involving a threshold, but the ROC curves have particularly useful properties for the likelihood ratio test. One is the fact that the slope of the ROC at a particular point on the curve represents the threshold value of the likelihood ratio η. Taking the derivative of Eqs. (30) and (31) with respect to η, we have

10

RADAR SIGNAL DETECTION

and

Also,

Taking the derivative of Eq. (34) with respect to η, we obtain

Combining Eqs. (32), (33), and (35), the slope of the ROC curve obtained is

In the Neyman–Pearson criterion, the slope of the ROC curve at a particular point represents the likelihood ratio threshold η of achieving Pd and Pfa at that point. In the Bayes criterion, the threshold η is determined by a priori probabilities and the costs. Consequently, Pd and Pfa are determined on the point of the ROC at which the tangent has a slope of η. Since the ROC curves are always concave and facing downward, it is possible to determine an “optimum” value (the knee) for Pfa , such that a small decrease of its value causes a fast decrease of Pd , while any increase has a very small effect (the saturation zone, where the rate of change is nearly 0). Finally, we note that the most important part of the ROC curve is the upper left-hand (northwest) corner. This is the so-called high-performance corner, where a high-detection probability occurs with a low false-alarm probability. This part of the plot could be stretched out by the use of appropriate (such as logarithmic) scales.

Analysis of a Matched Filter Derivation of the Matched Filter. The matched filter achieves the maximum output SNR, which is

Consider a signal s(t) with the spectrum S(f ) and finite energy

RADAR SIGNAL DETECTION

11

For the input signal s(t) and the filter with transfer function H(f ), the instantaneous power of the output signal y(t) is

For white noise with a two-sided noise power spectral density N 0 /2, the output power spectral density is |H(f )|2 N 0 /2. Therefore, the noise power at the filter output is

Using Eqs. (39) and (40) in Eq. (37) in leads to the following:

where T denotes the time at which the maximum value of |y(t)|2 occurs. Using Schwarz’s inequality

we obtain

It follows that the signal-to-noise ratio will be a maximum when

yielding the requirement of the matched filter. From discussions above it is evident that the maximum signalto-noise ratio can be expressed as

Equation (45) indicates that the detection capability of a particular signal depends only on its energy content, and not on the time structure of the signal. However, it is necessary to process the signal through a matched filter to obtain this condition in practice. We note that Es /N 0 is defined as the input SNR, and it is clear from Eq. (45) that the maximum output SNR for the matched filter is twice that of the input SNR if the noise is white.

12

RADAR SIGNAL DETECTION

In general case, when the noise is nonwhite (colored noise), the derivation of the matched filter can be carried out in a similar way. If the power spectral density of the nonwhite noise is N(f ), then Eq. (40) is written as

Therefore, by multiplying and dividing the integrand of the numerator of Eq. (41) by (46):

and using Eq.

and the maximum is achieved when

The conjugate is not needed in the denominator because N(f ) is always real (and nonnegative). If the noise is white—that is, if N(f ) is a constant over the band of H(f )—then Eq. (48) is the same as Eq. (44) for the white noise. The matched filter for nonwhite noise can be interpreted as the cascade of two filters. The first is the “whitening” filter. This filter makes the noise spectrum flat one, whose transfer function is 1/ (white). The second one is matched to the signal filtered by the whitening filter, that is, to the whitening signal . with the spectrum S(f )/ We note that it is not necessary that the noise be Gaussian for Eq. (45) to hold, but only that its power spectral density be flat over the frequency band of interest. To summarize, the matched filter maximizes the output SNR over all probability densities, provided the power spectral density (PSD) is a constant. In the event that the noise PSD is nonwhite (colored noise), the matched impulse response corresponds to the modified signal spectrum S∗(f )e − j2πfT /

rather than simply S∗(f )e − j2πfT . Justiﬁcation of the Signal-to-Noise Ratio Criterion. We derived the matched filter under the criterion of maximizing the output SNR. We remark here that the matched filter can also be derived under the likelihood ratio criterion (1). In this section, we want to justify the maximum output SNR criterion, and more specifically derive the relationship between the output SNR and the system performance in terms of the probability of error.

RADAR SIGNAL DETECTION

13

The total probability of error for a radar receiver consists of the false-alarm probability Pfa and the false-dismissal probability Pfd . A false dismissal declares no target when a target is present, that is,

For equal a priori probabilities P(H 0 ) = P(H 1 ) = 1/2, the total probability of error is

Supposing p(r|H i ), i = 0, 1 is a Gaussian distribution that is given by Eqs. (22) and (23), Pe can be expressed as

where

is the error function. The minimum of Pe occurs when R0 = µ/2, and

where

is the complementary error function. Recalling that µ is the expectation of the matched filter output at time T under the H 1 hypothesis

14

RADAR SIGNAL DETECTION

Fig. 8. Minimum error probability Pe |min versus maximum output signal-to-noise ratio SNR|max .

and

Defining σ2 to be the variance of y(T) under the H 0 or H 1 hypotheses, it is given that

The application of Eqs. (55) and (57) to Eq. (53) leads to the following minimum probability of error:

It is clear from Eq. (58) that Pe |min is inversely proportional to SNR|max , because erfc(x) is a monotonic decreasing function. In other words, a lower probability of error means a higher output SNR, and requires a higher input SNR. Figure 8 shows the relationship between Pe |min and SNR|max . This curve should be shifted to the left by 3 dB if it is plotted with respect to the input SNR, since SNR|max is twice the input SNR. For example, a 10 − 5 error probability corresponds to an output SNR of 18.6 dB, and 15.6 dB of input SNR is required. Therefore, it is justifiable to use the signal-to-noise ratio criterion in radar detection.

RADAR SIGNAL DETECTION

15

Noncoherent Detections A received radar signal is a bandpass random process because it is modulated on a carrier. Radar detection is classified into coherent and noncoherent detections depending upon whether the carrier phase at the receiver is available. Specifically, the matched filter and the cross-correlation discussed previously are coherent because they require the knowledge of the carrier phase. The envelope and all the other nonlinear detections are noncoherent due to their ignorance of the phase information in the received signal. To understand the nonlinear detections, we introduce the representations of bandpass signals and bandpass processes. Representation of Band-Pass Signals. The concept of a band-pass signal is a generalization of the concept of monochromatic signals. A bandpass signal is a signal x(t) whose spectrum X(f ) is nonzero for frequencies in a usually small neighborhood of some high frequency f 0 , that is,

where the frequency f 0 is referred to as the central frequency (carrier frequency) of the bandpass signal. A radar signal that is modulated on a carrier is a bandpass signal. It is assumed that the band-pass signal is real-valued. Figure 9(a) illustrates the spectrum of a bandpass signal x(t). A real-valued bandpass signal x(t) can be represented as the real part of a complex signal x+ (t), called the preenvelope or analytic signal of x(t), where

and

is the Hilbert transform of x(t). The spectrum of the preenvelope signal is readily found from the Fourier transform of Eq. (60) to be

The spectrum of the preenvelope signal is obtained by deleting the negative frequencies from X(f ) and multiplying the positive frequencies in X(f ) by two, as illustrated in Fig. 9(b). The spectrum of the complex envelope is obtained by shifting X + (f ) to the left by f 0 , that is,

and

The amplitude spectrum of x˜ (t) is illustrated in Fig. 9(c).

16

RADAR SIGNAL DETECTION

Fig. 9. (a) Amplitude spectrum of a band-pass signal x(t). (b) Amplitude spectrum of preenvelope x+ (t). (c) Amplitude spectrum of complex envelope x˜ (t). The spectrum of x+ (t) is twice the positive spectrum of x(t), and the spectrum of x˜ (t) is a low-pass version of that x+ (t).

It is clear that x˜ (t) is a low-pass signal, meaning that its frequency components are located around the zero frequency. x˜ (t) is the low-pass representation of the bandpass signal x(t). In general, x˜ (t) is a complex signal having xc (t) and xs (t) as its real and imaginary parts:

where xc (t) and xs (t) are low-pass signals, respectively, and are called the in-phase and quadrature components of the bandpass signal x(t). Notice that x(t) is the real part of x+ (t). Using Eq. (65), we obtain

RADAR SIGNAL DETECTION

17

Fig. 10. (a) Band-pass description and (b) complex envelope description of a system. Complex envelope description simplifies the analysis of a bandpass signal.

This is the canonical representation for a bandpass signal in terms of the in-phase component xc (t) and quadrature component xs (t) of the complex envelope associated with the signal. The complex envelope can be employed to find the outputs of bandpass systems driven by bandpass signals. Accordingly, by analyzing the complex envelope representation of a band-pass signal, we may develop the complex low-pass representation of the bandpass system by retaining the positive-frequency half of the ˜ ) denote the transfer function of the complex low-pass transfer function H(f ), and shift it to the left by f 0 . Let H(f system so defined. The analysis of the bandpass system with transfer function H(f ) driven by the bandpass signal with spectrum X(f ), as depicted in Fig. 10(a), is replaced by an equivalent but simpler analysis of a ˜ ) driven by a complex low-pass input with spectrum X(f ˜ ), complex low-pass system with transfer function H(f ˜ ). as shown in Fig. 10(b). The complex low-pass output y˜ (t) is obtained from the inverse Fourier transform of Y(f Having determined y˜ (t), we may find the desired band-pass output y(t) simply by using the relation

The bandpass to low-pass transformation is also true for bandpass random processes. X(t) is a bandpass process if its power spectral density Sx (f ) = 0 for |f − f 0 | ≥ W. X(t) can be represented by its in-phase component X c (t) and quadrature component X s (t) in the same way that a bandpass signal does. Specifically,

where X c (t) and X s (t) are two low-pass processes representing the real and imaginary parts of the complex ˜ ˜ envelope process X(t), respectively. X(t) can be found from the complex envelope process X(t) by

Envelope Detection and Square-Law Detection. The matched filter is the optimal detection for an exactly known signal (i.e., phase, amplitude, and Doppler frequency are known) in a background of white noise. However, both the matched filter and the cross-correlation need to generate a synchronous reference, which

18

RADAR SIGNAL DETECTION

is difficult to realize. In a typical radar application, the range between the target and the radar represents a very large number of transmitted signal wavelengths. This makes specifying the phase of the return signal extremely difficult, and we usually assume that the signal phase of the return signal is a random variable uniformly distributed over an angle of 2π rad. The matched-filter detection is often used to set a standard of performance, as it represents the optimal detection when all signal parameters are exactly known. The synchronization problem in the matched-filter detection is obviated in a practical system by employing an envelope detection. The envelope V(t) of a bandpass signal x(t) is given by

The complex envelope x˜ (t) can be represented more compactly as

where

The procedure for obtaining the envelope is shown in Fig. 11, which is the extraction of the in-phase and quadrature components and the derivation of the envelope from them. Specifically, the multiplication of x(t) by 2 cos(2πf 0 t) in the in-phase channel yields

The mixing operation produces two images besides the expected low-pass component. The product 2x(t) cos(2πf 0 t) is passed through an ideal low-pass filter (the integrator in Fig. 11), which rejects the images and leaves xc (t). Similar operations in the quadrature channel produce xs (t). The square sum of the quadrature components yields the envelope V(t). Removing the square-root operation from Fig. 11 yields the square-law detection. A detailed performance analysis of these detections is given in the section entitled “Performance Analysis of Coherent and Noncoherent Detections.” The envelope can be extracted alternatively by passing the band-pass signal x(t) through a rectifier and a low-pass filter, as illustrated in Fig. 12. Such a description can sometimes simplify the analysis and is easier to implement physically because the various rectifiers are readily available from the diodes and the transistors. The output of the full-wave linear rectifier is proportional to the magnitude of its input, while the output of the full-wave square-law (quadratic) rectifier is proportional to the squared magnitude of its input. The half-wave rectifier, of course, gives only the positive portion of its input. Fig. 13 shows these transfer characteristics. Referring to Fig. 12, we may write

RADAR SIGNAL DETECTION

19

Fig. 11. Block diagram of the envelope detection.

where LF indicates the low-frequency portion. Also, considering the full-wave quadratic rectifier in place of the full-wave linear rectifier, we may write

The band-pass signal x(t) in Eqs. (74) and (75) has the following form:

Then in the case of the square-law rectifier, we have

Since the envelope V(t) is slowly varying compared to the carrier frequency f 0 , the first term in Eq. (77) is concentrated around zero frequency. The fact that the term is the square of the envelope means that the bandwidth will be somewhat greater than that of V(t). The second term in Eq. (77) will be concentrated around 2f 0 with a bandwidth that depends on both the envelope square V 2 (t) and the phase modulation θ(t). In most cases of interest, the bandwidth of the total modulation will be small enough compared to f 0 so that the low-pass filter following the rectifier will easily separate the low-frequency portion of Eq. (77).

20

RADAR SIGNAL DETECTION

Fig. 12. Envelope detection with a linear rectifier.

Fig. 13. Various rectifier characteristics for (a) full-wave linear rectifier, (b) half-wave linear rectifier, (c) full-wave squarelaw (quadratic) rectifier, and (d) half-wave square-law (quadrature) rectifier.

In the case of the full-wave linear rectifier, we may write

The low-pass filter will remove all of the terms in the curly bracket except for the first term. Thus, if the bandwidth of V(t) is not too large, a very good approximation of the envelope can be obtained. A similar analysis can be carried out to show that the half-wave linear and the half-wave quadratic rectifiers extract the envelope V(t). We note that the envelope detection is referred to as a linear detection, due to its transfer characteristic stipulating that the output is proportional to the input when the input is positive, as illustrated in Fig. 13(a). The operation of the envelope detection, however, is of course highly nonlinear, and as a result the output consists of a dc term proportional to the envelope, plus an infinite number of harmonics of theinput at 2f 0 , 4f 0 , etc. It is for this reason that the envelope detection must be passed through a low-pass filter, thus eliminating the unwanted harmonics. Similar comments apply to the square-law detection. Justiﬁcation of the Noncoherent Detections. The justification of the envelope and the square-law detections by the likelihood ratio criterion is given in this subsection. The radar detection process may include

RADAR SIGNAL DETECTION

21

down-conversion of the carrier frequency to a more manageable intermediate frequency (IF). This step, however, is irrelevant to the results we are going to obtain and is therefore omitted. Consider the signal to be a carrier pulse of the form

The two hypotheses are

for 0 ≤ t ≤ T, and n(t) is a Gaussian white-noise process with two-sided spectral density N 0 /2. The detection problem described by Eq. (79) consists of examining the received waveform r(t) and determining whether it consists of a signal plus noise or noise alone. The optimal detection, as previously described, forms the likelihood ratio which is compared against a threshold. The sampling bandwidth B, which is the reciprocal of the sampling interval, must be sufficiently large to pass along essentially all of the signal energy, which will be the case if B ≥ 1/T. In this case, by the sampling theorem we know that the number of samples k is given by k = 2BT. Given these conditions, the likelihood ratio can be written as

where the noise variance σ2 is (N 0 /2)2B = N 0 B. Recall from the sampling theorem (2,3) that for any two band-limited functions u(t) and v(t) we can write

22

RADAR SIGNAL DETECTION

Applying this to Eq. (81), we have

Because the signal is of finite duration, the approximation in passing from the discrete to the continuous representation improves as B is allowed to become very large. In the noncoherent case, θ in Eq. (83) is unknown. Since no auxiliary information about θ is available, it is reasonable to assume θ to be uniformly distributed over 2π rad. An average likelihood ratio is

The exponent in the integrand can be written as

where

and Eq. (84) becomes

RADAR SIGNAL DETECTION where V = therefore

23

and I0 is the modified Bessel function of order zero. The likelihood ratio test is

where the signal energy Es = A2 T/2. Thus the natural logarithm of the modified Bessel function I0 is the optimum noncoherent detection characteristic. For a large SNR, the likelihood ratio test can be approximated as

and for a small SNR,

The implementation of the likelihood ratio test of Eq. (89) has been shown in Fig. 11, where the in-phase and the quadrature channels generate the xc and xs in Eq. (86), respectively. Summation of the square of xc and xs yields the square-law detection of Eq. (90). Taking the square root in addition to the square-law detection yields the envelope detection of Eq. (89).

Performance Analysis of Coherent and Noncoherent Detections Detection Probability Analysis. From the mean and the variances given by Eqs. (55) to (57), the false-alarm probability Pfa for the coherent detection is determined by

24

RADAR SIGNAL DETECTION

The detection probability Pd is given by

For the noncoherent detection, the in-phase and the quadrature components from Eq. (86) are

For white Gaussian noise with two-sided spectral density N 0 /2, we have the following quantities:

and therefore the joint probability density functions can be written as

RADAR SIGNAL DETECTION

25

under the H 1 and H 0 hypotheses, respectively. The averaged joint probability density function of p1 (X c , X s |θ) is

For an envelope (V =

detection, let X c = V cos θ, X s = V sin θ, then dX c dX s = V dV dθ. We have

which is a Rice distribution. Under the H 0 hypothesis, we substitute A = 0 and use I0 (0) = 1 in Eq. (97) to obtain the following Rayleigh distribution:

The threshold R0 is therefore determined from the false-alarm probability Pfa using

as

26

RADAR SIGNAL DETECTION

And the detection probability Pd is given by

This can be put into a more convenient dimensionless form with the changing of variable x = which

from

where

For a square-law (Z = X 2 c + X 2 s ) detection, let X c = Eq. (96) we have

cos θ, X s =

sin θ; then dX c dX s = 12 dZ dθ. From

which is a noncentral χ2 distribution. Under H 0 , we have the following central χ2 distribution:

The threshold R0 can be determined from the false-alarm probability Pfa , which is given by

to be

The detection probability Pd is given by

RADAR SIGNAL DETECTION

27

Changing the variable x = Z/N 0 T/4 in Eq. (108) yields a more convenient dimensionless form

where α is given by Eq. (103). Changing the variable y = in Eq. (109) leads to the interesting result that Pd is identical to that given by Eq. (102) for the envelope detection. Note that the difference between the envelope and the square-law detector concerns the presence or absence of the square-root operation. Supposing for a square-law detection that an observation value is greater than R0 of Eq. (107); then the square root of this value must be greater than R0 of Eq. (100) for the envelope detection. Therefore, the square-law detection and the envelope detection bear the same detection characteristics. The Pd ’s in Eqs. (92), (102), and (109) show explicitly that the detection probabilities depend only on Pfa and the signal-to-noise ratio Es /N 0 . Note that, in addition to Pfa , the detection threshold R0 depends on the received signal energy in the coherent case of Eq. (91), but only on the noise and the integration time in the noncoherent case of Eqs. (100) and (107), which is somewhat more convenient. A comparison of coherent and noncoherent detections is presented in Fig. 14 in terms of the value of Pd that can be achieved for a given value of Pfa as a function of signal-to-noise ratio Es /N 0 . We note the following: (1) For small values of Es /N 0 (say less than 3 dB) for any given value of Pfa , noncoherent detection requires from 2 dB to 3 dB more SNR than that required by coherent detection in order to achieve the same value of detection. (2) For large values of Es /N 0 (say greater than 10 dB) the difference in SNR required by the two schemes is less than approximately 1 dB, and it is clear that with further increase Es /N 0 , the difference eventually becomes negligible. Output Signal-to-Noise Analysis. To evaluate the detection performance, we can compare output SNRs of different schemes other than the detection probabilities we used in the last section. Recall that the SNR is the ratio between the signal power and the noise variance. The signal power is the square of the mean of the detection statistic under H 1 . Thus only the mean and the variance are required to compute the SNR. In contrast, the detection probability requires full knowledge of the probability density function, which may not be available sometimes. Therefore the output SNR, which is easier to obtain, is also used to evaluate the detection performance. Of course the detection-probability-based performance evaluation, if it is available, is more accurate than the SNR-based evaluation. In this section, the output SNR will be used to compare the detection performance between the matched filter and the square-law detection. The matched filter output Y for an input signal A cos(2πf 0 t) can be written as

The band-pass noise n(t) can be expressed in the form

28

RADAR SIGNAL DETECTION

Fig. 14. Detection probability for coherent and noncoherent detection.

Hence

Equation (112) indicates that the quadrature noise ns (t) sin(2πf 0 t) has been rejected by the low-pass filter from the output. The first term in Eq. (112) is the signal component with output signal power A2 T 2 /4, while the second term represents the in-phase noise component with output noise power E[n2 c (t)]T 2 /4. Using Ref. 2

RADAR SIGNAL DETECTION

29

we obtain the output SNR

The input signal power is A2 /2 and the input noise power is E[n2 (t)] = σ2 . The input SNR is therefore

and we have the following relationship between the input and output SNRs for coherent detection:

It is clear that coherent detection gives a 3 dB improvement in SNR. The reason for this improvement is that the multiplier and low-pass filter in Eq. (110) eliminate the quadrature noise component ns (t) sin(2πf 0 t). On the other hand, in the noncoherent case both the in-phase and the quadrature noise components come in to play. To analyze the square-law detection easily, we use the equivalent scheme of Fig. 12 with the square-law rectifier characteristic of y = x2 , as shown in Fig. 13(c), replacing the linear rectifier. The output Z of a square-law detection for the following received signal:

can be written as (4)

This output can be regarded as composed of three terms. The first term, A2 T/2, is the desired output signal component with output signal power (A2 T/2)2 . The second term, ATnc (t), represents the carrier-noise component with the associated output noise power A2 T 2 σ2 . The third term, 12 [n2 c (t) + n2 s (t)], is the self-noise component. The associated noise power is

where E[n4 (t)] = 3{E[n2 (t)]}2 = 3σ4 has been used in the last step. With these results, we can write the output SNR as

30

RADAR SIGNAL DETECTION

Fig. 15. Output signal-to-noise ratios for the matched filter and the square-law detection.

If the input SNR is much larger than 1, the output SNR is approximately equal to 12 SNRin , with the square-law detection thus causing a 3 dB reduction in signal-to-noise ratio. For input signal-to-noise ratios that are much less than 1, 2SNRin is negligible compared with 1, and Eq. (120) shows that SNRout is now equal to SNR2 in . In this case, the square-law detection causes a very serious degradation of the signal-to-noise ratio. The relationship between SNRout and SNRin for the matched filter and the square-law detection is shown in Fig. 15. It is clear from both Fig. 14 and Fig. 15 that the noncoherent detection is inferior to the coherent detection for low input signal-to-noise ratios and approximates the coherent detection for high input signal-tonoise ratios in the detection probability and the output signal-to-noise ratio.

Detection of the Linear Frequency-Modulated Signal In the previous sections, we have considered the detection of the narrow-band signal such as the singlefrequency sinusoidal pulses given by Eq. (79). The product of time duration T and bandwidth B is essentially 1. There is an inherent conflict between long-range detection and high-range-resolution capability for such signals. Because the received signal energy Es attenuates rapidly as the range increases, the long range requires large transmitted signal amplitudes in order to have a sufficiently large value of Es /N 0 for reliable detection and range estimation. But all radar systems have a limitation on the peak transmitted signal power, which imposes an upper limit on the transmitted signal amplitude. Of course the required value of Es can also be obtained by maintaining the transmitted signal amplitude at some maximum value A and increasing the signal duration T. In this case the signal bandwidth B, which is approximately 1/T, is small. But since the signal bandwidth B is inversely proportional to the range resolution (5), then achieving the required Es by increasing T reduces B, thereby degrading range-resolution capability. On the other hand, if B can be increased essentially independently of T, there is no such conflict. This is why modern radar systems employ large time–bandwidth product (BT) signals. In the radar system, the earliest and most widely used large BT signal is the linear frequency-modulated (LFM) signal (1,3). In addition to providing a solution to the long-range–high-resolution problem, the LFM signal is also a form of Dopplerinvariant waveform (3).

RADAR SIGNAL DETECTION

31

We note that in military communication systems, a large time–bandwidth product signal is referred to as a spread-spectrum signal (2). It provides resistance to jamming and has a low interception probability because the signal is transmitted at low power. Among various spread-spectrum signals, the direct sequence spread spectrum (DSSS) signal, where the transmitted signal is modulated by a pseudorandom sequence, is used in code devision multiple access (CDMA) communications (6). The frequency-hopped spread spectrum (FHSS) is another widely used spread spectrum signal in modern communication systems (2,7). Wigner-Ville Distribution and Ambiguity Function of an LFM Signal. For an LFM signal the analytic form is

The instantaneous frequency in Eq. (121) is

which has an initial frequency f 0 and increases at a frequency rate m. Since an LFM signal is a nonstationary signal, the best way to describe it is through such distribution functions as the Wigner-Ville distribution (WVD) and the ambiguity function (AF). The WVD of a signal s(t) is defined as

WVD is the Fourier transform (with respect to the delay τ) of the signal’s correlation function. It relates the time and the instantaneous frequency of a signal. Substituting the LFM signal of Eq. (121) into this definition yields [15]

where f i (t) is given in Eq. (122) and the sine function is defined as

Figure 16 shows the WVD for an LFM signal with f 0 = 20, m = 12, and T = 2. It is seen from the WVD that the instantaneous frequency linearly increases with the time in accordance with Eq. (122), whereas this relationship is not observable from the spectrum of the signal, which is also shown at the top of Fig. 16. The ambiguity function (AF) is defined as

32

RADAR SIGNAL DETECTION

Fig. 16. Spectrum (top) and WVD (bottom) of an LFM signal.

where ξ and τ denote the frequency shift and the delay, respectively. AF is the Fourier transform (with respect to the time t) of the signal’s correlation function, and it relates the delay and the Doppler frequency (or frequency shift). Note that the AF and the WVD form a two-dimensional (2-D) Fourier pair, that is,

where F and F − 1 denote the Fourier and its inverse operators, respectively. Applying Parseval’s theorem u(t)v∗(t) dt = U(f )V∗(f ) df to Eq. (125), we obtain

The AF can therefore be regarded as the matched-filter output with a different delay τ and frequency shift ξ. The AF has proven to be an important tool in analyzing and constructing radar signals by relating range and velocity resolutions. By constructing signals having a particular ambiguity function, desired performance

RADAR SIGNAL DETECTION

33

Fig. 17. Ambiguity function of the LFM signal. The AF is symmetric about the origin, and its greatest value appears above the origin.

characteristics are achieved. For example, the magnitude AF of the LFM signal in Eq. (121) is

which is shown in Fig. 17 for the same LFM signal in Fig. 16. The AF is symmetric about the origin τ = ξ = 0, and the greatest value appears above the origin. The time delay τ is related to the range, and the frequency shift ξ is related to the Doppler shift. Thus AF describes the range-Doppler ambiguity of the transmitted signal. An ideal radar signal is the one whose AF is a thumbtack function because it leaves the least ambiguity in resolving the range and Doppler shift. Detection of Multiple LFM Signals. The matched-filter detection is the optimal detection if all of the signal information (phase, initial frequency, and frequency rate) is available. However, these parameters are difficult to specify because accurate values of the range, the velocity, and the acceleration of a target are not available. Noncoherent detection is thus preferred. Next, we are going to consider the noncoherent detection of multiple LFM signals in a noise background. For multiple LFM signal detection, it is often the case that the frequency rate is the only parameter of interest in practice (8). In other words, the frequency rates distinguish different LFM signals. Such a scenario occurs in the radar detection of a small, fast-moving missile launched from a relatively slow-moving aircraft. Multiple LFM signals can be detected by locating maxima in the frequency rate in many applications. AF of Multiple LFM Signals. The input signal to be analyzed is modeled by a linear sum of two (may be extended to more than two) LFM signals with frequency rates m0 and m1 as given by

34

RADAR SIGNAL DETECTION

Here ωi represent the carrier (or the initial) frequency, which is proportional to the velocity of the target, and the frequency rate mi is proportional to the acceleration. The AF defined by

is computed for the signal r(t) of Eq. (129) to yield

where

The last two terms of Eq. (131) are interference terms generated by the two LFM components in the signal r(t), due to the nonlinearity of the AF. Using the following identity

Q1 and Q2 are combined to give

Figure 18(a) shows the AF [Eq. (133)] of a signal composed of two LFM signals which may represent two targets with different velocities and accelerations. Although there is cross-term interference, we can identify

RADAR SIGNAL DETECTION

35

Fig. 18. The ambiguity functions of a bicomponent LFM signal (a) without noise, and (b) with the additive white Gaussian noise (SNR = −6 dB).

the two straight lines representing the bicomponent signal in Fig. 18(a). However, the two LFM signals are not obvious if they are corrupted by noise. Figure 18(b) is identical to Fig. 18(a) except that the two signals are corrupted by Gaussian white noise with SNR = −6 dB. Detecting Multiple LFM Signals Using Radon-Ambiguity Transform. Recall that the Radon transform (9), commonly used for the reconstruction of images in computer tomography, is defined by

for −∞ < s < ∞ and −π/2 < φ < π/2, where the δ function specifies the direction of integration. The parameter s represents the shifted location of the origin. Equation (134) actually represents the sum of the values of f (x, y) along the line that makes an angle φ with the x axis and is located at a distance s from the origin. The Radon– Wigner transform 9, 10 is a special case when f (x, y) in Eq. (134) takes the WVD of a multicomponent LFM

36

RADAR SIGNAL DETECTION

Fig. 19. (a) The WVD of a bicomponent signal. (b) the AF of the bicomponent signal.

signal. The WVD of a bicomponent signal is graphically drawn in Fig. 19(a). The Radon–Wigner transform of Fig. 19(a) should produce two maxima in the resulting α-ω plane. Figure 19(b) is the AF of the same signal in Fig. 19(a). The AF is the 2-D Fourier transform of the WVD; thus they share the same angles of α0 and α1 as shown in the WVD. However, the initial frequencies shown in Fig. 19(a) have disappeared in Fig. 19(b), since they have been mapped into the phase of the AF. This also explains why the AFs of the two chirps pass through the origin in the τ-ξ plane. Thus, by applying the Radon transform to the phase-free ambiguity function, detection of multicomponent signals can be reduced from the 2-D search problem in the Radon–Wigner transform to a 1-D search problem. The advantage of the ambiguity function over the WVD has been shown in the kernel design for the time-frequency analysis (11). This work can be extended to the detection of multicomponent signals (12). Since all directions of interest pass through the origin of the ambiguity plane, the Radon transform with parameter s set to 0 is applied to the phase-free ambiguity function of Eq. (133). We essentially compute the line integral along a straight line with its direction specified by the δ function δ(ξ − mτ) in the ambiguity plane. Therefore the detection statistic can be formed by the so-called Radon-ambiguity transform (12) as

Since the infinite integrals in Eq. (135) usually diverge, it is necessary to first remove the constant term from the integrand. Specifically, for m = mi (i = 0, 1) and assuming m0 − m1 > 0, we have from Eq. (133)

RADAR SIGNAL DETECTION

37

with

Removing the constant from Eq. (136) and substituting it into Eq. (135) yield

For m = m0 or m1 (i.e., am = 0), it is clear that η(m) is finite. By Eqs. (137) and (138), we have η(m) → ∞ as m → m0 or m → m1 . Therefore, by calculating η(m) and comparing it to a preset threshold, the multicomponent signals can be detected. Finite-Length Signal. Now we consider a bicomponent finite-length signal as given by

with the assumption that ω0 = ω1 and m0 > m1 for simplicity purposes. The modulus of the ambiguity function of r(t) for ω = mτ can be calculated by making use of the following integral,

to yield

38

RADAR SIGNAL DETECTION

Fig. 20. The η(m) of two equal-amplitude LFM signals with T = 40, m0 = 200π/T 2 , m1 = 100π/T 2 . Solid line: η(m). Dashed line: auto terms only. Dotted line: cross-terms only. The two peaks indicate the existence of the equal-amplitude ∠FM signals.

where

are the Fresnel integrals, and

while am in Eq. (141) is defined by Eq. (137). For |τ| ≤ T, the first two terms in Eq. (141) represent the auto terms of the signal, while the rest express the cross-terms. Figure 20 shows the integral of Eq. (141) over τ, that is, η(m), for two LFM signals with equal amplitudes. Also shown in Fig. 20 are the integrals of the auto terms and cross-terms of Eq. (141).

RADAR SIGNAL DETECTION

39

Output Signal-to-Noise Ratio Analysis. The output SNR of the statistics η in Eq. (135) can be analyzed by making use of the following quantities (12):

to find

It is seen from Eq. (142) that there is a 3 dB loss in SNR between the input and the output when the input SNR is high, and the output SNR degrades severely when the input SNR is low, illustrating a typical nonlinear detection characteristic.

Conclusion We have presented the techniques of radar signal detection, as well as the related performance analyses. The following conclusions can be drawn. • • • • • •

Among various detection criteria, the Neyman–Pearson criterion is particularly well suited to radar detection, owing to its concepts of a priori fixed Pfa and maximized Pd . The coherent detection, in the form of a matched filter or a cross-correlation, is the optimal detection for an exactly known signal (i.e., phase, amplitude, and Doppler frequency are known) in a background of white noise. In a typical radar application, the range between the target and the radar represents a very large number of transmitted signal wavelengths. This makes specifying the phase of the return signal extremely difficult, and a noncoherent detection has to be used. The noncoherent detection is inferior to the coherent detection for low input signal-to-noise ratios and approximates the coherent detection for high input signal-to-noise ratios. There is an inherent conflict between long-range detection and high-range-resolution capability for the unity time-bandwidth signal. Large time-bandwidth signals such as an LFM signal do not have such a conflict. Large time-bandwidth signals can be described by the ambiguity function or the Wigner–Ville distribution. The Radon-ambiguity transform can be used to detect multiple LFM signals.

BIBLIOGRAPHY 1. 2. 3. 4.

C. E. Cook, M. Bernfeld, Radar Signals: An Introduction to Theory and Application, New York: Academic Press, 1967. J. G. Proakis, M. Salehi, Communication Systems Engineering, Englewood Cliffs, NJ: Prentice-Hall, 1994. J. Minkoff, Signals, Noise, and Active Sensors, New York: Wiley, 1992. J. Brown, E. V. D. Glazier, Signal Analysis, New York: Reinhold, 1964.

40 5. 6. 7. 8. 9. 10. 11. 12.

RADAR SIGNAL DETECTION D. L. Mensa, High Resolution Radar Cross-Section Imaging, Norwood, MA: Artech House, 1991. A. J. Viterbi, Principles of Spread Spectrum Communication, Reading, MA: Addison-Wesley, 1995. J. G. Proakis, Digital Communications, New York: McGraw-Hill, 1995. P. M. Djuric, S. Kay, Parameter estimation of chirp signals, IEEE Trans. Acoust. Speech Signal Process., ASSP-38: 2118–2126, 1990. J. C. Wood, D. T. Barry, Tomographic time-frequency analysis and its application toward time varying filtering and adaptive kernel design for multicomponent linear-FM signals, IEEE Trans. Signal Process., 42: 2094–2104, 1994. J. C. Wood, D. T. Barry, Radon transformation of time-frequency distributions for analysis of multicomponent signals, IEEE Trans. Signal Process., 42: 3166–3177, 1994. B. Ristic, B. Boashash, Kernel design for time-frequency signal analysis using the radon transform, IEEE Trans. Signal Process., 41: 1996–2008, 1993. M. Wang, A. K. Chan, C. K. Chui, Linear frequency modulated signal detection using radon-ambiguity transform, IEEE Trans. Signal Process., 46: 571–586, 1998.

READING LIST M. Barkat, Signal Detection and Estimation, Norwood, MA: Artech House, 1991. B. Bouachache, Time-frequency signal analysis, in S. Haykin (ed.), Advances in Spectral Estimation and Array Processing, Englewood Cliffs, NJ: Prentice-Hall, 1991, Vol. 1, Chap. 9, pp. 418–517. J. V. DiFranco, W. L. Rubin, Radar Detection, Englewood Cliffs, NJ: Prentice-Hall, 1968. J. L. Eaves, E. K. Reedy (eds.), Principles of Modern Radar, New York: Van Nostrand-Reinhold, 1987. G. Galati (ed.), Advanced Radar Techniques and Systems, Stevenage, UK: Peregrinus, 1993. H. V. Poor, An Introduction to signal Detection and Estimation, New York: Springer, 1988. D. C. Schleher, MTI and Pulsed Doppler Radar, Norwood, MA: Artech House, 1991. H. Urkowitz, Signal Theory and Random Processes, Norwood, MA: Artech House, 1983.

MINSHENG WANG Texas Instruments Incorporated ANDREW K. CHAN Texas A & M University

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering c 1999 John Wiley & Sons, Inc. Copyright

RADAR TARGET RECOGNITION Recent advances in radar provide sufficient resolution and enough information to recognize tactical targets from radar returns (1). For target recognition applications, different types of radar, such as synthetic aperture radar (SAR), millimeter wave (MMW) real aperture radar (RAR), and interferometric synthetic aperture radar (IFSAR), have been explored. Modern radar systems also provide extensive data, including fully polarimetric and Doppler channels, but there still are many challenges for target recognition using radar returns. The special characteristics of radar and recent advances in radar technology make the recognition of targets difficult. For example, the radar profile changes drastically for a small change in look angle, and the recent development of stealth technology significantly alters the radar signature. This article covers many of the different approaches for different radar types of radar target recognition. Synthetic aperture radar provides high resolutions for both range and azimuth directions and has been used widely for target recognition applications. Since SAR is an imaging radar, it can provide detailed images of target areas with cloud-penetrating characteristics. When its fully polarimetric data are utilized, targets can be recognized with high accuracy. The SAR target recognition is typically done in multiple stages (2). The first stage of a typical SAR target recognition algorithm is a prescreener, where regions of interest are located. A constant false-alarm rate (CFAR) detection algorithm is often used in the first stage. The second stage may be a discriminator, where natural clutter false alarms are rejected. In this stage, textural features are extracted from the target-size windows applied to the neighborhood of pixels located by CFAR detection. The third stage is typically a classifier, where the targets are classified and human-made discretes are rejected. Extensive studies in statistical pattern recognition have been done, and many different pattern classifiers are used in target recognition applications. For SAR target recognition, spatial matched filters are investigated (2) for the recognition of ground vehicles. Novak’s (2) spatial matched filter approach in target recognition is explained in later sections. Although SAR technology provides many advantages over RAR there are many applications for which RAR is important. The azimuth resolution of SAR is much better than that of RAR, but it is typically based on the assumption that the movement of the radar is at constant speed and direction. When the radar is moving at a rapidly changing speed and trajectory, as in the case of the radar mounted on fighter airplanes or self-guided weapons, the assumptions for SAR processing are not correct. Millimeter wave RAR represents the next generation of military smart sensors for detection, tracking, and surveillance due to its high-range resolution, which is critical in target recognition. The MMW RAR technology is sufficiently advanced, and the range resolution of MMW radar is sufficiently high to discriminate tactical targets at a distance of several kilometers (3). The target recognition by hierarchical modeling of high-range resolution (HRR) MMW radar signatures is discussed in this article. Artificial neural networks (ANN) have been widely used in target recognition. The use of ANN in radar target recognition is also discussed in this article. ANNs are used as pattern classifiers and feature extractors and in model adaptation and fuzzy classification in radar target classification. Feed-forward neural networks have been used as pattern classifiers after being “trained” with known target samples. The leaning algorithm of neural networks provides a powerful tool for adaptation of the classifier to input vectors (4). A self-organizing 1

2

RADAR TARGET RECOGNITION

feature map (SOM) has been used as a feature extractor (5) for radar target recognition. In combination with Kohonen’s learning vector quantizer (LVQ) for supervised classification, SOM has been applied to the recognition of ground vehicles from MMW HRR radar signatures (5). Perlovsky, et al. (6) suggested the modelbased neural network to include a priori information to an ANN. This approach can reduce the search space of the neural network by incorporating a priori information to the adaptability of an ANN. The fuzzy neural network approach (7) is also suggested to classify targets that may belong to more than one class. The advances in neural network approaches can improve the performance of target recognition algorithms further. There are many approaches to incorporating information from many different sources for radar target recognition. By fusing the information from more than one sensor, the accuracy of radar target recognition may be improved. Two approaches in utilizing information from multiple sensors are discussed in this article. IFSAR provides elevation information, in addition to two-dimensional radar images, by processing interference between radar returns received by two different antennas. By processing IFSAR images and fusing to SAR of visual images, the accuracy of the target recognition can be improved substantially. The approach of combining IFSAR and visual images using image registration approach is discussed in this article. There are statistical approaches in data fusion, and Bayesian data fusion approaches are used in radar target recognition (8). In this approach, features from polarimetric SAR images are fused to improve the recognition accuracy. Radar target recognition is a complex problem, and no single algorithm performs better than other algorithms with different types and modes of radar. In this article, different approaches for radar target recognition are discussed in terms of radar types and approaches.

Hierarchical Modeling Approach for Noncooperative Target Recognition Target recognition using hierarchical modeling of MMW RAR radar signatures is considered in (9), and the algorithm is tested with MMW radar data from a noncooperative target identification (NCTI) database. In radar target recognition problems, the targets may be shifted and scaled from the trained data. The change in look angle results in a scale change in radar signature, and the change in azimuth angle results in a shifting of radar signature. Therefore, a classifier using features extracted from the radar signatures should classify targets reliably even when radar signatures are shifted or scaled. In general, parameters of time series models such as ARMA models are not scale invariant. Thus, the classification of scaled targets with model parameter features can result in poor classification accuracy. There are a few approaches to make a classifier more reliable to shifting and scaling. For example, the use of scale invariant features or a training with scaled samples can make a classifier more robust to scaling. The hierarchical ARMA modeling approach is briefly discussed in the following. Suppose that a continuous signal x(t) is a training sample. The classifier needs to classify the scaled signal of x(t) correctly as the same class as x(t). The scaled signal of x(t) is given by

One approach to achieve this is to train a classifier with features extracted from scaled signals of x(t). For example, different features at m different scales are extracted from scaled signals yα1 (t), yα2 (t), . . ., yαm (t); then the classifier is trained with these multiscale features. If the number of scales included in the training is large enough, the classifier will classify signals having large-scale changes. However, there are at least two potential problems with this approach if the signal is a discrete signal {x(i), i = 1, . . ., N}. First, the original signal is defined only at discrete points, and the signal at the finer scale is not defined at certain points. Second, feature extraction is performed multiple times with a single training sample,

RADAR TARGET RECOGNITION

3

Fig. 1. A decimation filter.

and the computational complexity increases linearly as the number of scales increases. These difficulties can be solved by the hierarchical modeling approach. The hierarchical modeling approach presented in this section extracts multiscale features without adding much computational complexity. A discrete signal can be scaled to a coarser scale or a finer scale by decimation filtering or interpolation, respectively. We will first consider the decimation filtering of a signal and its effect on the statistical model, and then we will consider the scaling to a finer scale as a modeling process. A decimation filter is defined as a local averaging (finite impulse response [FIR] filtering) followed by a down-sampling process, as shown in Fig. 1. If the down-sampling rate is m, the decimation-filtered signal represents the signal at the scale reduced by the factor of m. Let H be a FIR filter of length r and ↓ be the down-sampling operator of factor m.

Suppose that a signal at a coarser scale ym (i) is obtained by decimation-filtering of the original signal x(i).

Suppose that the signal {x(i), i = 1, . . ., N} follows an ARMA(p,q) model.

where {w(i)} is a zero mean white noise sequence with variance σ2 w , and aj ’s and bj ’s are real coefficients. Equation (5) can be rewritten as

where

and δ is the unit delay operator, and we assume that the roots of Ap (δ) and Bq (δ) lie inside of the unit circle for stability and invertability of the model.

4

RADAR TARGET RECOGNITION

To find features at coarser scale, the model at a coarser scale should be considered. The following theorem summarizes the results on the modeling of a decimation-filtered ARMA process. The decimation-filtered process {ym (i)} defined in Eq. (4) follows an ARMA(p,q∗) model, where the order of AR polynomial is p, the order of the MA polynomial is q∗ = [(p(m − 1) + r + q − 1)/m], and the model parameters can be obtained from the model parameters of x(i).

where

Also, the AR polynomial Cp (δ) of aggregated data satisfies the following relation:

where r1 , . . ., rp are roots of Ap (δ). The proof can be found in (9). The ARMA model parameters are shift invariant because this model’s parameters depend only on mean and correlations. Power spectral density can be estimated by using ARMA model parameters and is also shift invariant and provides features that are intuitively appealing. For example, spectral peaks or notches represent the presence or absence of a frequency component in the signal. For radar signal classification, ARMA power spectrum features at multiple scales are used. Power spectral density of an ARMA process can be estimated by an extended least squares (ELS) method. Suppose that x(i) is an ARMA(p,q) process and w(i) is the input white sequence with variance σ2 as defined in Eq. (5). Let Rxx (k) be the autocorrelation functions of x(i), and Rxw (k) be the cross correlation between x(i) and w(i). The Yule–Walker equation for the ARMA process x(i) is given by the following:

The AR parameters are estimated by solving the above Yule–Walker equations. By using the estimated AR parameters, the MA component of x(i) can be obtained by filtering AR component from x(i).

The power spectral density of the ARMA process x(t) is estimated from the correlations of xma (t) and the AR parameters estimated by Yule–Walker equations. The ELS power spectrum estimation algorithm is summarized as follows.

RADAR TARGET RECOGNITION

5

ELS Spectrum Estimation Algorithm Step 1: Compute sample correlations Rxx (k) for k = 0, . . ., p + q.

Step 2: Estimate AR parameters a1 , . . ., ap by

Step 3: Compute the sample correlation of MA component xma (i) that is obtained by removing the AR component.

Step 4: Compute the ARMA power spectrum.

For each training sample x(i), the models at the other scales (both coarser and finer scales) are obtained by the hierarchical modeling approach presented in the previous section. The model at a coarser scale is obtained using Theorem 1. The AR polynomial is obtained by Eq. (10), and the correlation of the signal at the coarse scale is obtained with a proper choice of smoothing filter H, such as a Gaussian filter. Thus, the spectral density of the signal at a coarser scale is obtained by the ELS algorithm. The model at a finer scale is obtained by the approach explained in step 2. The AR polynomial of the signal at a finer scale is obtained under no-hidden-periodicity assumption. The correlation function at a finer scale is obtained by disaggregation (9),

6

RADAR TARGET RECOGNITION

and the ARMA spectrum at a finer scale is obtained by the ELS estimation algorithm. The multiscale feature extraction algorithm is summarized as follows. Multiscale Spectral Feature Extraction Algorithm: Step 1: Each radar return is normalized to zero mean and unit variance by

wherem ˆ and σˆ 2 are sample mean and sample variances of x(i). M K-dimensional features from M scales (including coarser and finer scales) are obtained from the normalized radar returns by the following procedure. Step 2: For each training sample, the AR parameters and correlations are estimated by the ELS algorithm. For k = 0, 1, . . ., K − 1, the power spectrum is estimated at ω = πk/K. The logarithm of the power spectral density forms a K-dimensional feature vector. Step 3: At each coarser scale, a feature vector is obtained by estimating the power spectrum using the ELS method, with model parameters obtained by the hierarchical modeling approach. The logarithm of the power spectral density forms a K-dimensional feature vector at a coarser scale. Feature vectors at multiple scales are obtained by repeating this step at coarser scales. Step 4: At each finer scale, a feature vector is obtained by estimating the power spectrum using the ELS method, with model parameters obtained by the hierarchical modeling approach. This is repeated for other finer scales, and multiple K-dimensional feature vectors are obtained from the logarithm of the power spectral density. Classification is done by a minimum distance classifier with multiple prototypes. In this approach, each training sample generates M prototypes corresponding to M scales. Therefore, if there are N training signals for each class, then NM prototypes will be available for each class. Let us assume that there are N k prototypes z1k , . . ., zkNk in the class k ∈ {1, . . ., K}. For a test pattern x, the distance to the class k is defined by

where the intersample distance d(x, z) is the Eucledian distance between x and z. The distance Di is the smallest of the distances between x and each of the prototypes of the class k. The test pattern x is classified by the minimum distance decision rule: x is classified into class k if Dk < Di for all i = k. In Ref. 9, the hierarchical model-based features are tested with NCTI data. Figure 2 shows a typical NCTI radar signature and estimated power spectral density. In Ref. 9, about 95% of classification accuracy is reported with 5000 MMW RAR radar signatures.

Neural Network Approaches Neural networks have been widely used in radar target recognition applications (4,5,6,7). A good survey on the use of neural networks in target recognition can be found in Ref. 4. Neural network approaches have been popular because they have many promising qualities in target recognition applications. Neural networks have the capability for adaptation to additional targets and environments. The existence of powerful learning algorithms is one of the main strengths of the neural network approach (4). Neural networks also offer techniques

RADAR TARGET RECOGNITION

7

Fig. 2. A HRR radar signature from NCTI database and its power spectrum estimated by hierarchical modeling.

for selecting, developing, clustering, and compressing features into a useful set, and they provide automatic knowledge acquisition and integration techniques for target recognition systems. Feed-forward neural networks have been used as a pattern classifier for the target recognition. (10). Let xi be the multidimensional feature vector extracted from a radar image, and let S be the index set of the target patterns.

The optimum receiver (10) defined by

8

RADAR TARGET RECOGNITION

can be implemented as a neural network:

where T is the threshold and (x) is the heavyside step function. Roth (10) showed that detection of target patterns out of a set of P of patterns can be handled by the preceding feed-forward neural network. Neural networks have been also used as feature extractors for target recognition. Kohonen’s selforganizing map (SOM) and LVQ have been used in the two-stage target recognition approach (5). SOM is based on unsupervised competitive learning where only one output node, or one per local group of nodes at a time, gives the active response to the current input signal, and it clusters input vectors into preselected C classes by adapting connection weights to nodes in the network, and is used as a feature extractor in Ref. 5. At each iteration of the SOM algorithm, the best matching node c is selected by

where x is the current input vector, and {m1 , . . ., mC } is the set of nodes (cluster centers). Then each node mi located in the neighborhood of the node c is adapted by the learning rule:

where the gain ηi (t) can be a simple monotonically decreasing function of time or a Gaussian gain function defined by

The learning rate α(t) and the kernel width σ(t) are monotonically decreasing functions, and their exact forms are not critical. LVQ is used as a supervised classifier of the features extracted by SOM in Ref. 5. In Ref. 5, more than 94% of accuracy is reported in the target recognition experiment, with MMW data having five types of ground vehicles. Recently, the model-based neural network (MBNN) was introduced (6) to combine a priori knowledge of models of data with adaptivity to changing data properties. The learning and adaptation of the MBNN is done by iterative estimation of association weights and model parameters. Different statistical models for different physical processes, background clutter, outlier, target pixels, and so on, are also introduced in Ref. 6. This approach has the potential to improve target recognition performances by allowing inclusion of a priori information in addition to the adaptability of a neural network. Fuzzy neural networks are also used in radar target recognition. Fuzzy ARTMAP and EMAP neural networks are suggested (7) for radar target recognition. Fuzzy neural networks allow us to make soft decisions in classifying a target, and each input vector can belong to more than one class. The fuzzy association between the input vector and the classified target can improve the performance and the complexity of the adaptation.

Exploitation of Elevation Data from IFSAR IFSAR is a technique to generate high-resolution digital elevation model (DEM) based on the phase difference in SAR signals received by two spatially separated antennas (11). There are drawbacks in height maps derived

RADAR TARGET RECOGNITION

9

Fig. 3. An IFSAR image.

from IFSAR data: The data are noisy and the spatial resolution is much inferior to that of visual data. The spatial resolution is further degraded by the noise removal step. Figure 3 shows a height map produced by a real IFSAR. A typical IFSAR elevation image is noisy and needs to be filtered before it can be reliably used. Also, there are regions with “no data” that result either from the fact that the original scene was not on a rectangular grid or from radar geometry effects, which cause some points not to be mapped. Interpolation and nonlinear filtering techniques are used to filter the elevation data. Positioning of IFSAR and visual data allows for the fusion of clues from both sensors for target recognition. It is needed to overcome various difficulties resulting from the limitations of the sensor. For example, building detection requires the extraction and grouping of features such as lines, corners, and building tops to form buildings (12). The features extracted from visual data usually contain many unwanted spurious edges, lines, and so on that do not correspond to buildings. The grouping stage requires complex and computationally intensive operations. Further, the height of a building is typically estimated by extracting shadows and sun angle when available and is not reliable when the shadows are cast on adjacent buildings. Another drawback of methods based exclusively on visual data lies in their sensitivity to imaging conditions. IFSAR elevation data can be used in conjunction with visual data to overcome the aforementioned difficulties. Current IFSAR technology provides sufficient elevation resolution to discriminate building regions from surrounding clutter. These building regions are not well defined from a visual image when the buildings have the same intensity level as their surrounding background. Similarly, a building having different colors may be wrongly segmented into several buildings. IFSAR data are not affected by color variations in buildings and therefore are better for building detection. Figure 4 shows a visual image and edges detected by the Canny operator for the area shown in Fig. 3. The top part of Fig. 4 shows a building with two different roof colors and roof structures on many buildings. Many spurious edges not corresponding to the building appear in the edge map shown on the bottom right of Fig. 4. Using the IFSAR elevation map shown in Fig. 3, buildings and ground regions are labeled using a two class

10

RADAR TARGET RECOGNITION

Fig. 4. Visual image and edges detected by the Canny operator.

classifier. The IFSAR and visual images are registered. Figure 5 shows the result of registration of a visual image and the segmented elevation image. Features corresponding to roads, parked cars, trees, and so on are suppressed from the visual images using the segmented buildings derived from the IFSAR image. The locations and the directions of edges in the segmented image are estimated and are used to locate edges of buildings in the visual image. In the visual image, an edge pixel corresponding to each edge pixel in the registered height image is searched in the direction perpendicular to the estimated direction in the height

RADAR TARGET RECOGNITION

11

Fig. 5. Buildings segmented from the IFSAR image overlaid to visual image.

image. If an edge is found within a small neighborhood, the edge pixel is accepted as a valid edge of a building. If such a pixel is not found in the neighborhood, the edge is not accepted. Figure 6 shows the refined edges obtained by searching in the neighborhoods of height edges. Most of building edges in the height image are found while the unwanted edges are removed.

Spatial Matched Filter Classiﬁers for SAR (2) A typical target recognition using SAR is done in multiple stages and is illustrated by the block diagram in Fig. 7 (13). In the first stage, a CFAR detector prescreens by locating potential targets on the basis of radar amplitude. Since a single target may produce multiple detections, the CFAR detections are clustered (grouped together). Then a region of interest (ROI) around the centroid of each cluster is passed to the next stage of the algorithm for further processing. The second stage takes each ROI as its input and analyzes it. The goal of this discrimination stage is to reject natural-clutter false alarms while accepting real targets. This stage consists of three steps: (1) determining the position and orientation of the detected object, (2) computing simple texture features, and (3) combining the features into a discrimination statistic that measures how “targetlike” the detection object is. The third stage is classification, where a 2-D pattern-matching algorithm is used to (1) reject clutter false alarms caused by man-made clutter discretes (buildings, bridges, etc.) and (2) classify the remaining detected objects. Those detected objects that pass the second stage are matched against stored reference templates of targets. If none of the matches exceeds a minimum required score, the detected object is classified as clutter; otherwise, the detected object is assigned to the class with the highest match score. Matched filters are investigated in 2 as pattern-matching classifiers in the target recognition system shown in Fig. 7. They are synthetic discriminant function (SDF), the minimum average correlation energy (MACE) filter, the quadratic distance correlation classifier (QDCC), and the shift-invariant 2-D

12

RADAR TARGET RECOGNITION

Fig. 6. Edges refined using IFSAR result.

pattern-matching classifier. The basic structure of the SDF and MACE filter is characterized in the frequency domain by

where H denotes the DFT of the spatial matched filter. The matrix X is composed of a set of target training vectors obtained by taking the DFT of the target training images. The vector U represents a set of constraints imposed on the values of the correlation peaks obtained when the training vectors are run through the spatial matched filter. The matrix A represents a positive definite weighting matrix. A is an identity matrix for SDF

RADAR TARGET RECOGNITION

13

Fig. 7. Block diagram of a typical baseline target recognition system. [Adapted from Novak et al. (13)].

and is the inverse of the following matrix D.

where N is the number of training images and p is the dimension of the training vectors. In the QDCC, the DFT of the spatial matched filter is expressed by

where m1 and m2 are means of the DFTs of the training images for classes 1 and 2, respectively. S is a diagonal matrix defined by

where M 1 and M 2 are matrices with elements of m1 and m2 placed on the main diagonal, and X i and Y i are ith training vectors from classes 1 and 2, respectively. In the shift-invariant 2-D pattern-matching classifier, the correlation scores are calculated by

where T is the DFT of the dB-normalized test image and Ri is the ith reference template. Novak et al. (2) did extensive experiment with the high-resolution (1 ft × 1 ft) fully polarimetric SAR data. In the four-class classification experiment using four types of spatial matched filter classifiers, it is reported that all targets are correctly classified (2).

Multisensor Fusion (8,14,15) Research has been conducted on multisensor fusion for target recognition. Some of the motivating factors of such research are increased target illumination, increased coverage, and increased information for recognition. Significant improvement in target recognition performance has been reported (8) when multiple radar sources are utilized using sensor fusion approaches. Tenney and Sandell (14) developed a theory for obtaining the distributed Bayesian decision rules. Chair and Varshney (15) presented an optimal fusion structure given that

14

RADAR TARGET RECOGNITION

Fig. 8. A typical data fusion approach for target recognition. [Adapted from Heuter et al. (8)].

detectors are independently designed. The target recognition using multiple sensors is formulated as a twostage decision problem in Ref. 8. A typical radar target recognition approach using data fusion is illustrated in Fig. 8. After the prescreening, single-source classifications are performed first; then the fusion of decision are performed. The data fusion problem is treated as an m-hypothesis problem with individual source decisions being the observations. The decision rule for m-hypothesis is written as

For Bayes’ rule, gi (u) is a posterior probability. That is,

Since the prior probability and the distribution of features cannot be estimated accurately, a heuristic function is used (8). It is a direct extension of Bayesian approach introduced by Varshney (16), and the function gi (·) is generalized to include the full threshold range:

where P0 and P1 are prior probabilities; 1 and 0 are the sets of all i such that {gi (u) ≥ T i } and {gi (u) < T i }, respectively, with T i being the individual source threshold for partitioning decision regions; and the probabilities Pf i and Pd i are false alarm rates and probabilities of detections of each local sensor. The probabilities Pf i and Pd i are defined by the cumulative distribution functions (CDF) for each decision statistic. In practice, the CDFs are quantized and estimated from training on the individual sensor’s classifier error probabilities. In a distributed scenario, the weighting can be computed at each sensor and transmitted to the fusion center, where they will be summed and compared to the decision threshold. In 8, the data fusion approach is applied to multiple polarimetric channels of a SAR image, and substantially improved classification performance is reported.

Summary In radar target recognition, different types of radar are employed for different applications. In this article, radar target recognition approaches for different radar systems are discussed.

RADAR TARGET RECOGNITION

15

BIBLIOGRAPHY 1. A. W. Rihaczek, S. J. Hershkowitz, Radar Resolution and Complex-Image Analysis, Norwood, MA: Artech House, 1996. 2. L. M. Novak, Radar target identification using spatial matched filters, Pattern Recognition, 27 (4): 607–617, 1994. 3. J. D. Wald, D. B. Krig, T. DePersia, ATR: Problems and possibilities for the IU community, Proc. ARPA Image Understanding Workshop, January 1992, San Diego, CA, pp. 255–264. 4. M. W. Roth, Survey of neural network technology for automatic target recognition, IEEE Trans. Neural Netw., 1: 28–43, 1990. 5. A neural clustering approach for high resolution radar target classification, Pattern Recognition, 27 (4): 503–513, 1994. 6. L. I. Perlovsky et al., Model-based neural network for target detection in SAR images, IEEE Trans. Image Process., 6: 203–216, 1997. 7. M. A. Rubin, Application of fuzzy ARTMAP and ART-EMAP to automatic target recognition using radar range profiles, Neural Netw., 8: 1109–1116, 1995. 8. A. Hauter, K. C. Chang, S. Karp, Polarimetric fusion for synthetic aperture radar target classification, Pattern Recognition, 30 (5): 769–775, 1997. 9. K. B. Eom, R. Chellappa, Non-cooperative target classification using hierarchical modeling of high range resolution radar signatures, IEEE Trans. Signal Process., 45: 2318–2327, 1997. 10. M. W. Roth, Neural networks for extraction of weak targets in high clutter environments, IEEE Trans. Syst. Man Cybern., 19: 1210–1217, 1989. 11. H. A. Zebker, R. M. Goldstein, Topographic mapping from interferometric synthetic aperture radar observations, J. Geophys. Research, 91: 4993–4999, 1986. 12. R. Chellappa et al., On the positioning of multisensor imagery for exploitation and target recognition, Proc. IEEE, 85: 120–138, 1997. 13. L. M. Novak, A comparison of 1-D and 2-D algorithms for radar target classification, Proc. IEEE Int. Conf. Syst. Eng., August 1991, pp. 6–12. 14. R. R. Tenney, N. R. Sandell, Detection with distributed sensors, IEEE Trans. Aerosp. Electron. Syst., 17: 501–510, 1981. 15. Z. Chair, P. K. Varshney, Optimal data fusion in multiple sensor detection systems, IEEE Trans. Aerosp. Electron. Syst., 22: 98–101, 1986.

READING LIST J. S. Baras, S. I. Wolk, Model-based automatic target recognition from high-range-resolution radar returns, SPIE Proc., 2234: 57–66, 1994. M. Basseville, A. Benveniste, A. S. Willsky, Multiscale autoregressive processes, Part I & II, IEEE Trans. Signal Process., 40: 1915–1954, 1992. B. Bhanu, Automatic target recognition: State of the art survey, IEEE Trans. Aerosp. Electron. Syst., 22: 364–379, 1986. V. Cantoni et al., Recognizing 2D objects by a multi-resolution approach, Proc. Int. Conf. Pattern Recognition, Vol. 3, Jerusalem, Israel, October 1994, pp. 310–316. N. C. Currie, R. D. Hayes, R. N. Trebits, Millimeter-Wave Radar Clutter, Norwood, MA: Artech House, 1992. I. Daubechies, The wavelet transform, time-frequency localization and signal analysis, IEEE Trans. Inf. Theory, 36: 961– 1005, 1990. D. M. Dunn, W. H. Williams, T. L. DeChaine, Aggregate versus subaggregate models in local area forecasting, J. American Statistical Assoc., 71: 68–71, 1976. J. Geweke, Temporal aggregation in the multiple regression model, Econometrica, 46: 643–662, 1978. C. W. J. Granger, M. J. Morris, Time series modeling and interpretation, J. Royal Statistical Soc., A-139: 246–257, 1976. S. Kingsley, S. Quegan, Understanding Radar Systems, New York: McGraw-Hill, 1992. D. C. McKee et al., Model-based automatic target recognition using hierarchical foveal machine vision, SPIE Proc., 2755: 70–79, 1996. R. A. Mitchell, R. Dewall, Overview of high range resolution radar target identification, Proc. Automatic Target Recognition Working Group Conf., Monterey, CA, November 1994.

16

RADAR TARGET RECOGNITION

F. A. Pino, P. A. Morettin, R. P. Mentz, Modelling and forecasting linear combinations of time series, Int. Statistical Rev., 55: 295–313, 1987. O. Rioul, A discrete-time multiresolution theory, IEEE Trans. Acoust. Speech Signal Process., 41: 2591–2606, 1993. M. L. Skolink, Introduction to Radar Systems, New York: McGraw-Hill, 1980. N. S. Subotic et al., Multiresolution detection of coherent radar targets, IEEE Trans. Image Process., 6: 21–35, 1997. L. G. Telser, Discrete samples and moving sums in stationary stochastic processes, J. American Statistical Assoc., 62: 484–499, 1967. D. R. Wehner, High Resolution Radar, Norwood, MA: Artech House, 1987. W. Wei, The effect of temporal aggregation of parameter estimation in distributed lag model, J. Econometrics, 8: 237–246, 1978. W. W. S. Wei, D. O. Stram, Disaggregation of time series models, J. Royal Statistical Soc., B-52: 453–467, 1990. M. A. Wincek, G. C. Reinsel, An exact likelihood estimation procedure for regression ARMA time series models with possibly non-consecutive data, J. Royal Statistical Soc., B-48: 303–313, 1986.

KIE B. EOM George Washington University

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering c 1999 John Wiley & Sons, Inc. Copyright

RADAR TRACKING Radar tracking is the ability to determine the position and velocity vector of a target at any particular instant in time, to predict its position in the future, and to distinguish the desired target from other targets and clutter. For a typical radar, the direction from the radar antenna (or antennas) to the target is generally determined in the polar coordinates of range (distance), azimuth (horizontal) angle, and possibly vertical angle. For a sophisticated coherent radar, tracking targets in Doppler frequency space may also be required. Thus radar tracking can be one dimensional (range, angle, or Doppler), two dimensional (range and azimuth angle), three dimensional (range, azimuth angle, and elevation angle), or four dimensional (range, azimuth angle, elevation angle, and Doppler). For some systems, radar information is converted to Cartesian coordinates, and the tracking functions are performed in coordinates such as latitude, longitude, and height. Target tracking is necessary for a number of reasons. In order to direct a weapon such as a missile or a projectile to a target, the range, future range, and angles from the radar to the target must be determined by the radar. By knowing the position of the target relative to that of the missile, the guidance computer can direct the missile to the target. Aircraft controllers must know an aircraft’s location relative to other aircraft in the vicinity, and by tracking the positions of all the aircraft in their assigned sectors, they can control the spacing of the aircraft to ensure flight safety.

Examples of Radar Trackers A police radar can determine the speed of the vehicle in the field of view of the radar by measuring the Doppler frequency of the return (echo) signal from the vehicle because the Doppler frequency is directly proportional to the vehicle’s velocity. Most police radars must track the Doppler frequency over a given period of time to ensure measurement. A missile guidance radar must continually track the target’s range, azimuth angle, and elevation angle in order to predict the future target position; thus, it is an example of a three-dimensional tracker. An airborne radar such as the APG-70 in the F15-E aircraft utilizes Doppler processing for clutter rejection, as well as range, azimuth angle, and elevation for target-tracking purposes, and is thus an example of a four-dimensional tracking radar. A phased-array radar must be capable of maintaining track simultaneously on multiple targets, while still scanning its field of regard for new targets.

History of Radar Tracking In the early days of radar, range and angle-tracking functions were performed manually. Using a device such as a track ball, the operator could keep the cross-hairs positioned on the range and azimuth angle of a detected target viewed on a display such as a plan position indicator (PPI) display. The PPI display, such as that shown in Fig. 1, provides a two-dimensional display of range and azimuth angle for a radar with an azimuth-scanning antenna. Targets result in blips on the display where the brightness (and size) of the blips are related to 1

2

RADAR TRACKING

Fig. 1. PPI radar display showing targets and clutter.

the amplitude of the target echoes at the receiver. The output of the track ball can provide readout of the target range and azimuth angle or provide the required range and angle information to weapons systems for targeting purposes. Although this was a satisfactory technique for tracking slow-moving targets such as ships, it is certainly a tedious process. To aid in the tracking of ships and aircraft, a rate-aided device was added to some systems. With rateaided tracking, the operator needed to make only fine adjustments to account for changes of the target range and angle rates with respect to the radar. With this configuration, the radar operators were better able to track faster-moving objects such as aircraft. Still, this tracking function required the constant attention of the radar operator. Automated target tracking evolved as a necessary tool to allow the radar operator to perform the tracking function efficiently. After range and angle trackers are locked onto the target, the tracker then senses any error between the current target position and that predicted by the tracker and automatically and continuously adjusts the tracker functions either on a pulse-to-pulse or scan-to-scan basis. As a result, automatic radar tracking can maintain target track more accurately than a human operator and can better follow fast maneuvering targets.

RADAR TRACKING

3

Tracking Basics For automatic target tracking, a sequential procedure must be used to acquire the target and initiate track. The three steps are target detection, target acquisition, and target track. Target Detection. In order for the received echo signal from the target to be detected by the radar, the receive signal strength in that particular range cell must be stronger than the residual noise in the radar and other interfering signals in that range cell. For a target separated from clutter, the primary interfering source is receiver noise. Although it is desired to declare a target’s presence with high probability, it is also necessary to keep the probability of false alarm (declaring a target detection when no target is present) as low as possible. The two values are tied closely together: for a given signal-to-noise (SNR), lowering the detection threshold to increase the probability of detection threshold also increases the probability of false alarm. Depending upon the target detection criteria, a SNR of 8 to 15 dB is generally required to keep the probability of detection reasonably high, while keeping the probability of false alarms at or below 10 − 6 . Probability of detection vs. false alarm curves are available in Blake (1) and a number of other sources. In many cases, the single-pulse SNR may be below the threshold, but the SNR can be improved by integrating a number of pulses. For coherent operation, the SNR improvement is directly proportional to n, the number of pulses coherently integrated. For noncoherent operation, the SNR improvement for small n, is usually near n0.8 in practical radar systems where n < 20. Most real targets are composed of complex reflecting surfaces; the scattering contributions of these separate reflecting surfaces tend to add and subtract vectorially to the overall radar cross section (RCS) of the target. The fluctuations in RCS caused by these surfaces will affect the probability of detection and false alarm. Swerling (2) has derived the probability of detection and false alarm curves for both slowly varying and rapidly varying target RCS fluctuations. For these cases, the required SNR required can be obtained from this set of curves. Target Acquisition. Target acquisition for tracking can be done either manually or automatically. For manual target acquisition, the operator needs to point the radar antenna (or an angle cursor) on the azimuth angle to the target and designate the desired target range. Alternately, the operator could use a light pen if available to designate the target azimuth angle and range to the tracker. When the particular target is within the acquisition limits of the tracker, the acquisition process can be initiated to lock the tracker up on the target range and azimuth. For automatic target acquisition, the tracker must have either a designated philosophy for selecting the target for track acquisition, or the tracker must have sufficient capability of tracking all the targets satisfying the track initiation criteria. For example, for a radar altimeter, the track would be initiated on the closest radar returns to the radar. For a scanning surveillance radar, the tracker would need to have sufficient capability to track all the targets satisfying the track criteria.

Range Track When the target has been acquired by the tracker, the tracker must determine not only the range and angular positions but also the velocity vector of the target, and it must determine the velocity components in range and angle in order to maintain track on the target. This is especially important in order to maintain track during conditions of track fade or during momentary passage of other targets or clutter returns. Target trackers differ in complexity and include: (1) dedicated single target trackers, (2) track-while-scan target trackers, and (3) multiple target trackers. For scan-to-scan and multiple target trackers, association algorithms are required to keep track of the targets, especially during crossing target events. Dedicated Range Trackers. Dedicated target trackers generally use radars with antennas that spotlight the desired target with the antenna beam and keep the antenna beam spotlighted on the target during the entire tracking process. This type of tracker is generally used with weapon systems that require continuously

4

RADAR TRACKING

Fig. 2. FMCW transmit and receive waveforms.

updated position information on the target. This is a relatively simple type of tracker and will be used to explain the principles of tracking. The tracking process will be described as composed of the following process: range tracking, angle tracking, and Doppler tracking. For a radar system, the range from the radar to a target is precisely determined from the time delay between the transmission of the radar signal and the receipt of the radar echo from the target arriving back at the radar’s receive antenna. The range (R) from the radar antenna to the target is then given by

where c = speed of light (2.997 × 108 m/s), τ = delay time between transmit and receive target echo. Because radar signals travel at the speed of light, the range to the target is approximately 150 m for each microsecond of time delay between the time the radar signal is transmitted and when the return echo signal reflected from the target arrives at the receiver. Range Tracking with an FMCW Radar. The simplest type of radar used for range tracking is that of frequency modulated carrier wave (commonly referred to as an FMCW) radar. One of the prime advantages of using an FMCW radar is that, for a given signal to noise ratio, the average transmit power is much less than the peak power required for a pulse-type radar. Transmit signal frequency is generally swept linearly over a period of time, such as shown in Fig. 2. This signal is transmitted toward the target and returns with a time delay (τ). By comparing the frequency of the received signal with that currently being generated by the transmitter, the time delay (and hence the range) can be determined from the equation

where f = difference frequency between the received signal and the transmit signal, df /dt = rate of change in frequency versus time for the transmit signal. The circuitry for a FMCW ranging system is rather simple, as shown in Fig. 3. The transmitter is coupled to the antenna through a circulator, for example to isolate the transmit signal from the receiver input. Transmit signal is reflected by the target, received by the antenna, and then mixed with the current transmit signal. The mixed signal is then amplified, filtered to remove the radio-frequency (RF) transmitter and receive signal

RADAR TRACKING

5

Fig. 3. Typical FMCW circuit.

components, and coupled to a frequency discriminator circuit. The frequency discriminator provides an output voltage that is proportional to the input frequency. Thus the output signal is proportional to the range to the target. For a moving target, the frequency of the returns from the target are not only affected by the range to the target but also by the target velocity with respect to the radar. In order to separate frequency change effects resulting from range from those resulting from target velocity, an up/down ramp waveform, such as that shown in Fig. 4 can be used. The frequency change caused by velocity essentially moves the entire receive frequency up or down, and by averaging the frequency difference between the up frequency and down frequency portions of the waveform, both the range and the target velocity can be determined from the following equations. The range is determined from the average of the frequency differences during the positive frequency ramp and the negative frequency ramp, thus

The velocity of the target relative to that of the radar is a function of the frequency difference between the positive ramp portion and the negative ramp portion. The velocity of the target in the direction toward the radar (positive Doppler frequency) is then

where λ is the wavelength of the transmit frequency. Range tracking using an FMCW radar can be accomplished simply by averaging the output voltage from the frequency discriminator, which is proportional to the time delay between the transmit and receive signals,

6

RADAR TRACKING

Fig. 4. FMCW waveform for resolving target range from velocity.

Fig. 5. Basic pulsed radar range tracker.

and hence is proportional to the target range. Extreme linearity and slope calibration of the frequency sweep is required for accurate range determination. For example, a 1% error in the linearity of the linearity or slope, can have an equivalent error in the range determination. Range Tracking with Pulsed Radar. For pulsed radar, the target range is measured from the time delay (τ) between transmit pulse and received echo from the target. Figure 5 shows the basic configuration of a range tracker used with pulsed radar. The “heart” of a range tracker is the time discriminator that enables the tracker to determine the time difference between the range reference (estimated delay time) and the actual range of the target return. The range error (r ) is normally bipolar and proportional to the range (or time) difference between the estimated range and measured range. The range error is then input to a range and velocity estimator (and possibly acceleration estimator) circuit. The function of the range error output is to drive estimated range to the measured range. In most cases, an initial range (and possibly range rate) in the general vicinity of the target range must be input to the range, velocity estimator circuit in order to enable it to acquire the target. There are three basic classes of range trackers, which will be designated as analog, digital, and computer tracker range trackers. The most common analog-type tracker circuit uses early and early–late gates, such as those shown in Fig. 6. The detected target video is input to both early and late gates. During the early gate time, the portion of the video signal existing during that time period is fed through to an integrator circuit, which integrates the signal energy during that time period. The late gate likewise feeds the video signal during the late gate time period to a second integrator. The outputs of the two integrators are compared in a difference circuit. If there is more video energy in one of the integrators, an error signal proportional to the difference is

RADAR TRACKING

7

Fig. 6. Analog early–late gate range tracker.

generated. The polarity of the error signal depends upon which integrator output is greater. The error voltage then is provided to the range servo loop circuit, which generates voltages proportional to the estimated range, velocity, and possibly acceleration. The range voltage (estimated range) drives the timing generator, which generates the early and late gate times dependent upon the range voltage. If more video energy is in the late gate time, the error voltage causes the range voltage to increase so that the partition between the early and late gates moves out in range and becomes aligned on the centroid of the video pulse. In order for the range tracker to initially acquire track, the early–late gates must be positioned so that a significant portion of the target video energy appears in the early or late gate times. An operator can accomplish this by observing a radar display and setting the initial range into the track circuit. The range tracking accuracy of the range tracker is dependent upon the signal to noise power ratio (SNR) of the signal compared to the noise in the early and late gate time periods. According to Barton (3), the standard deviation of the range error (σr1 ) on a single pulse basis is given by

where B = receiver frequency bandwidth τ0 = pulse width. Normally, the servo loop integrates a number of pulses to provide smoothing of the range voltage, which reduces the effects of noise jitter upon the range determination. For noncoherent operation, the range error is effectively reduced by 1/n, the number of pulses integrated. The resulting range error is then given by

where f r = pulse repetition frequency, t0 = observation time.

8

RADAR TRACKING

Fig. 7. Digital range track sampler.

Digital range trackers can be implemented using a number of techniques. In most cases, the range information (estimated range) is stored in a digital counter and is updated (up or down counted) depending upon the actual range compared to the estimated range. An early–late gate discriminator, such as that shown for the analog range tracker can be used, and the error voltage then drives the up/down count. A simpler method for accomplishing the discrimination function is shown in Fig. 7. For this case, a range window is positioned about the radar video, and the video voltage is sampled at equal increments across the pulse. It should be noted that the digital discriminator of Fig. 7 requires that the signal be passed through an approximately matched filter prior to sampling, if the SNR is to be optimized. The split-gate tracker performs the matched filter function by averaging over the gates, and hence can be preceded by an IF amplifier with wider bandwidth. The digital circuit then drives the center of the range window to the centroid of the target video signal by equalizing the voltages in the early and late sample times. Three samples are required, as a minimum, for this type of discriminator: an early sample, a late sample, and an on-target sample. When the tracker is centered on the target, the on-target sample voltage is maximized, thus indicating that a true target is being tracked, rather than noise. Again, the range window must initially be set to the approximate target range or caused to slew automatically until a target is detected. The analog and digital trackers described earlier are primarily intended for range tracking a single target, and in most cases the radar antenna is boresighted on the target either manually or by using an angle tracker. Tracking circuits, either analog or digital, can be designed to track targets using continuously scanning antennas. For this case, the target returns are received only by the radar during the time when the antenna beam a scans by the target, and the tracker must use prediction algorithms to estimate the position of the target on the next scan. If multiple targets are to be tracked, then individual analog or digital tracking circuits must be used for each target tracked. In most cases where multiple targets are to be tracked, especially in scanningtype radars, the tracking functions are performed in a computer using specialized tracking algorithms. Because track-while-scan tracking normally involves angle tracking as well as range tracking, the discussion on multiple target computer tracking will be deferred.

Angle Tracking Angle tracking can differ depending upon the application. For dedicated target-tracking radars, the antenna is kept boresighted on the target by the angle-tracking circuits and the antenna servo. With a continuously scanning antenna, the centroid of the target returns is measured each time the radar scans by the target, and uses an estimator to predict the position of the target on the next scan. For multifunction or phase array radars, the target track is updated each time the antenna is scanned to the target location. Because track-while-scan

RADAR TRACKING

9

Fig. 8. Conical scan radar.

and multitarget trackers normally require range (and possible Doppler tracking), the angle tracking described in this section is limited to a single target, boresighted angle tracking systems. The most common types of on-boresight trackers use conical scan, sequential lobe, or monopulse-angle-sensing techniques. Conical Scan Angle Trackers. Conical scan is the simplest angle-sensing technique in that only a single receiver channel is required. As shown in Fig. 8, the antenna beam is squinted off the antenna rotational axis. The squinted antenna beam is rotated about the antenna boresight by either rotating the antenna or nutating an offset feed. If the target is located on the antenna boresight, the target video signal maintains a constant amplitude as the antenna rotates. However, if the target moves off boresight, the target video signal will have a sinusoidal amplitude variation given by

where E0 = average magnitude of received signal, = angular distance of target from boresight, K s = antenna error slope, ωs = antenna rotator scan frequency, φ = phase angle of the return modulation relative to the scan rotation. In order to determine the transverse (azimuth) and elevation angle error components, the equation can be rewritten in the form

where t = transverse (azimuth) angle error component, e = elevation angle error component. By using the preceding equation, the angle resolver can determine the azimuth and elevation error components of the target direction from boresight. These angle error components are then coupled into the azimuth and elevation inputs of the antenna servo positioner, which then drives the antenna boresight onto the target direction. Although this is conceivably the easiest angle-sensing technique, it is susceptible to tracking errors produced by amplitude fluctuations of the target. Also, for military applications, because the conical scan modulation can be detected, modulation jammers can drive the antenna off the target. Sequential Lobe. Sequential lobe angle sensing is similar to that of conical scan, except that the beam is switched electronically between beam positions. For dual-axis (azimuth and elevation) angle sensing,

10

RADAR TRACKING

generally four beam positions are used (up, down, left, right). By comparing the amplitude of the received signals in the upper and lower beams, and knowing the shape of the antenna beams, the angular elevation angle of the target from the antenna boresight can be determined. A similar technique can be used for azimuth angle sensing. The technique can either use a single receiver channel or separate receivers for azimuth and elevation angle sensing. The advantage of the technique is that the switching of the beams can be accomplished on a pulse-to-pulse basis, thus making it less vulnerable to target radar cross-sectioned fluctuations. It, however, still has a vulnerability to modulation-type jammers. Lobe on receive only (LORO) is a variation of sequential lobe sensing. With this technique, the transmitter either uses a separate transmit horn on boresight or transmits simultaneously through all four horns. The sequential lobing is accomplished only on receive, through sequential sampling of the signals in the four horns. The advantage of LORO is that modulation jammers cannot detect the sequential modulation pattern of the receivers. Monopulse. Monopulse sensing provides the ability to determine the angle of arrival in a single pulse by simultaneously processing the signals in multiple receive beams. Figure 9 shows an example of a four-horn monopulse configuration for dual-plane (elevation and azimuth) angle sensing. The four-horn configuration shown in Fig. 9 is useful for a description of the basic process, but practical radars built since the 1960s have used more complex feed systems to optimize the sum gain, difference error slopes, and sidelobes of all channels. Amplitude-type monopulse uses simultaneous antenna beams squinted at angles off the elevation and azimuth boresights. The relative amplitude of receive signals determines the angular distance of the target off the boresight. Another type of monopulse, referred to as phase-sensing monopulse, uses separate receive apertures spaced a short distance apart, but with the beams pointed parallel with the antenna boresight. For this type of monopulse sensing, the phase difference between the receive signals determines the angular distance of the target from boresight. The monopulse feed, such as shown in Fig. 9, is normally used either to illuminate a parabolodial reflector directly or to illuminate a subreflector for a Cassegrain-type antenna. The monopulse feed is normally attached directly to a and comparator. The and comparator combines the received signals in the four beams to form a signal, a AZ signal, and a EL signal. According to Rhodes (4), amplitude sensing and phase sensing are equivalent and can be converted to and sensing. Within the 3 dB beamwidth of the pattern of the monopulse antenna, the function / is approximately linear. The target azimuth angle θ off the azimuth boresight and the elevation angle β off the elevation boresight can be determined from

where K = antenna slope (a function of the squint angle of the beams), φAZ = phase angle of the AZ signal relative to the signal, φEL = phase angle of the EL signal relative to the signal. For a point like target (target extent less than the antenna beamwidth), the phase angle between the and signals is normally either 0◦ or 180◦ , depending upon which side of the boresight the target is located. A typical configuration for a three-channel monopulse receiver is shown in Fig. 10. The and signals, after down-conversion to IF are amplified in gain-controlled amplifiers. The IF outputs from the gain-controlled amplifiers are input to amplitude-sensitive phase detectors, along with the IF outputs. The

RADAR TRACKING

11

Fig. 9. Four-horn monopulse antenna beam patterns.

Fig. 10. Three-channel monopulse circuit.

phase-sensitive amplitude detector provides a video output signal proportional to the amplitude of the signal and the cosine of the phase angle between the signal and the signal. In order to maintain a constant number of volts per degree for the output phase amplitude phase detectors, the gain of the receivers must be maintained to provide a constant output signal level at the range of the target. To do this, the sum signal is detected to provide a video signal to a range tracker circuit, which then locks the range onto the target. The output video signal is then sampled at the range of the target, and this is then used to form the gain control voltage in the three receiver channels. This signal normalization maintains the desired number of output volts per degree from the channel receivers. Close tolerances on gain and phase track of the three gain control amplifiers are required to provide the integrity of the angle error calibration. Figure 11 shows an example of a two-channel monopulse receiver. The , AZ , and EL microwave signals out of the monopulse comparator are switched in a RF commutator so that on receive pulse 1, + AZ , and

12

RADAR TRACKING

Fig. 11. Two-channel monopulse circuit.

− AZ signals are in receiver channels 1 and 2, respectively. On the second receive pulse, + EL , and − EL are coupled in to receive channels 1 and 2. On the third receive pulses, the polarities are switched, so that the − AZ , and + AZ signals are input to channels 1 and 2 on the third pulse, and the − EL , and + EL signals are in channels 1 and 2 on the fourth pulse. The receive signals in channels 1 and 2 are down-converted from RF to intermediate frequency (IF), amplified in gain-controlled amplifiers and subsequently converted to video. The decommutator circuit then uses the difference in the video outputs to form the AZ error and the EL error signals. The error signals are then coupled into the antenna servo to maintain the antenna boresight on the tracked target. In order for the angle circuit to maintain a constant number of volts per degree for the angle error output, the gain of the receivers must be maintained to provide a constant (on the average) output signal level at the target range. In order to accomplish this, the sum of the video receive signals is provided to a range tracker circuit, which then determines the range to the tracked target. The output video signal is then sampled at the range of the target, and this is then used to form the gain control voltage to both the receiver channels, in order to maintain relatively constant target output levels in the receivers. The advantage of the two-channel receiver is that it eliminates the need for a third receiver channel, and that it eliminates any zero drift in the phase-sensitive amplitude detectors. This is at the expense of 3 dB less efficiency (as compared to the three-channel configuration), and potential sensitivity to target amplitude fluctuation if the two channel gains are not identical. The disadvantage is that the angle error is only determined on alternate pulses, and any noise or differential losses in the switching process will tend to degrade the accuracy and precision of track. Thus, some sacrifice in tracking precision will be suffered in comparison to a full three-channel monopulse angle tracker. Angle Error Sources. The accurate determination of the angle to the target is influenced by a number of factors including radar-dependent errors, target-dependent errors, and propagation effects. Radar-dependent errors include the effects of thermal noise, antenna misalignment and cross coupling errors, and radar instrumentation error sources. The angular errors resulting from thermal noise can be quantified and are primarily dependent upon the signal-to-noise ratio. For a conical scan radar, the variance in the angle determination is

RADAR TRACKING

13

given by

where K s = conical scan angle error slope, θc = antenna 3-dB bandwidth, SNR = signal-to-noise power ratio, f r = pulse repetition frequency, βn = servo bandwidth, B = receiver bandwidth, τ = pulse width. For a monopulse angle tracker, the variance is given by

where K m = monopulse error slope, θm = antenna 3 dB beamwidth. Glint is one of the most significant target dependent, angle error sources for complex targets such as aircraft and ships. Complex targets consist of multiple scatterers separated in angle and range. Rather small variations in target aspect angle can change the phase relationship of the separate scatterers, resulting in large variations of the amplitude and indicated angle to target. Depending upon the extent of the scatterers and their phase relationships, the indicated angle to the target can actually be outside the physical dimensions of the target. In order to understand the phenomena, the slope of the phase front resulting from two isolated point targets is given by Dunn and Howard (5) as

where a = relative amplitude of the one scatterer to the stronger scatterer, L = lateral distance separating the two scatterers, φ = relative phase of the two scatterers, ψ= angle between the perpendicular bisector of the scatterers and direction to the radar. If ψ is set equal to zero, then

14

RADAR TRACKING

Fig. 12. Phase front warpage caused by two scatterers.

The preceding equation has been plotted in Fig. 12 for a = 0.9. As can be seen from the plot, when the relative phase angle between the two scatterers approaches 180◦ , the indicated angular position is outside the directions to the two scatterers. Propagation effects such as multipath and ducting can also affect the angular indication, especially for elevation angle sensing. Multipath is a severe problem for low angle tracking of targets, where the multipath return from the terrain is in the main beam (or possibly even the sidelobes) of the antenna. Multipath contributions can be both from specular and diffuse reflections from the terrain, and their contributions are a function of the surface roughness. For specular reflections, the return signals can be expressed as

where At = free-space target amplitude at the antenna, Ar = free-space multipath (target image) amplitude at the antenna, f (θt ) = antenna voltage gain in the direction of the target, f (−θr ) = antenna voltage gain in the direction of the multipath return, ρo = magnitude of the reflection coefficient, α = relative phase angle between the direct and the multipath return. In a sense, the angular errors caused multipath effects are similar to those associated with the two-point scatterer situation.

RADAR TRACKING

15

Doppler Tracking Tracking targets in a clutter background is one of the major problems for radar trackers. Fortunately, terrain clutter generally has a narrow specular extent. If the target is moving, the Doppler frequency of its return is normally outside that of the terrain clutter. Doppler filtering can then be used to reject clutter, while keeping the target returns. The simplest type of Doppler filtering is obtained by using moving target indication (MTI) processing. More advanced Doppler processing enables the determination of the Doppler frequency (and hence the radial velocity) of the target. MTI or Doppler filtering must be applied to both sum and difference channels of a monopulse tracker, and in conical scan or lobing radar must be able to cancel the modulation induced on the clutter by scanning. MTI Processing. For a ground-based radar system, MTI processing provides the capability to reject clutter by filtering out the returns whose spectral content is close to the pulse repetition frequency (PRF) of the radar. This is accomplished by comparing the phase and amplitude of the target returns on successive pulse intervals. Coherent radar operation is normally used for MTI processing, however, coherent on-receive MTI processing can be used with noncoherent radars to provide most of the benefits achieved with coherent radar processing. In MTI processing, if the phase and amplitude of the returns stay constant over two, three, or more pulse intervals, then the returns are assumed to be associated with clutter and are rejected. The phase (and possibly the amplitude) of moving target returns will change on a pulse-to-pulse basis and are not rejected by the MTI filtering. MTI-filtered target returns can then be tracked by range and angle tracking circuits. Doppler Filtering. Full Doppler tracking requires coherent radar operation and can improve the tracking ability of the radar using narrow filter bandwidths, thus increasing the sensitivity of the radar. The Doppler filtering can also enable the determination of the actual Doppler frequency of the radar returns, thus providing an exact determination of the target radial velocity. In addition, for airborne radars, the Doppler frequencies of the clutter returns are a function of the aircraft velocity and the aspect angles to the clutter patch. Thus MTI processing cannot be used for clutter rejection with airborne radars. Continuous wave (CW) radar provides the ability to track a moving target while rejecting clutter. Normally, separate transmit and receive antennas are used for CW tracking radars. An example of a Doppler phase lock loop, a simplified version of that shown by Morris (6), is given in Fig. 13. The signal input is mixed down to the center frequency (f 2 ) of the narrow band-pass filter. The input to the signal mixer is derived from a combination of the output of the phase locked oscillator (PLO) which is then mixed with the IF local oscillator (IF LO) frequency to provide the IF signal necessary to mix the signal down to frequency f 2 . Any increase or decrease in the Doppler frequency (f D ) will cause the PLO output frequency to change in order to maintain the input to the band-pass filter at frequency f 2 . The AZ and EL signals are also mixed down to frequency f 2 . The AZ and EL signals are narrowband filtered, and used to derive the AZ error and EL error signals. For a high PRF pulsed Doppler radar, a narrow pass-band filter is normally used to limit the receive spectrum to f o ± PRF/2. This has the effect of converting the pulsed signal to CW, at which time the CW Doppler tracking configuration described earlier can be used for the Doppler and angle tracking. If range tracking of the signal is also required, the signal must first be sampled at the range of the target prior to narrow band filtering. A minimum of two adjacent range cell samplers, each followed by narrow band Doppler filter, are required to accomplish range tracking. In this case, the range samplers act as early and late gate samplers, and by comparing the output Doppler amplitudes, the range tracker can keep the received pulses centered between the two range samplers. Acquisition with a pulsed Doppler tracker can be a complicated process. In order for the Doppler tracker to acquire the target, both the range and Doppler frequency must be established in order to provide the Doppler output signals required for tracking. Thus, unless the range and the Doppler frequency is known (and normally they are not), a search process both in range and Doppler frequency must be initiated to find the target and initiate track. Other configurations exist for Doppler filter trackers. Barton (3) describes a technique using narrow-band filters offset above and below from a center frequency. By

16

RADAR TRACKING

Fig. 13. Doppler phase lock loop.

comparing the amplitudes out of the high- and low-frequency narrowband filters, an estimate of the Doppler frequency can be obtained on a single pulse basis. Digital Doppler Processing. With the advent of high-speed digital processors, the Doppler frequencies can be computed directly. For this type of processing, the receive signals are normally converted to I and Q digital format using high-speed analog to digital converters. The range-sampled I and Q signals can be stored for a selected number of pulse repetition intervals (PRI), and input to fast Fourier transform (FFT) computational routines. The FFT computes the detected amplitude versus Doppler frequency for each sampled range. Tracking algorithms can then use the detected targets out of the FFT processor to establish the range track, and subsequently angle and Doppler track.

Radar Ambiguities Generally radars are classified as low, medium, or high PRF radars. For low PRF radars, all the target (and clutter) returns are received prior to transmission of the next radar pulse. With high PRF radars, the Doppler frequencies of all the target (and clutter) signals are less than that of the radar PRF. Low PRF radars, which are unambiguous in range, are generally ambiguous in Doppler, whereas high PRF radars are normally highly ambiguous in range. Medium PRF radars can be ambiguous in both range and Doppler. Range Ambiguities and Eclipsing. Figure 14 shows receive signals over several PRI. The returns from target 1 occur within the same PRI as the transmit pulse that initiated the target returns, and so target 1 range is unambiguous. The returns from target 2 occur at the same times when other transmit pulses are being generated. Because receiver returns are normally disabled during the transmit pulse times to prevent receiver saturation (and possibly burn-out), target 2 returns are eclipsed, and not detected in the receiver. The returns from target 3 are from a range exceeding the unambiguous range, so that the returns in the current PRI are associated with pulses transmitted several pulses earlier. Thus, from the radar display, the returns from target 3 appear to be from a much closer range.

RADAR TRACKING

17

Fig. 14. Range ambiguities and eclipsing.

Range eclipsing occurs quite frequently in high PRF radars because of the relatively high transmit time duty factors. Even on medium PRF radars eclipsing must be avoided for reliable target detection. Eclipsing can be avoided by changing the PRF when the radar determines that the tracked target range is approaching an eclipse situation. An alternate solution is to switch between two or more PRI, so that the target will be visible in the PRIs in which it is not eclipsed. Clutter returns with delay times exceeding the PRI (second-time around returns) can cause serious problems to an MTI radar. This is because many MTI radars employ pulse-to-pulse stagger to avoid blind ranges. With PRF-staggered MTI radar, second time around clutter returns are not cancelled because the apparent range changes from pulse to pulse. In general, range ambiguities need be resolved, especially for medium and high PRF radars. Even for relatively low PRF radar, such as the AN/MPS-36 instrumentation tracking radar with a 320 Hz PRF (unambiguous range of 253 nm), the radar when its return is augmented by a transponder can track missiles many thousands of miles. A number of methods are available for resolving range ambiguities. One method is to use a form of PRF stagger in which the transmission time is varied on a pulse-to-pulse basis. The only receive pulses that align on a pulse-to-pulse basis are those corresponding to the destagger associated with that specific number of PRI. Another method is to apply intrapulse coding on the transmit pulse in which the coding is changed on a pulse-to-pulse basis. On receipt, receive signals can then be associated with the particular transmit pulse responsible for the target returns. Doppler Ambiguities and Blind Speeds. Figure 15 illustrates receive signals (in frequency space) for a pulsed coherent radar. The spectral content of the clutter returns are centered about the PRF frequency lines denoted by f o ± nPRF . Target 1 has a Doppler frequency that is less than the PRF, and so its Doppler can be determined unambiguously. Target 2 Doppler frequency is at a multiple of the PRF, and because the clutter returns are normally much higher than those of the target, it is highly unlikely that the target will be detected. In fact, most coherent radars intentionally reject frequencies around the f o ± nPRF frequencies, specifically to reject clutter. Target speeds associated with Doppler frequencies of f o ± nPRF are referred to as blind speeds. Target 3 Doppler frequency exceeds that of the PRF so that the actual Doppler frequency cannot be determined from the receive spectrum. Blind speeds can be avoided by several methods. Many coherent MTI radars avoid blind speeds by varying the PRF on a pulse-to-pulse basis. By appropriately selecting a number of different PRFs, and switching PRFs on a pulse-to-pulse basis, blind speeds can be avoided over a large range of target velocities. However, as noted previously, second-time-around clutter returns will pose a problem to this type of processing. Most Doppler radars require a constant PRF during the coherent processing interval (CPI). Doppler radars can avoid blind speeds by switching to a different PRF when it notes that it is approaching a blind speed. Alternately, the radar could transmit groups of pulses at different PRFs so that at most only one group would be at the blind speed.

18

RADAR TRACKING

Fig. 15. Doppler ambiguity and blind speeds.

Resolving Doppler ambiguities can be accomplished by several techniques. If the radar is tracking the target range, the range rate determination is generally accurate enough to enable determination of which PRF multiple the target Doppler is located. If two or more groups of Doppler PRFs are used, the ambiguity can often be resolved from the measured Doppler frequencies resulting from the multiple PRFs. Multiple Target Tracking. In many cases there is a need to track multiple targets simultaneously. Continuously scanning surveillance radars, such as the FAA’s ASR-9 airport surveillance radars, must track all the targets (airplanes) within their coverage regime. This tracking must be performed on a scan-to-scan basis, and thus this type of tracking is commonly referred to as track-while-scan processing. Phased array radars are also multiple target trackers, because they normally interleave switched beam locations to track a number of targets, with scanning for new targets, as well as performing other possible functions. With these radars, the targets (or aircraft) are only viewed for a number of pulses on a scan-to-scan (or look-to-look) basis. Most modern scan-to-scan (or look-to-look) radars use computers for multiple target tracking. Because the aircraft positions typically change on a look-to-look basis, tracking algorithms must be derived to predict the estimated target positions on the next scan, based upon on previous scans. The accuracy of these predicted positions is limited by the maneuver capabilities of the targets being tracked, so that the predicted positions are only estimates of their actual positions. Association algorithms must then be used to determine (1) if detected target is associated with an established track, (2) which established target track the target should be associated with, and (3) to determine if a new track should be established, if no track association is made. Figure 16 shows a typical flow diagram for a multiple target tracker. The raw target position information, such as range and azimuth angle (and possibly elevation angle or height), is derived in the radar. Most multiple target tracker associative algorithms prefer to track in rectilinear coordinates (RN , RE , RV ) rather than polar coordinates so that the conversion must be made from polar coordinates. If North is assumed to be at zero

RADAR TRACKING

19

degrees, then

where R = range, θ = azimuth angle, β = elevation angle. The current target position (RC ) is then

After the coordinate transformations are performed on the incoming radar target data, present target detections are then compared in association algorithms to determine if the target data are associated with established target tracks. For the FAA airport surveillance radars, the radar target detections are also associated with the secondary (beacon) radar target reports. The beacon returns also include aircraft identification and reported aircraft height. The combined associations are then used to update the target tracks and predict the aircraft locations on the next scan using the track prediction and smoothing algorithms. If the target detection is not associated with any of the present target tracks, then the position information is considered for the establishment of a new target track. In order to establish a new target track, generally the target must be associated with m on n of the previous scans in order to establish a new track. After the m out of n association is made, a new track is established, and the past position information on those target detections is used to predict the target location on the next scan. Target tracks are generally dropped after a certain number of successive target track associations are missed. The information on established target tracks is then routed to radar displays and possibly to weapons systems. Smoothing and Prediction Algorithms. Radar measurements of target positions and velocities are often imprecise as a result of a number of factors, such as signal-to-noise ratio, target RCS fluctuations, multipath, and clutter contamination. Various algorithms can be used for the track smoothing and prediction to mitigate the effects of scan-to-scan position and velocity measurement errors, and thus to improve the accuracy of tracking. Kalman filters (7) are probably the best -known smoothing and prediction algorithms. Alpha, beta (α, β or α, β, γ) trackers are a subset of the Kalman filters and are the simplestbecause they use precomputed fixed gains. The α, β, γ equations applied to position and velocity smoothing are

20

RADAR TRACKING

Fig. 16. Multiple target tracking.

and for prediction are

where T = sampling period, RC = measured position, ˆ C = smoothed estimate of current position, R ˆ PC = predicted position at the time of the measurement, R ˆ RP(C+1) = predicted position T s later, ˆ˙ = Smoothed estimate of current velocity, RC ˆ˙ R PC = predicted velocity at the time of the measurement, ˆ ˙ RP(C+1) = predicted velocity T s later, ˆ¨ = smoothed current acceleration, R C ˆ¨ R P(C+1) = predicted acceleration T s later, T = time between measurements.

RADAR TRACKING

21

The precomputed fixed gains α, β, γ can vary between zero and one, with values toward one giving the greatest emphasis toward the current measurements, whereas values toward zero provide the greatest smoothing. Benedict and Bordner (8) analyzed the gains for an α, β for track-while scan application and determined the optimal selection for this application as

The performance of the α, β, γ trackers are limited by the selection of the fixed gains, which may not be optimal for all situations. Bar-Shalom and Li (9) discusses the use of Bayesian data association techniques, as well as multiple model estimators for providing superior performance for multitarget tracking.

BIBLIOGRAPHY 1. 2. 3. 4. 5. 6. 7. 8. 9.

L. V. Blake, Prediction of radar range, in M. I. Skolnik, (ed.), Radar Handbook, New York: McGraw-Hill, 1990, chap. 2. P. Swerling, Probability of detection for fluctuating targets, IRE Trans., IT-6: 269–308, 1960. D. K. Barton, Radar System Analysis, Englewood Cliffs, NJ: Prentice-Hall, 1964. D. R. Rhodes, Introduction to Monopulse, New York: McGraw-Hill, 1959, p. 41; reprint, Norwood, MA: Artech House, 1980. J. H. Dunn, D. D. Howard, Radar target amplitude, angle, and Doppler scintillation from analysis of the echo signal propagating through space, IEEE Trans. Microw. Theory Tech., MTT-16: 715–728, 1968. G. V. Morris, Doppler frequency tracking, in J. L. Eaves and E. K. Reedy (eds.), Principles of Modern Radar, New York: Van Nostrand Reinhold, 1987, Chap. 19. R. E. Kalman, New results in linear filtering and prediction theory, ASME Trans., 83D: 95–108, 1961. T. R. Benedict, G. W. Bordner, Synthesis of an optimal set of track-while-scan smoothing equations, IRE Trans. Autom. Control, AC-7: 27–32, 1962. Y. Bar-Shalom, Xiao-Rong Li, Multitarget-Multisensor Tracking: Principles and Techniques, Storrs, CT: Yaakov BarShalom, 1995.

JOSEPH A. BRUDER Air Force Research Laboratory

RADIO DIRECTION FINDING Radio direction ﬁnding is the technique of measuring radio wave angle of arrival (AOA), as illustrated in Fig. 1. A transmitting antenna radiates radio energy toward the direction ﬁnding site. At distances greater than ten wavelengths from the transmitting antenna, the radio wave can be represented as a plane wave, with linear contours of constant amplitude perpendicular to the direction of propagation. Ideally, the radiated energy propagates along the most direct path from the transmitter to the receiver. The receiving system is conventionally called a direction ﬁnder, or DF for short. Figure 1 shows AOA expressed in terms of an azimuth component in the horizontal plane and an elevation component measured in the vertical plane, relative to the horizon. A direction ﬁnder employs one or more antennas in a DF array, used to detect the incoming radio wave. The output of each antenna is applied to a radio receiver, and this signal is sampled and input to a DF computer processor for determining AOA. The DF processor may either (1) determine the direction of energy ﬂow toward the direction ﬁnder, (2) measure the direction of maximum rate of phase change across the DF array, or (3) estimate the direction of the velocity vector, normal to the plane wave fronts. A well-known example of a simple DF system, which is still in use, is a rotatable loop antenna connected to a radio receiver as shown in Fig. 2. AOA is measured by determining the direction of energy ﬂow toward the DF antenna. This is accomplished by rotating the loop for minimum audible output as indicated by the headphones, and thereby placing the null response of the loop on the AOA of the received signal. Thus, the direction to the transmitter is indicated by the broadside angle of the loop. The loop also has a null response on the reciprocal bearing, 180◦ from the true AOA. This problem of ambiguity can be resolved with an auxiliary “sense antenna.” The advent of radio communication in the late 1890s launched the development of direction ﬁnding techniques for navigation and radio transmitter location. Radio navigation and radio location are complimentary technologies exploiting a common methodology. For example, signals received at sea from known shore-based radio beacons are used for navigation by triangulation to ﬁx the position of a ship [Fig. 3(a)]. Conversely, two or more direction ﬁnders at known locations can be used to locate a radio transmitter by triangulation as shown in Fig. 3(b). For location applications, the DF result is the observed line of bearing (LOB) or bearing to the transmitter. A LOB is expressed either as the true bearing, that is, an angle measured clockwise from true North, or as a relative bearing which is measured clockwise from a reference direction such as the heading of a mobile DF platform (ship, aircraft, vehicle). True bearings are generally used to locate a radio transmitter on a map. A major factor affecting DF system performance is the process of conﬁrming that each reported bearing is associated with the correct signal. Since there may be many signals on the air with overlapping frequencies, conﬁrmation is sometimes a very difﬁcult task. Often the achievable

accuracy of a DF net is determined by the reliability with which each bearing is conﬁrmed to be associated with the correct signal. Operational DF measurements are always subject to error, and minimizing DF error is a major consideration in direction ﬁnder system design. DF error may be divided into three categories: 1. Site error is caused by reradiating structures and ground plane characteristics at the DF antenna site which distort the arriving wavefronts into nonplanar conﬁgurations. Under these conditions the estimated AOA will vary depending on the location of the DF antenna array in the waveﬁeld. 2. Measurement error may be caused by imperfections in the DF system instrumentation but it is more frequently due to perturbed conditions in the received waveﬁeld. Multipath propagation and cochannel interference create a multicomponent waveﬁeld and are common sources of error in DF algorithms that are based on a single plane wave model. Multipath occurs when the signal arrives at the DF site via two or more propagation paths. Cochannel interference is caused by other on-the-air signals transmitted at frequencies that overlap the signal of interest. 3. Propagation error, the most fundamental source of error, is introduced by the propagation medium which may deviate the radio wave off the most direct path to the DF receiver. At best, the DF system accurately measures AOA as received at the DF site, and the estimated bearing may not indicate the “true” direction to the transmitter. Propagation error, which is beyond the control of the DF system engineer, imposes a fundamental limit to achievable DF accuracy. Conventional shipboard DF operations in the high frequency band (2 to 30 MHz) illustrate all of these DF errors in a single situation (Fig. 4). Signals experience multipath propagation through the ionosphere. Also, the ionosphere introduces propagation error by deviating each of the arriving propagation modes out of the great circle plane. A conventional single plane wave DF algorithm produces DF error by treating the superimposed ionospheric modes as a single plane wave. Finally, the ship’s superstructure introduces additional site error by reradiating the incident waves into the DF array, thus creating a second and more complex source of multipath. This brief introduction has provided an overview of the science and technology of radio direction ﬁnding. For a more thorough overview of the ﬁeld, the interested reader is referred to Refs. (1–3). An extensive bibliography of direction ﬁnding literature is provided in Ref. (4). The following sections discuss various approaches to DF system design. Conventional DF design techniques based on the assumption of a single incident plane wave signal are considered. The impact of operational conditions on conventional DF system performance is emphasized along with important methods for mitigating site error. Also, modern design techniques are described which are based on the decomposition of multicomponent waveﬁelds. These DF tech-

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright © 2007 John Wiley & Sons, Inc.

2

Radio Direction Finding

Figure 1. Radio wave received at a direction ﬁnding site. Contours of constant amplitude propagate radially from the transmitting antenna, and the angle-of-arrival is characterized by azimuth and elevation.

where r = (rx , ry , rz ) is the spatial coordinate, B is signal amplitude, fo is the frequency of the wave, ko = vfo /|v|2 is the wavenumber which is a function of the scalar frequency and v is the vector velocity of propagation (typically assumed to be the speed of light in free space), and γ is a random starting phase which is uniformly distributed over [0, 2π]. The AOA information is contained in the 2πko ·r phase term and is given as

Figure 2. Rotating simple loop antenna direction ﬁnding system. The headphones are used to detect when the AOA is broadside to the loop and the received signal is minimum.

niques are generally referred to as superresolution methods. Finally, current trends in DF research are surveyed and performance beneﬁts are assessed. APPLIED DIRECTION FINDING TECHNOLOGY A radio direction ﬁnding system performs both time and spatial sampling of the ﬁeld distribution and processes the samples to estimate AOA. A DF system acquires spatial samples through a combination of individual antenna placements and/or antenna patterns. The local description of any spatial ﬁeld distribution may be estimated either from a set of spatially separated samples or from a set of spatial derivatives at a single point in space. A local description of a ﬁeld at an arbitrary point r in a Cartesian coordinate system may be developed by considering a monochromatic plane wave propagating in free space as

If it is assumed that there are M antennas in the array, then a simultaneous sampling of the output at each antenna may be expressed as a column vector X(t, r) = [x1 , . . . , xM ]T , where T denotes the transpose operation. The vector X is known as an array snapshot. If the AOA term in Eq. (1) is separated, then the array snapshot for a single incident signal may be characterized as

where s(t) = B exp(j2jπfo t + γ) is the time varying part of the signal, and the thermal noise in each receiving channel is represented by N(t) = [n1 , . . . , nM ]T . A(r) is an M × 1 column vector which represents the antenna array response for a signal arriving from an arbitrary direction (φ, ψ) as

for an array of isotropic antennas. The column vector given by Eq. (4) is referred to as the array steering vector. The collection of all array steering vectors as a function of AOA, polarization, and frequency is referred to as the array manifold. Two examples of waveﬁeld sampling are shown in Fig. 5. Figure 5a illustrates an aerial view looking down on the surface of the Earth at a single plane wave propagating

Radio Direction Finding

3

Figure 3. Similarity of radio navigation and DF radio-location technology. (a) Shipboard navigation technique using ﬁve radio beacons for position ﬁxing. (b) Radio transmitter location technique with ﬁve direction ﬁnding sites.

Figure 4. Shipboard direction ﬁnding scenario illustrating sources of bearing error. An error will result from the fact that the radio wave is not traveling on a direct path from the transmitter. The presence of multiple signals will induce a measurement error due to wave interference, and the metallic structure of the ship will cause errors due to reﬂections near the receiving antenna.

across a circular array of eight antennas. In this case, the amplitude term B Eq. (1) is constant throughout the space. The phase term 2π(fo t + ko ·r) + γ is illustrated for constant contours of nπ as a function of r. The contours plot as parallel straight lines, and AOA is orthgonal to the contours of constant phase. In the example plot of Fig. 5b, it is assumed that four signals are propagating across an interferometer array of seven antennas. The signals are of equal amplitude and are assumed to be arriving from azimuth and elevation AOAs of (45◦ , 15◦ ), (50◦ , 30◦ ), (60◦ , 45◦ ), and (45◦ , 50◦ ) respectively. In this case, the amplitude of the composite signal is not constant but varies with r. The dark contours illustrate those regions where the composite am-

plitude exceeds a normalized threshold of 0.95 units. The thin lines are contours of constant phase showing a somewhat distorted pattern. Clearly, the antenna array is not able to adequately sample all the features of the spatial interference pattern and a plane wave solution will incur multipath error; decomposing the wave ﬁeld into its individual components is required for a complete DF solution. The processing objective of radio direction ﬁnding is to determine a unique AOA which is consistent with the set of data measured on the array. The key to this process is a knowledge of the array manifold that includes site effects of the operational environment. In the most general form, the bearing estimation process requires an iterative comparison between the observed data and the array manifold for every possible combination of polarization and AOA. The remainder of this section describes DF techniques that progress from simple closed-form solutions (where stringent constraints apply to the antenna patterns, array geometry, local scattering environments, and number of simultaneous signals), to more general calibration based DF processing algorithms, and ﬁnally to DF methods that decompose multisignal wave ﬁelds. Single Plane Wave Direction Finding Direction Finding in the Absence of Site Effects. Traditional direction ﬁnding techniques assume a single uniform plane wave incident upon the DF antenna array. These DF techniques require noninvasive electromagnetic ﬁeld measurements across a region in space in which the wave front must maintain its properties as a uniform plane wave. This requirement imposes stringent constraints on the installation site, on the selection of the antenna ele-

4

Radio Direction Finding

Figure 5. Contrasting single and multiple plane waveﬁelds incident on DF arrays. (a) Single plane wave incident on a circular array of antennas. Amplitude is everywhere constant and contours of constant phase are parallel straight lines. (b) Multiple plane waves incident on orthogonal baseline interferometer array. Contours of constant amplitude and phase are distorted. (Contour plot provided by D. N. Travers.)

ments, on the array geometry, and in some cases on the class of signals against which the system can operate. The requirement that the system be noninvasive demands that (1) no antenna element within the array perturbs the response characteristics of any other antenna element (i.e., mutual coupling must be negligible) and (2) the array of antennas must be installed in a region that is sufﬁciently separated from structures which would disturb the planarity property of the propagating wave. The selection of array geometry, choice of antenna elements, and site selection are all design factors that inﬂuence the simplicity of the DF process and system performance for various signal conditions (e.g., signal-to-noise ratio [SNR], polarization, elevation). If the noninvasive constraints are satisﬁed and simple antenna elements are deployed in a favorable array geometry, then the characteristic response of the system (array manifold) may be analytically predicted as a function of AOA and polarization. In this case, AOA may be calculated by closed-form analytic solution, thus avoiding the more general requirement for an iterative search. For example, solution of Eq. (2) for phase measurements made among identical elements in a circularly disposed antenna array (CDAA) produced an estimate of AOA using a closed-form arctangent solution. Many CDAA closed-form processing techniques relax the mutual coupling constraint among the antenna elements, provided array rotational symmetry is preserved. Eleven examples of traditional DF systems that are based on the assumption of a single uniform plane wave incident upon the antenna array are listed in Table 1. The table summarizes antenna array conﬁguration, basis for AOA determination, primary advantages and disadvan-

tages, and provides literature references to more detailed descriptions. Table 2 illustrates the corresponding antenna array geometries and analytic processing algorithms associated with the eleven examples of Table 1. DF results are derived from solutions of Eqs. 1 and 2 with appropriate modiﬁcations to account for antenna patterns and array geometry. Under more general conditions, it may not be possible or desirable to impose the restrictive design constraints that are essential for traditional DF techniques. Under more general installation conditions, arbitrary and diverse antenna elements that experience mutual coupling are installed in irregular array geometries (sometimes dictated by the site) on a location that distorts the characteristics of the incident uniform plane wave. Under these conditions, the DF process must be generalized to perform an iterative comparison between the observed response vector and the array manifold to determine AOA. Further, calibration measurements to determine installed antenna response patterns are required to characterize the array manifold as a function of AOA (and polarization). DF calibration and its application to iterative DF processing are discussed in the next section. DF Under Conditions of Site Interaction. The previous discussion focused on DF systems that were isolated from electrically conducting structures, and each antenna element within the array was excited by a single, uniform plane wave. However, an incident wave induces currents on conducting structures in the vicinity of the array, and these induced currents become sources of secondary radiation which are also coupled into the DF antennas. In the presence of conducting structures, DF antennas experience

Radio Direction Finding

phase and amplitude perturbations that distort their ideal patterns. If one assumes ideal array patterns, these distortions result in erroneous DF estimates. To the extent that the structures remain stationary, the distorted antenna patterns and the erroneous DF estimates are repeatable functions of the incident signal AOA, polarization, and frequency. This repeatable characteristic provides the basis for using calibration measurements to improve DF performance under conditions of stationary site interaction. Calibration measurements to accommodate site interaction can be performed at either of two levels: (1) measurement of AOA error correction values or (2) measurement of installed antenna response patterns. The application and effectiveness of these two basic approaches for DF operation in the presence of site interaction are the focus of

5

the following paragraphs. Under either scenario, the basic procedure for performing the calibration is to record the appropriate measurement from the installed DF system while exposing it to a controlled (calibration) incident plane wave under every appropriate combination of signal parameter (i.e., azimuth, elevation, polarization, and/or frequency). Calibration for AOA Error Correction. For a low degree of residual site interaction with the DF antennas, moderate pattern distortion and moderate DF error exist. In this case, calibration for AOA error correction is effective. One conventional approach for reducing AOA errors in DF performance is to start with a carefully controlled site in which the DF array is removed as far as practical from

6

Radio Direction Finding

Radio Direction Finding

perturbing structures. To the extent that this objective can be achieved, AOA errors due to site interaction are minimized. In this situation, a reasonably good AOA approximation is obtained using the single plane wave assumption, and residual AOA error due to site interaction is reduced through a calibration correction curve. A primary criterion for AOA error correction to be effective is that the curve of corrected bearing versus observed bearing (i.e., the uncorrected AOA estimated by the DF system) is single valued. Figure 6 shows two calibration correction curves illustrating both moderate and severe AOA error. The calibration curve of Fig. 6(a) shows a moderate case of site interaction and a single valued AOA correction curve. The calibration curve of Fig. 6(b) illustrates the impact of severe site interaction and displays

7

reentrant regions. There are two intervals in which one value of observed bearing estimated by the DF system is associated with more than one corrected bearing. Calibration for AOA errors is not effective in reentrant regions, and experimental determination of the array manifold is generally necessary. Calibration of antenna responses. DF is a process that requires a priori knowledge of the installed array steering vectors. Conventional DF system development assumes a carefully controlled array of antennas having response patterns that are closely approximated by ideal predictions. When the DF array is installed on an adverse site, antenna patterns can suffer severe distortion from nearby conducting structures. In such cases, required knowledge of the installed array steering vectors must be obtained by

8

Radio Direction Finding

Figure 6. Moderate and severe site interaction AOA calibration curves. (a) Moderate site interaction with AOA errors which are correctable. (b) Severe site interaction with reentrant regions in which AOA error correction is ambiguous.

measuring antenna responses and storing this information in an array manifold consisting of antenna patterns versus frequency, AOA, and polarization. For ship and aircraft platforms, the installed array steering vectors may be measured by repeatedly turning the platform in circles to expose the antennas to all possible AOAs from a farﬁeld calibration station transmitting a wide range of signal frequencies. As an alternative, array manifolds have been obtained by performing calibrationlike measurements of antenna responses from scale-model arrays installed on highly detailed miniature models of the platform. Iterative Search DF Techniques. The essential process of any iterative search DF technique is to select the AOA of the calibrated array steering vector which best agrees with the unknown measured array response. The primary difference between DF iterative search schemes is the criterion for obtaining best agreement. A commonly used procedure for iteratively comparing observed array response to the calibrated array response vectors is a beam steering process which acts as an equivalent bank of matched ﬁlters. In this case, best agreement is deﬁned in a least mean squared sense. A digital beamformer may be viewed as a matched ﬁlter that processes the observed antenna response to produce a single (scalar) response that is maximum when the preferred direction of the beamformer best agrees with the AOA of the signal. Beam steering DF iteratively processes an observed response vector through a progression of matched ﬁlters (viz., array steering vectors), each of which represents a different AOA. Under the constraint of a normalized input vector (i.e., unit-norm), the output level of the ﬁlter is maximum when excited by a vector that matches the ﬁlter parameters. Or stated another way, maximum output is obtained from the ﬁlter whose steer-

ing vector AOA best agrees with the bearing of the observed signal. In this manner, the process scans (or steers) a simulated beam over all possible AOAs, searching for the steering direction that maximizes the beamformer response. The beam steered iterative process may be characterized mathematically by considering an array manifold of L steering vectors for a particular frequency. The array manifold is the set of array steering vectors A for i = 1, L. This implies that the DF system was calibrated at L bearings in ª

the interval [0, 360] degrees. If X denotes the observed array response vector for a signal of interest, then the output y of each matched ﬁlter is given as

where H denotes the Hermitian or conjugate transpose operation. The best AOA estimate is obtained when yi is a maximum. A typical plot of y is shown in Fig. 7. In this illustration, L = 25, and the AOA estimate is 195◦ corresponding to the maximum value associated with the array steering vector at y14 . This process results in an AOA estimate that is best in a least squared sense. Three areas of particular importance that impact the performance of DF systems which are based on iterative search techniques are (1) antenna array design, (2) adequacy of calibration, and (3) polarization effects. Antenna array design critically inﬂuences the pattern of the steerable beam. The characteristics of the steerable beam are determined by the geometry of the array, the intrinsic element patterns and the number of antennas in the array. Of primary concern in the design of the steerable beampattern is the beamwidth of the main lobe and relative level of the side lobes. Beam steering DF system performance is also controlled by the accuracy and completeness of the calibration process

Radio Direction Finding

Figure 7. Beam steered iterative search indicating a maximum when steering angle matches AOA of arriving signal.

used to characterize the array manifold. To obtain robust maxima in the matched ﬁltering process, the array manifold should be acquired at high SNR to minimize the possibility of contaminated data. The calibration data must be measured in AOA increments that permit accurate interpolation. The array manifold varies not only with AOA, but also with signal polarization. Under arbitrary polarization conditions, the steering vector must be adjusted to match the polarization and the AOA of the incident signal. For arbitrary polarization applications, the array manifold at each steering direction must consist of a pair of steering vectors for two different polarizations, for example, AV (φ, ψ) and AH (φ, ψ) for vertical and horizontal polarization. The additional maximization step required by the polarization extension of beam steering DF at each bearing is to maximize the beamformer response over all possible linear combinations of AV (φ, ψ) and AH (φ, ψ). The matched ﬁlter thus becomes a digital beamformer that processes the observed antenna response producing a single (scalar) response that is maximum when the steered direction of the beamformer matches the AOA of the received signal. DF Error Mitigation There are a number of sources of error in radio direction ﬁnding, and an exhaustive discussion is well beyond the scope of this survey; however, there are three categories common to all DF systems: (1) site errors, (2) measurement errors, and (3) propagation errors. These are considered in the following discussion. Site Errors. The antenna array generally designed to operate on a site that is clear of obstructions and reradiating structures. As discussed in the previous section, array calibration is generally used to mitigate site error in those cases where the DF antenna must operate in a cluttered

9

environment. The concern in this section is selecting a site that least perturbs the incident signal. Generally, DF systems are deployed on level, unobstructed terrain where the ground dielectric constant and conductivity are reasonably uniform within 10 to 20 wavelengths of the antenna array. Abrupt discontinuities in the electrical properties of the terrain such as nearby rivers, lakes, or coastlines should be avoided. Also DF sites in the vicinity of rock or mineral outcroppings are undesirable. Abrupt changes in terrain topology such as nearby mountains, high cliffs, or deep ravines should be avoided. In addition to selection of natural features, proximity to manmade reradiating structures should also be considered. The DF site should be clear of above ground conductors such as utility lines and wire fences. The site should not have buried conductors beneath the antenna array such as pipelines or utility distribution lines. The site should be clear of tall structures such as buildings, bridges, or water towers. It is rarely possible to achieve all of these conditions at a DF site. Generally, there are a number of other antennas deployed at a DF site, as well as buildings and facilities which support other signal intelligence gathering activities. In practice, the DF system engineer attempts to locate the antenna array on a site which contains fewest perturbing factors. Measurement Errors. Leaving aside instrumental accuracy and resolution determined by SNR, signiﬁcant DF measurement errors are caused by multipath and cochannel wave interference, which tends to corrupt the planarity of an incident wavefront. Two approaches for mitigating measurement errors are: (1) acquire data under favorable conditions and exclude corrupted data, and (2) extract the desired signal from the interference. The ﬁrst class of techniques includes wavefront tests which determine whether a single plane wave is present or if the waveﬁeld is corrupted by interference. A test which has been used in CDAA systems is to compare the array scan of amplitude versus azimuth with an ideal sum or difference array response (1). In the case of the sum beam response, one tests for a symmetrical main lobe having an expected width depending upon frequency. When this condition is satisﬁed, it is assumed that a single planewave is present and the bearing data are accepted; otherwise, the bearing data are rejected. Another test which has been used successfully in interferometer systems is that of a linear phase progression across the array (12). If the waveﬁeld consists of a single planewave, then the relative phase between separated antennas is linearly dependent upon distance. In the presence of wave interference, this condition is generally not satisﬁed. The second class of techniques includes Fourier analysis (FFT) methods which decompose the received signal into highly resolved spectral bins. In many cases, interference due to other on-the-air signals may be separated from the desired signal through a difference in spectral occupancy. In this situation, accurate DF results may be obtained by DF processing each spectral bin occupied by the signal of interest and ignoring the spectral bins corrupted by the interfering signal. Another approach is to

10

Radio Direction Finding

use the superresolution techniques described in the next section. These techniques decompose the waveﬁeld into its constituent components and permit accurate DF for each component of the waveﬁeld.

Propagation Errors. DF error caused by propagation off the great circle path is essentially beyond the control of the system designer. In this case, the signal arrives at the antenna array from a direction that is not on the direct path to the transmitter. An example is a cellular telephone signal, propagating in a dense urban environment, received after reﬂection from several buildings. The received AOA may be considerably different from the true bearing to the transmitter. Another example is the refractive error introduced by the ionosphere on HF ionospheric paths (3). The essential approach to minimimze the effect of offpath DF errors is to perform DF on the ﬁrst arriving signal. One method for implementing this technique is to acquire data on the leading edge of the waveform after a signal off-to-on transition and before the multipath components arrive. Another alternative is to decompose the complex waveﬁeld and identify the ﬁrst arriving signal through cepstral delay analysis (13). In both implementations, the intent is to estimate AOA for the ﬁrst arriving signal on the premise that it will most nearly represent the true bearing to the transmitter. A technique which has been developed to mitigate offpath errors for HF skywave signals is that of elevation angle discrimination. This technique is based on the premise that signals arriving at higher elevation angles spend more time in the ionosphere and are therefore more prone to offpath refraction than are signals arriving at lower elevation angles. In those cases where multipath propagation is present, the mode arriving at the lower elevation angle produces an azimuth estimate closer to the great circle bearing than a higher angle mode. This situation is almost always true in practice. The limiting case is a surface wave with a simultaneous skywave component over a path at sea. The surface wave invariably produces a more accurate bearing estimate.

Superresolution Direction Finding Methods DF techniques described in the previous section were primarily concerned with system response to a single component waveﬁeld. In contrast, the problem considered in this section is that of decomposing a multicomponent waveﬁeld. Superresolution methods are particularly attractive for solving the multicomponent problem since the waveﬁeld is generally undersampled in the spatial domain. That is, the dimensions of the antenna array are small relative to the scale of the spatial interference pattern. An important building block used in superresolution spectrum anaysis is the spatial covariance matrix. If the vector X denotes an array snapshot at time t0 , then X(t0 , r) = [x(t0 , r1 ), x(t0 , r2 ), . . . , x(t0 , rM )]T and for M antennas,

the spatial covariance matrix is given as

where H is the Hermitian operation and E{·} denotes statistical expectation. Each matrix element R(ri , rj ) is the averaged product of the output of the antenna located at point ri times the conjugated output of the antenna located at point rj . Subspace Based Superresolution. A large volume of work has been presented over the past two decades on superresolution techniques that are based on an eigen decomposition of the spatial covariance matrix. Many modern algorithms have their origin in the early work of Pisarenko (14) which was revived and expanded by Schmidt (15). Schmidt’s MUltiple SIgnal Classiﬁcation (MUSIC) algorithm is the most widely cited superresolution technique in the present day literature. It has been the springboard for a seemingly endless ﬂow of methods that are variations of the original approach. In this treatise, the original MUSIC algorithm is considered; however, for an extensive survey of MUSIC related techniques, the reader is referred to Krim and Viberg (16). The initial step in the MUSIC algorithm is to solve the following eigen equation

where R is the M × M spatial covariance matrix, λ is an arbitrary eigenvalue, and E is an arbitrary eigenvector. This formulation implicitly assumes that the noise background is uncorrelated white Gaussian noise. The n eigenvalues may be ordered such that λM ≥ λM−1 ≥ ··· ≥ λ1 . The corresponding eigenvectors are arranged to form the matrix RE

A threshold value is determined such that the eigenvalues greater than the threshold are assumed to be associated with eigenvectors residing in the signal subspace. Likewise, eigenvalues that are smaller than the threshold produce eigenvectors which are assumed to be in the noise subspace. If the spatial covariance matrix were M × M and there were d signals in the waveﬁeld, then the resulting eigenvector matrix would be partitioned such that the ﬁrst d columns are vectors spanning the signal subspace, and the rightmost p = (M − d) columns are vectors spanning the noise subspace. If the matrix partition corresponding to noise were denoted Rp , then the MUSIC spectrum would be given by

where A is the M × 1 array steering vector deﬁned in the previous section. AOA is determined using an approach similar to the technique described in the previous section on iterative search methods. A scalar value P is computed

Radio Direction Finding

11

Figure 8. Two signals incident on a linear antenna array that were resolved using the MUSIC superresolution algorithm. The signals are received from azimuths of 151◦ and 180◦ , and the peak at 209◦ is an ambiguity. The peak at 180◦ is somewhat smeared due to the fact that the AOA of the signal is aligned with the longitudinal axis of the array (or end ﬁre).

Figure 9. Two unresolved signals arriving from azimuths of 160◦ and 164◦ on a linear antenna array. The peaks at 196◦ and 200◦ are due to ambiguities in the array. In this case the effective aperture of the array was too small to resolve an AOA separation of 4◦ . A possible solution would be to lengthen the array by adding more antennas.

using Eq. (9) for each array steering vector A, and AOA is given by the array steering vector that maximizes P. If multiple signals were present in the incident waveﬁeld, then the MUSIC spectrum would exhibit multiple peaks. This is illustrated in the next section.

resolution can be realized for signals arriving on or near the bore sight of the array (i.e., orthogonal to the array).

Multisignal DF Example. The MUSIC algorithm is capable of simultaneous DF on multiple incident signals. To illustrate the capability, a linear antenna array geometry is considered with two interfering signals incident on the array (17). The antennas were deployed in a nine-element minimum redundancy array conﬁguration and this provided an equivalent 30 element ﬁlled array measurement. The plot of Fig. 8 shows the results obtained for one signal on array end ﬁre at 180◦ and a second signal at 61◦ off array bore sight at 151◦ . The signal at 151◦ also produced a peak at 209◦ due to the inherent bearing ambiguity present in a linear array. The effect of the end ﬁre grating lobe of the antenna array is evidenced by the relatively broad peak about 180◦ . The signal arriving at a bearing removed from the end ﬁre condition produced the more robust peaks observed at 151◦ and 209◦ . Although the MUSIC algorithm was able to provide DF results for both signals, these data clearly indicated that angular resolution becomes poorer for signals arriving from directions near array end ﬁre. A situation in which the MUSIC algorithm was unable to correctly resolve the two signals is shown in Fig. 9. In this case, the AOA separation of the two signals was 4◦ , and the AOAs were near array end ﬁre, 160◦ and 164◦ respectively. The image solutions were evident at 196◦ and 200◦ azimuth. Because the AOAs were close together and near array end ﬁre, the MUSIC algorithm was not able to resolve the two signals. Two peaks were evident in the MUSIC spectrum; however, neither one indicated an AOA correctly associated with an arriving signal. Improved AOA

Superresolution Implementation Issues. One of the primary difﬁculties encountered in the implementation of the superresolution techniques is the requirement for a precise characterization of the antenna array response (viz., array manifold). In general, the array manifold must be known for all frequencies, polarizations, and AOAs. In practice, an array deployed on highly conducting soil and on a site free of interacting structures may be characterized mathematically using ideal antenna responses (17). However, for antenna arrays deployed on shipboard, airborne, satellite, or other cluttered sites, a mathematical characterization is generally not possible and the array manifold must be determined by calibration using a transmitter at known locations. Due to the highly robust nature of the superresolution techniques, the array calibration must be done in increments of AOA, frequency, polarization, etc., which will be immune to signiﬁcant interpolation error. These issues were discussed in the previous section relating to array calibration. Another source of difﬁculty is detecting the number of signals in the waveﬁeld. One rule-of-thumb is based on the relative magnitudes of the eigenvalues. Larger eigenvalues are associated with signals and smaller eigenvalues are associated with noise. This process works reasonably well in high SNR situations, but it is unreliable for low SNR. It has been shown that underestimation of the number of signals results in poor AOA performance (18), and for this reason, system designers generally try to overestimate the number of signals; however, overestimation may be a problem in low SNR situations due to the fact that the eigen based techniques tend to produce extraneous peaks corresponding to the number of estimated signals. A num-

12

Radio Direction Finding

ber of techniques for signal detection have been developed and one which continues to be used as a baseline for performance comparisons is the Minimum Description Length (MDL) criterion developed by Wax and Kailath (19). Coherent signal interference causes a difﬁculty in the application of superresolution technology. The spatial covariance analysis proceeds upon the assumption that the matrix elements are a function of the separation between antennas and are not dependent upon the location of the antennas within the waveﬁeld. If the signals are incoherent, then the spatial interference pattern will move relative to the antenna array, and the elements of the spatial covariance matrix will depend only upon the relative separation between antennas. However, if the signals are coherent, then the spatial interference pattern will be ﬁxed in space and the elements of the spatial covariance matrix will depend both upon separation and location of the antennas. In this case, the analysis will fail. To overcome the difﬁculty caused by coherent interference, Shan et al. (20) have proposed a technique of spatial smoothing which partitions the antenna array into identical subsets and averages several covariance matrices of reduced size. This effectively moves the array relative to the interference pattern. In general, the eigen decomposition part of the computation is not a signiﬁcant burden, but the search for the AOA solution may be computationally intense. In particular, one must perform the matrix multiplication within Eq. (9) for each array steering vector leading to a solution. For an azimuth-only solution, the task is greatly simpliﬁed and the computational burden generally depends on the number of antennas in the array. In the case of 2D azimuth-and-elevation AOA solutions, a number of numerical search techniques have been applied and they encounter the well known difﬁculties of dependence on starting point, convergence to local maxima, etc. It should also be noted that the superresolution techniques require a radio receiver connected to each antenna to permit simultaneous sampling of the array. For this reason, superresolution implementations are frequently called N-channel systems. Multiple matched receiver channels are a major cost driver in the implementation of superresolution methods. Generally, the system designer must evaluate hardware cost versus the expected performance improvement as compared with a more conventional DF architecture consisting of a reference receiver connected to one antenna and another receiver that sequences through the remaining antennas in the array.

TRENDS IN DF RESEARCH Two primary areas of research in the science and technology of radio direction ﬁnding are: (1) efforts to improve DF system performance in the presence of reradiating structures and (2) investigations to improve the performance of the superresolution waveﬁeld decomposition techniques. The discussion in this section focuses on a representative subset of the many important research efforts going on.

Ongoing Developments in Array Calibration Technology Array manifold errors arise from differences between the overall system response (i.e., including site interactions, antenna elements, interfacing networks, cabling, and characteristic impedances versus frequency) and the model from which the array manifold was derived. Errors in the array manifold generally cause DF performance degradation that equals or exceeds degradations caused by measurement errors. Array manifolds are frequently based upon the model of ideal array geometry, perfect channel amplitude/phase match, complete interelement isolation, and absence of site interaction. To the extent that these assumptions are valid, the array manifold can usually be generated analytically. When these simplifying assumptions do not apply, a conventional procedure has been to generate the array manifold through exhaustive calibration measurements with the installed array responding to cooperative transmitters, a procedure that can be expensive and/or impractical. Because DF performance is critically dependent upon the array manifold, the development of efﬁcient methods for accurately and completely characterizing the array manifold is a topic of active investigation. These investigations typically begin with an initial estimate of the array manifold and an assumed underlying relationship between the installed responses of the array and the initial array manifold characterization. The objective is to minimize the quantity and difﬁculty of controlled measurements required to accurately characterize the installed patterns as a function of azimuth, elevation, polarization, and/or frequency of interest. The initial array manifold characterization is analytically estimated and is assumed to be correct except for discrete, unknown factors that account for (1) antenna mutual coupling, as was done by Friedlander and Weiss (21), (2) directionally independent amplitude and phase errors, and (3) perturbations in the location of each antenna element, as proposed by Weiss and Friedlander (22). Many research efforts seek to use a bootstrapping technique to develop a calibrated array manifold. In this approach, system responses are measured using signals of opportunity, and the initial estimate of the array manifold is adaptively modiﬁed to agree with the measured data. These developments usually exploit various constraints on array geometry, nature of modeling error for generating the initial array manifold estimate, knowledge and distribution of AOAs, statistical nature of the incident signal temporal characteristics, and so on. Constraints on knowledge of calibration signal AOA range from requiring complete knowledge to a complete lack of knowledge, in which case the signal AOA and the array manifold correction factors must be jointly estimated from multiple measurements of responses to the unknown signals. In a related approach, Gustafsson et al. (23), characterize the statistical distribution of the random differences between the initial model and the installed system response, and use this information as a basis for developing linear transformations (through weighting factors) to estimate system responses and reduce modeling errors.

Radio Direction Finding

Research in Superresolution Direction Finder Techniques Detection of the Number of Signals. Since eigen DF processing is dependent on an accurate estimate of the number of signals present, research is continuing to deﬁne more effective estimation techniques. Two techniques of hypothesis testing for signal detection using eigenvalues were developed byWax and Kailath (19) and are called the AIC and MDL methods. One of the difﬁculties encountered in the application of the AIC and MDL techniques is that performance tends to degrade signiﬁcantly for short data rapidly as the noise back ground departs from the white Gaussian assumption. In a recent work, Chen et al. (24) have proposed a technique called the canonical correlation test (CCT) method which is designed to work well in unknown colored noise and does not require a subjective threshold setting. Also Wu and Wong (25) have developed an algorithm called parametric detection (PARADE), and they contend that the performance of their method is better than that of the CCT approach, particularly in the presence of correlated signals. It has been observed that the performance of the AIC and MDL criteria tends to degrade in the presence of coherent or highly correlated signals. Ma and Teng (26) report a technique to detect the number of coherent signals through the use of a modiﬁed spatial smoothing scheme called weighted subspace smoothing (WSS). The authors suggest that the WSS technique provides signiﬁcant performance improvement over the conventional MDL method in the presence of highly correlated or coherent signals. Another difﬁculty in the application of the AIC and MDL criteria is that the performance tends to degrade signiﬁcantly for short data records or in the small-sample case. This problem has been considered by Shah and Tufts (27), who propose a nonparametric procedure designed to improve performance for short data records over moderateto-high SNRs. The authors contend that the AIC and MDL criteria do not provide ﬂexible control over the false alarm rate while their method permits a choice of false alarm rate and corresponding probability of error. In a related work, Zhu et al. (28) develop two new criteria based on informational theoretic and eigen decomposition methods and an assumed noise covariance structure. The development results in modiﬁed AIC and MDL criteria which the authors claim produce signiﬁcant performance improvement over the conventional AIC and MDL methods in the case of small-sample populations. Wu et al. (36) derived an information-theoretic criterion for estimating the number of signals in a multi-array scenario and spatially correlated noise. Simulation results demonstrate that the approach performs well with a small number of array samples and low SNR (−0.67 dB). A weakness of the AIC and MDL is that they explicitly assume the array manifold vectors are linearly independent. Manikas and Proukakas (36) showed that linear dependencies among array manifold vectors lead to ambiguities. An ambiguous array manifold can, under certain AOAs, create erroneous MUSIC peaks and cause the AIC and MDL to incorrectly estimate the number of signals. Abramovich et al. (37) address this problem by developing a sequential noise-subspace equalization (SNSE) detection

13

algorithm that estimates the number of signals when the array manifold has ambiguities. The SNSE algorithm handles an overloaded array, i.e. when the number of sources exceeds the number of sensors. Simulation results showed that the SNSE performance for the overloaded array is similar to the performance of the AIC and MDL when the array is not overloaded. Real-Time Subspace Based Implementations. Two difﬁculties routinely encountered in a DF operational scenario are: (1) tracking radio transmitters which are in motion and (2) the continual appearance and disappearance of radio transmissions. In both cases, there is difﬁculty in realizing real-time adaptive responses with the eigen based techniques since the averaging operation is used to estimate the spatial covariance matrix estimate. Averaging is necessary to obtain wide-sense stationarity and to achieve statistical stability. In a real-time signal environment, it is highly desirable to update the estimate of the number of signals present and corresponding AOAs with each new snapshot of array data. A class of fast subspace tracking (FST) algorithms has been recently developed by Rabideau (29). The techniques require on the order of O(Md) operations per update (where M is the number of antennas and d is the number of estimated signals). This is contrasted with the O(M2d) operations required in eigen decomposition methods. The author also implements a rank adaptive version of the approach called RA-FST which tracks changes in the number of signals present in the waveﬁeld. Another tracking technique has been developed by MacInnes and Vaccaro (30), and the method requires O(M2) operations per update. The authors also propose a scheme for detecting a change in the number of signal sources using each new array snapshot. The authors contend that the new algorithm is capable of tracking rapidly changing AOAs, and that the response is virtually identical to that obtained through the use of a complete eigen decomposition at each time step. Moreover, the technique is able to accurately track appearing and disappearing sources. Zha (31) has recently proposed the use of newly developed fast subspace decomposition methods which exploit the redundancy in large antenna arrays to reduce the computational burden of estimating AOA. The computational complexity is O(Md) + O(d3) per data vector update. Another proposed technique to achieve real-time implementation has been developed by performing only linear operations on the data. It is called subspace method without eigen decomposition (SWEDE). The technique was developed by Eriksson et al. (32) and is computationally simpler than the exact implementations of the eigen based methods and is also more robust to assumptions regarding noise spatial correlation. The authors assert that the approach is computationally more efﬁcient than the approximate fast eigen decomposition algorithms, and that the computational complexity of the method is O(Md2). Coherent Signal Analysis. The problem of correlated signals has been attacked primarily through the application of spatial smoothing as suggested by Shan et al. (20). In the presence of coherent signals, the primary difﬁculty arises

14

Radio Direction Finding

from the fact that the spatial covariance matrix is not statistically stationary in the wide sense. That is, the terms in the matrix depend upon antenna location as well as spatial separation. Spatial smoothing is a process whereby the array is partitioned into smaller subarrays, and the resulting covariance matrices from the subarrays are averaged. The resulting averaged covariance matrix effectively decorrelates the coherent signals. A technique has been recently proposed by Li et al. (33) to estimate AOA for coherent signal components without the use of spatial smoothing and eigen decomposition. The authors have developed a new method for 2-D spatialspectrum estimation of coherent signals using a rectangular planar array, and the method works in the presence of unknown noise environments. The authors claim that the performance of the proposed technique is similar to that of spatial smoothing in the presence of spatially white noise, and it provides improved performance in spatially colored noise environments. In another approach, Delis and Papadopoulos (34) propose an enhanced forward/backward spatial ﬁltering method that provides improved performance over spatial smoothing techniques. The authors contend that their enhanced spatial ﬁltering approach requires the same number of antenna elements as the spatial smoothing methods, and it achieves improved performance. An improved spatial smoothing technique has been proposed by Du and Kirlin (35). Two problems with the spatial smoothing method are it reduces the effective aperture of the array, and it does not take into account the cross correlation between the subarrays. The authors propose an averaging technique which utilizes the correlations between the subarrays to produce a more statistically stable estimate of the averaged covariance matrix. The authors suggest that the technique provides improved performance when the subarrays are small compared to the size of the overall array. If the transmitter is moving, then a concept known as temporal smoothing can be used to decorrelate coherent signals (38). Gu and Gunawan (39) showed that, for simulated VHF and UHF signals from a moving transmitter, temporal smoothing can resolve more closely spaced coherent signals than spatial smoothing. They showed temporal smoothing requires M+1 antennas to estimate the bearings of M coherent signals, whereas spatial smoothing requires 3M/2 antennas.

OPERATIONAL ISSUES All direction ﬁnding operations proceed on the axiomatic assumption that the bearing measured is conﬁrmed on the signal of interest. The advance of DF technology has produced DF systems with the potential for excellent operational performance. Experience shows this potential may never be realized in practice unless equal consideration is given to explicit conﬁrmation that the AOA reported is on the signal of interest. Traditionally, DF conﬁrmation has been the responsibility of the DF operator, while DF system engineers have concentrated on bearing accuracy, sensitivity, and response time. The growing speed and com-

plexity of communication signals plus the need to control DF operating costs strongly favor automatic, unmanned DF operations where automatic DF conﬁrmation is crucial for success. A review of DF conﬁrmation techniques is beyond the scope of this article. However, DF conﬁrmation processing is fast becoming an integral and indispensible part of DF system engineering. As a result, DF processing choices are increasingly concerned with the acquisition of signal parameter data directly tied to the corresponding DF measurement to conﬁrm that the bearing is associated with the signal of interest.

BIBLIOGRAPHY 1. P. J. D. Gething, Radio Direction Finding and Superresolution, London: Peregrinus, 1991. 2. H. H. Jenkins, Small-Aperture Direction-Finding, Norwood, MA: Artech House, 1991. 3. L. F. McNamara, The Ionosphere: Communications, Surveillance, and Direction Finding, Malabar, FL: Krieger, 1991. 4. D. N. Travers (ed.), Abstracts on Radio Direction Finding, 2nd ed., San Antonio, TX: Southwest Research Institute, 1996. 5. R. E. Franks, Direction-ﬁnding antennas, Antenna Handbook: Theory, Applications, and Design, Y. T. Lo andS. W. Lee (eds.), New York: Van Nostrand Reinhold, 25.4–25.9, 1988. 6. D. N. Travers, Characteristics of electrically small-spaced loop antennas, IEEE Trans. Antennas Propag., 13: 639–641, 1965. 7. J. E. Hipp, Experimental comparisons of sky wave DF algorithms using a small circular array of loop antennas, 4th Int. Conf. HF Radio Syst. Techniques, pp. 215–220, 1988. 8. H. D. Kennedy, W. Wharton, Direction-ﬁnding antennas and systems, Antenna Eng. Handbook, H. Jasik (ed.), New York: McGraw-Hill, 39.16–39.18, 1961. 9. J. E. Hipp,Adaptive Doppler DF system.U.S. Patent No. 5, 321,410, 1994. 10. R. M. Wundt, Wullenweber arrays, Signal Processing Arrays; Proc. 12th Symp. Agard Conf. Proc., Dusseldorf, (16), 128–152, 1966. 11. D. E. N. Davies, Circular arrays, The Handbook of Antenna Design, A. W. Rudge, K. Milne, A. D. Olver andP. Knight (eds.), London: Peregrinus, pp. 999–1003, 1986. 12. W. M. Sherrill, D. N. Travers, P. E. Martin,Phase linear interferometer system and method, U.S. Patent No. 4,387,376, 1983. 13. R. L. Johnson, Q. R. Black, A. G. Sonsteby, HF multipath passive single site radio location, IEEE Trans. Aerosp. Electron. Syst., 30: 462–470, 1994. 14. V. F. Pisarenko, The retrieval of harmonics from a covariance function, Geophysical J. Roy. Astronom. Soc., 33: 347–366, 1973. 15. R. O. Schmidt, Multiple emitter location and signal parameter estimation, IEEE Trans. Antennas Propag., AP-34: 276–280, 1986. 16. H. Krim, M. Viberg, Two decades of array signal processing research, IEEE Signal Process. Magazine, 13 (4): 67–94, 1996. 17. R. L. Johnson, An experimental investigation of three eigen DF techniques, IEEE Trans. Aerosp. Electron. Syst., 28: 852–860, 1992.

Radio Direction Finding 18. R. L. Johnson, Eigenvector matrix partition and radio direction ﬁnding performance, IEEE Trans. Antennas Propag., AP34: 1986. 19. M. Wax, T. Kailath, Detection of signals by information theoretic criteria, IEEE Trans. Acoust., Speech Signal Process., ASSP-33: 387–392, 1985. 20. T. J. Shan, M. Wax, T. Kailath, Spatio temporal spectral analysis by eigenstructure methods, IEEE Trans. Acoust., Speech Signal Process., ASSP-32: 817–827, 1984. 21. B. Friedlander, A. J. Weiss, Direction ﬁnding in the presence of mutual coupling, IEEE Trans. Antennas Propag., 39: 273–284, 1991. 22. A. J. Weiss, B. Friedlander, Array shape calibration using eigenstructure methods, Signal Process., 22: 251–258, 1991. 23. K. Gustafsson, F. McCarthy, A. Paulraj, Mitigation of wing ﬂexure induced errors for airborne direction-ﬁnding applications, IEEE Trans. Signal Process., 44: 296–304, 1996. 24. W. Chen, J. P. Reilly, K. M. Wong, Detection of the number of signals in noise with banded covariance matrices, IEE Proc. Radar, Sonar Navig., 143: 289–294, 1996. 25. Q. Wu, K. M. Wong, Determination of the number of signals in unknown noise environments – PARADE, IEEE Trans. Signal Process., 43: 362–365, 1995. 26. C-W Ma, C-C Teng, Detection of coherent signals using weighted subspace smoothing, IEEE Trans. Antennas Propag., 44: 179–187, 1996. 27. A. A. Shah, D. W. Tufts, Determination of the dimension of a signal subspace from short data records, IEEE Trans. Signal Process., 42: 2531–2535, 1994. 28. Z. Zhu, S. Haykin, X. Huang, Estimating the number of signals using reference noise samples, IEEE Trans. Aerosp. Electron. Syst., 27: 575–579, 1991. 29. D. J. Rabideau, Fast rank adaptive subspace tracking and applications, IEEE Trans. Signal Process., 44: 2229–2244, 1996. 30. C. S. MacInnes, R. J. Vaccaro, Tracking direction-of-arrival with invariant subspace updating, IEEE Trans. Signal Process., 50: 137–150, 1996. 31. H. Zha, Fast algorithms for direction-of-arrival ﬁnding using large ESPRIT arrays, Signal Process., 48: 111–121, 1996. 32. A. Eriksson, P. Stoica, T. Soderstrom, On-line subspace algorithms for tracking moving sources, IEEE Trans. Signal Process., 42: 2319–2329, 1994. 33. P. Li, J. Sun, B. Yu, Two-dimensional spatial-spectrum estimation of coherent signals without spatial smoothing and eigendecomposition, IEE Proc. - Radar, Sonar Navig., 143: 295–299, 1996. 34. A. Delis, G. Papadopoulis, Enhanced forward/backward spatial ﬁltering method for DOA estimation of narrowband coherent sources, IEE Proc. - Radar, Sonar Navig., 143: 1996. 35. W. Du, R. L. Kirlin, Improved spatial smoothing techniques for DOA estimation of coherent signals, IEEE Trans. Signal Process., 39: 1208–1210, 1991. 36. Y. Wu, K. Tam, and F. Li, Determination of number of sources with multiple arrays in correlated noise ﬁelds, IEEE Trans. Signal Process., 50: 1257–1260, 2002. 37. Y. I. Abramovich, N. K. Spencer, and A. Y. Gorokhov, Detectionestimation of more uncorrelated gaussian sources than sensors in nonuniform linear antennaarrays—Part I: fully augmentable arrays, IEEE Trans. Signal Process., 49: 959–971, 2001.

15

38. D. R. Van Rheeden and S. C. Gupta, A temporal smoothing approach to direction of arrival estimation of coherent signals in fading channels, in Proc. WCNC, 286–290, 1999. 39. Z. Gu, and E. Gunawan, A performance analysis of multipath direction ﬁnding with temporal smoothing, IEEE Signal Process. Letters, 10: 200–203, 2003.

RICHARD L. JOHNSON JACKIE E. HIPP WILLIAM M. SHERRILL Southwest Research Institute, 6220 Culebra Road, San Antonio, TX, 78238

RADIO NAVIGATION

RADIO NAVIGATION

Trans p

A key function of navigation is the estimation of current position of a vessel. The reception of radio signals from transmitters whose location is known is a common means of implementing the position estimation functions. Several different schemes have been developed. They can be classified according to the means of determining a position from radio signals. Figure 1 provides the geometric relationships for the different schemes. A theta–theta system determines the position by the vessel’s bearing with respect to two transmitters. This scheme is not common in aviation due to its low accuracy when compared with other available systems. Rho–theta systems use radio signals to determine distance and bearing with respect to the transmitter. This scheme has been in common use in aviation for many years. Most of the airspace can be flown using this basic means of navigation. Errors in bearing measurement will result in position errors that depend on the distance from the station. Rho–rho systems are based on distance measuring equipment (DME) that determines the position of an aircraft using two or more distance values. When only two distance values are available, there is potentially an ambiguity of position. This ambiguity is usually resolved by using the last computed position to determine the most reasonable position. The position accuracy of the rho–rho solution is dependent upon the accuracy of the measured distance and the bearing angles to the stations. If the aircraft is close to the line through two stations, the error in the position solution using only those two stations becomes large. Hyperbolic systems measure the time delay of signals simultaneously transmitted from three or more stations. The

N N

θ1

N

θ ρ

θ2 Theta–theta

Rho–theta

LOP

ρ2

ρ1

ρ3

Rho–rho

Hyperbolic

Figure 1. Geometric relationship of different radio navigation schemes.

Interr

onded

ogatio

pulse

n puls e

123

pair

pair

DME station

Figure 2. DME operation.

locus of all points which have a constant delay of signals from two stations is hyperbolic in shape and is called a line of position (LOP). The intersection of two LOP determine a unique position. The accuracy of each navigation scheme is dependent upon the particular implementation. The following sections describe the characteristics of the different equipment. DME CHARACTERISTICS Distance measuring equipment (DME) is a transponder system combining both airborne and ground equipment to provide distance information. The distance information may be used by other equipment or provided on an indicator to the pilot. In addition, the DME may provide ground speed information. Data from DME is used for both rho–rho and rho– theta navigation systems (see Fig. 2). The major units of the airborne DME equipment are a receiver-transmitter together with an antenna. In addition, a control unit and distance display may be provided in the airborne equipment. The ground-based equipment is a transponder consisting of a receiver-transmitter and an antenna. The DME ground-based facility is usually part of a VOR/ DME, ILS/DME, VORTAC, or TACAN facility. A VOR/DME station is a VHF omnidirectional station combined with the DME. An ILS/DME facility is an instrument landing system with DME. TACAN is a military navigation system providing both azimuth and distance information. A VORTAC is a VOR facility together with TACAN equipment. DME ground stations are capable of handling approximately 100 aircraft simultaneously. If more than 100 aircraft interrogate the ground station, the ground station reduces its sensitivity and replies to the 100 strongest interrogations. Most airborne DME units will operate down to a 50% reply efficiency, so operation is continued even when the ground stations does not respond to all interrogations. The ground station continually transmits a 2700 pulse pair per second squitter signal. At 30 s intervals, the ground station transmits a 1350 pulse pair per second signal that encodes the station identifier in Morse code. When interrogated by an airborne DME pulse pair, the ground station replaces the squitter pulse pair with a pulse pair 50 애s after interrogation. The airborne equipment operates in two modes, search and track. In the search mode, the airborne equipment transmits 90 or more pulse pairs per second. The transmission rate is randomly shifted to prevent possible confusion effects due to another airborne DME transmitter. After each pulse pair is transmitted, the receiver equipment waits for a reply pulse pair that arrives with a consistent delay after the transmis-

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

124

RADIO NAVIGATION

sion. If such a reply is found, the airborne DME switches to the track mode and tracks the regular delays of the replies. In the track mode the airborne equipment reduces the interrogation rate to 20 or less pulse pairs per second. The reduction in the interrogation rate allows more airborne units to be serviced by the ground based DME facility. The airborne equipment converts the time duration between the transmission of the interrogation pulse pair and the reception of the reply pulse pair into distance. Because the speed of the radio signal is a known constant, the distance from the DME facility can be determined. To be more precise,

DME Channels 1 to 16 17 to 56 Even channels with ILS Odd channels with VOR 60 to 69 57 to 59, 70 to 126

Assignment

VHF Frequencies

Unpaired

134.4 MHz to 135.9 MHz

ILS and VOR

108.00 MHz to 112.00 MHz 133.3 MHz to 134.2 MHz 112.00 MHz to 117.9 MHz

Unpaired VOR

Distance in nm = (Time duration in microseconds − 50 ms)/12.359 ms/nm The airborne DME unit has memory that handles the situation when the reception of the DME reply is momentarily interrupted. The equipment uses the memory to remain in the track mode and provide distance data during the reply interruption. The memory allows the DME to provide distance information for up to 10 s after loss of reception. It is common for airborne DME units to handle up to five DME ground facilities simultaneously by multiplexing the receiver-transmitter circuits. This allows the equipment to simultaneously provide distance information for up to five DME navaids. The airborne DME transmits and receives on one of 252 channels. There are 126 X and 126 Y channels. The transmit and receive frequencies of any one channel are separated by 63 MHz. In the first 63 X channels, the ground-to-air frequency is 63 MHz below the air-to-ground frequency. For X channels 64 through 126, the ground-to-air frequency is 63 MHz above the air-to-ground frequency. For Y channels the situation is reversed. The ground-to-air frequency of the first 63 Y channels is 63 MHz above the air-to-ground frequency. Channels 64Y to 126Y, the ground-to-air frequency is 63 MHz below the air-to-ground frequency. The 252 ground-to-air frequencies are each whole MHz frequencies from 962 MHz to 1213 MHz. The air-to-ground frequencies are each whole MHz frequencies from 1025 MHz to 1150 MHz. The duration between each pulse of the pulse pair transmitted by the airborne equipment and that of the ground equipment is different for X and Y channels. The table below shows the pulse spacing.

X-channel Y-channel

Air-to-ground

Ground-to-air

12 애s 36 애s

12 애s 30 애s

DME operation requires that the aircraft and the ground facility be in a direct line-of-sight connection. Terrain and the curvature of the earth limit the range. The formula below provides an approximate limit on the range of DME equipment due to earth curvature with an assumption that the ground station antenna is about 16 ft. above the surface. √ DME range limit (nm) = 1.23 aircraft altitude (ft) + 4 The DME provides the slant range distance from the aircraft to the DME navaid facility. If the DME range is large com-

VHF RF Carrier 108–112 MHz even-tenths MHz or 112–117.95 MHz

9960 Hz subcarrier

30 Hz reference

9960 Hz frequency modulated at 30 Hz (reference)

VHF carrier and 9960 Hz subcarrier (reference)

Variable 30 Hz AM

Most DME channels are paired with a VHF frequency allocated to VOR or ILS. That is, for each VOR or ILS frequency there is an assigned DME channel for use when DME equipment is part of the navaid facility. The X channels are paired with VHF frequencies in 100 kHz increments (108.00, 108.10, 108.20, etc.). The Y channels are paired with VHF frequencies in 100 kHz increments but offset by 50 kHz (108.05, 108.15, 108.25, etc.). The table below shows the DME channel pairing with VHF frequencies.

VHF carrier and 30 Hz space modulation

Figure 3. Components of VOR signal.

RADIO NAVIGATION 0˚

180˚

125

360˚

Reference symbol

10440 Hz 9960 Hz 9480 Hz

Variable signal north

0˚ Magnetic north 0˚

180˚

360˚

0˚

Reference symbol

180˚

360˚

Reference symbol VOR station

270˚

90˚

Variable signal west

Variable signal east 180˚

0˚

180˚

360˚

Reference symbol

Variable signal south

pared with the aircraft altitude the slant range is essentially the same as the ground distance. Navigation equipment that uses the DME range data can remove the effect of the aircraft altitude by incorporating in the calculation both the measured aircraft altitude and the elevation of the navaid as retrieved from the navigation database. The accuracy of DME range measurement is dependent upon the range, environmental conditions, and the equipment being used. A nominal 95% accuracy is about 0.1 nm at shorter ranges. For longer ranges the accuracy degrades due to atmospheric conditions and lower signal-to-noise ratio. VOR EQUIPMENT CHARACTERISTICS VHF Omnidirectional Range (VOR) equipment consists of a ground station (transmitter) and an airborne receiver. The VOR ground station continuously transmits a signal that may be used by all aircraft within reception range of the signal. Using the VOR signal, the receiver determines the bearing from the ground station to the receiver. VOR stations are of-

Figure 4. Phase relationship of VOR signal.

ten colocated with other navigational aids such as DME and TACAN stations. The frequency of VOR stations ranges from 108.00 MHz to 117.95 MHz. There are two types of VOR transmitters, Doppler VOR and conventional VOR. Doppler VOR has limited usage. The details of conventional VOR follow. The transmitted signal from the VOR station consists of a VHF carrier and a 9960 Hz subcarrier. The VHF carrier is amplitude modulated by a variable 30 Hz signal whose phase is dependent upon the bearing with respect to the ground station. The 9960 subcarrier is frequency modulated by a 30 Hz reference signal. The subcarrier is modulated between 10440 Hz and 9480 Hz (see Fig. 3). Figure 4 illustrates the phase-bearing relationship of the two components of the VOR signal. At 0⬚ bearing, the phase of the variable signal is the same as the phase of the reference signal. At 90⬚ bearing from the ground station, the variable signal is 90⬚ out of phase with respect to the reference signal. The phase difference between the two signals is proportional to the bearing from the ground station to the receiver.

126

RADIO NAVIGATION

The phase difference between the variable signal and the reference signal is used by the receiver to determine the bearing from the ground station to the receiver. Essentially, the receiver separates the subcarrier from the VHF carrier, detects the phase of the 30 Hz signal in each and then determines the phase relationship. In a conical area above the VOR station, the phase difference between the two signals cannot be detected reliably. This area is called the VOR cone of confusion. Receivers have monitors that detect this condition and provide alerts to the pilot that the signal is unreliable. In general, the VOR ground station antenna is physically aligned so that the VOR signal indicates 0⬚ bearing when the receiver is magnetically north of the ground station. That is, in general, the VOR bearing from the ground station to the receiver is the same as the magnetic bearing to the receiver. However, the magnetic field of the earth is constantly changing. In Europe and the United States there are regions where the magnetic north is changing 1⬚ every seven years. Therefore even if the VOR signal was aligned with magnetic north at one instant in time, after some period of time there can be a significant difference. When the difference becomes too large (2⬚ or so), the VOR antenna is usually realigned. The accuracy of VOR signals is degraded by range and terrain. Generally, the VOR signal error is less than 1⬚. As the range from the VOR station is increased, the VOR signal bearing often oscillates around the true bearing. This effect is known as scalloping. There are several types of VOR indicators that are in common use. Most of the indicators provide bearing as a rotating arrow pointing to the bearing angle on an azimuth card. The azimuth card is slaved to the heading sensor so that the current heading is at the top directly under the lubber line. The indicators often include a distance indication that is connected to the DME. ILS CHARACTERISTICS An instrument landing system (ILS) consists of ground-based transmitters and airborne equipment that provides lateral, along-track, and vertical guidance. The lateral signal is provided by a localizer transmitter and the vertical signal is provided by the glideslope transmitter. The airborne ILS receiver is capable of receiving and processing both the localizer and the glideslope signals. The along-track information is provided by marker beacons (transmitters located along the descent path that provide a narrow vertical radio signal) or distance measuring equipment (DME). The marker beacon receiver can be part of the ILS receiver or it can be a separate receiver (see Fig. 5). The localizer beam is almost always aligned to guide the aircraft directly over the runway threshold. In certain situa-

DME

Bia

s

Runway DME

Bias

Figure 6. Biased DME antenna situations.

tions, only a localizer signal is provided and no electronic glideslope signal is provided. In some localizer-only situations, the localizer signal is not aligned to the runway but instead provides guidance to some location from which the pilot has other means to complete the landing. In some cases, the ILS DME transponder delay is adjusted so that the sensed DME distance is zero at the runway threshold instead of at the DME antenna. Such DMEs are known as biased DMEs and the bias is indicated on the approach chart. There are two general arrangements of biased DMEs (see Fig. 6). In one case the DME antenna is located at the glideslope antenna and adjusted to read zero at the corresponding runway threshold. In the other case, the DME antenna is located midway between the two runway thresholds at opposite ends of the runway. In this case the single DME will support approaches from either direction and will read zero at both runway thresholds. The localizer signal is transmitted on assigned frequencies between 108.1 MHz to 111.95 MHz. As shown in Fig. 5, the localizer antenna is usually located past the far end of the runway very near the extended runway centerline.The signal pattern are two main lobes on each side of the center line. The left lobe is predominantly modulated at 90 Hz and the right lobe is predominately modulated at 150 Hz. Along the center line, the two modulated signals are equal (see Fig. 7). The localizer signal also extends backwards and is called the back course. This signal can be used for guidance but the modulation convention is reversed. There is no glideslope signal provided on the backcourse region. When flying the backcourse, either the equipment must reverse the localizer indications or the pilot must recognize and fly the reversed indications. The glideslope signal is transmitted on assigned frequencies between 328.6 MHz and 335.4 MHz. The glideslope antenna is located on the side of the approach end of the runway. The glideslope signal consists of two main lobes on each

Antenna beam pattern

Glideslope and DME antenna Localizer antenna

Runway

Zero DME distance indication

Runway

Localizer beam centerline

Figure 5. Common layout of ILS facility.

Back course

150 Hz 90 Hz Figure 7. Localizer antenna lobe patterns.

Localizer course

RADIO NAVIGATION

Glideslope antenna

Glideslope centerline

90 Hz 150 Hz

Runway

Figure 8. Glideslope antenna lobe patterns.

side of the desired glideslope path. The glideslope descent angle is usually three degrees. To provide obstacle clearance or to reduce noise, steeper glideslope paths are used. The upper glideslope lobe is predominantly modulated at 90 Hz and the lower lobe is predominantly modulated at 150 Hz (see Fig. 8). Marker beacons signals are transmitted at 75 MHz and are modulated at 400 Hz, 1300 Hz, or 3000 Hz. The transmitters are located along the descent path of to the runway. Figure 9 shows the general arrangement of the beacons. When the aircraft passes over the beacons, the marker beacon receiver detects the signal and provides an indication to the pilot of the passage. The exact location of the marker beacons is given on the approach procedure chart. Inner marker beacons are installed at runways with Category II and Category III operations. In typical operation, the pilot maneuvers the aircraft to cross the localizer signal centerline. At this time, the localizer receiver provides an indication that the localizer signal is being received and provides a lateral deviation indication showing the aircraft displacement from the centerline. Using the lateral deviation indication, the pilot steers to the localizer centerline until the glideslope receiver indicates reception of the glideslope signal. At that time the pilot has both lateral and vertical indications to guide the aircraft on the desired glideslope path. The marker beacons or DME indications provide along-track indications of the progress of the descent. ADF NAVIGATION Automatic direction finder (ADF) is the oldest and most widely used radio navigation system. The ADF system consists of a ground-based transmitter and an airborne receiver. The ADF system provides an indication of the bearing of the station from the aircraft centerline. The ADF receiver is capable of receiving AM signals from 190 kHz to 1750 kHz. The transmitter can be either commercial AM broadcast stations or nondirectional beacons (NDB) that are installed expressly for radio navigation.

400 Hz 1300 Hz

1300 Hz Glid

pe eslo path

Runway

Inner marker

Middle marker

Outer marker

Figure 9. Marker beacon locations.

127

An omnidirectional antenna is used to receive the signal to aid in tuning the receiver. The ADF receiver determined the bearing of the station using the directional sensitivity of loop antenna. The loop antenna may be physically rotated to determine the bearing to the signal or the bearing may be determined using electronic sensing of the signal strength from more than one loop. LONG-RANGE NAVIGATION (LORAN) Low frequency long range navigation (LORAN) was first developed for military applications during World War II. Since then it has evolved into today’s LORAN-C system that is also used for civil applications. Marine applications were the first to appear, followed by aviation applications. The LORAN system consists of a group or chain of transmitting stations and a receiver. Within each chain there is one master station and several slave stations identified by a single letter. Associated with each chain is a unique group repetition interval (GRI) that identifies the chain. A single chain provides navigation coverage for several hundred miles from the master station. The frequencies of transmitted signal are between 90 kHz and 110 kHz. The pulse patterns of the transmitters are different allowing the receiver to separate the received signals. Within each chain, each station simultaneously transmits a pulse at the specified GRI. Each station has an assigned pulse pattern that allows the receiver to distinguish between the different signals. By determining the time differences between the set of received pulses, the receiver can determine the difference in distance from each of the transmitting stations. If three or more signals are received, the receiver can determine the best estimate of the location of the receiver. The accuracy of the LORAN position estimate depends upon the position of the receiver with respect to the location of stations providing signals. The nominal accuracy of the LORAN system is 0.25 nm when within the groundwave range. To improve accuracy, LORAN receivers compensate for the differences in velocity of the ground wave when the signal is propagating over land rather than water. GLOBAL POSITIONING SYSTEM (GPS) The space segment of the global positioning system (GPS) consists of a set of orbiting satellite transmitters that provide two L-Band, 1575.42 MHz (L1) and 1227.6 MHz (L2) signals. The two frequencies allow the appropriately equipped user to correct for errors due to ionospheric refraction. Civil receivers use only the L1 signal. The basic satellite configuration is a set of 24 satellites in 6 orbital planes. The signals provided by the satellites are modulated with two pseudorandom noise (PRN) codes: a coarse/acquisition (C/ A) signal and a precise (P) signal. The P code component of the signal allows higher precision ranging and information necessary to decode the signal which has restricted distribution. The airborne GPS receiver receives the L1 signal of all satellites in view and by the use of correlation techniques can detect the unique C/A code for each satellite. The C/A code has a chipping rate of 1.023 MHz and a length of 1023 bits so it repeats every millisecond. By use of signals from four or more satellites, the receiver can determine the time reference

128

RADIO NOISE

and the range to each satellite and hence estimate the receiver position. To deny the accuracy of GPS to unfriendly forces, the satellite signals are intentionally degraded using a concept known as selective availability (SA). This technique degrades to L1 signal characteristics to the extent that navigation accuracy is about 100 m (95%). MICROWAVE LANDING SYSTEM (MLS) Microwave landing systems consist of an azimuth and elevation microwave transmitters, a conventional DME transponder, and the airborne receivers. The azimuth transmitter provides coverage for 40⬚ to each side of the centerline. The elevation transmitter provides coverage up to 15⬚ of elevation. Microwave landing transmitters operate on one of 200 assigned frequencies between 5.031 GHz and 5.1907 GHz. The azimuth transmitter provides a narrow beam signal that sweeps the azimuth coverage area (⫾40⬚) at a rapid rate. By detecting the timing between the reception of the microwave signal, the receiver can determine the azimuth angle from the centerline. A preamble microwave signal is transmitted from a broad beam antenna to indicate the beginning of the azimuth sweep. Various information is digitally encoded in the preamble signal. The elevation function is provided in the same manner as the azimuth function. High sweep rates provide about 40 samples per second for azimuth and elevation. GERALD E. BENDIXEN Rockwell Collins, Inc.

730

SEARCH RADAR

SEARCH RADAR Search radar is used widely to provide electronic surveillance of the environment to detect objects that would otherwise be invisible to the unaided observer. These systems usually function without operator interaction to provide information rates commensurate with high-speed decision making by the user. That user might, for example, be an automated weapon launch system, an air traffic controller, or a traffic policeman. OVERVIEW This section provides a summary of the function, applications, elements, and design challenges of modern search radar. Function The single function of the radar is to provide volume surveillance over all or portions of a sphere centered at the radar antenna. This is accomplished by radiating high-energy microwave pulses into the volume and detecting these pulses as they are reflected from objects of interest (targets). An antenna focuses the radiation to create a narrow beam of energy. This selectivity and the short duration of the pulses allow the radar to measure target location in distance (range) and in one or two angular dimensions (bearing and elevation). The antenna may be rotated mechanically in angle or the beam may be steered electronically in one or two dimensions. A receiver amplifies the target-reflected pulses and removes the microwave carrier frequency through a heterodyning process. The received pulses are applied to a signal processor for target detection and location measurement. In modern designs, the signals are converted to digital format with subsequent processing carried out digitally. Target detection is performed by comparing the magnitudes of received signals to a preset threshold. Signals exceeding this threshold are declared targets and their parameters are passed on to a location measurement process. Occasionally, internal receiver noise will exceed threshold. These occurrences are termed false alarms. Adaptive thresholds are employed to maintain a constant average false alarm rate (CFAR). Target range is determined by noting the time of detection relative to the time of transmission. The translation from time to range is predicated on the fact that the pulses travel at the speed of light. Angular measurement is obtained by noting the position of the antenna at the time of detection. In simpler applications, target information is applied to an electronic display. This display provides the operator with a picture of the environment that is updated on each antenna scan. Usually, a plan view is presented in x and y coordinates. Automatic designs rely on a general-purpose digital computer to interpret target detection information. This computer provides scan-to-scan correlation, trajectory extrapolation, and, in military applications, threat assessment. In applications supporting weapon engagement, a kill assessment function may be provided. J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

SEARCH RADAR

In earth-based applications, there are a number of extraneous signals that may cause interference. Among these are unwanted clutter returns from the local terrain, sea surface, or weather. In military applications, an adversary may radiate noise in an attempt to hide returns from their own vehicles. These interference signals are referred to as electronic countermeasures (ECM) or jamming. More sophisticated techniques are in use that attempt to confuse the radar by generating false targets. Finally, a large number of radar systems are in use today. Each has its own peculiar function and set of parameters, but there may be several systems colocated in the same general area. Direct reception of radiation from another radar may be interpreted as a target return. A function of the well-designed radar is minimization of clutter, ECM, and friendly interference effects. Applications There are a number of ways to categorize search radars. They may be developed for the military, be produced commercially, or be sponsored by governmental agencies. The platforms may be surface-based at sea or on land, airborne in aircraft or missiles, or space-based. A radar is either two-dimensional (range and bearing) or three-dimensional (range, bearing, and elevation). It also may be characterized, loosely, as long, medium, or short range. Finally, microwave transmission frequency is an important attribute. Radars are in use having carrier frequencies from 1 GHz, or lower, to 20 GHz, or higher. Examples of search radar applications include general surveillance, weapon support, aircraft and ship navigation, air traffic control, harbor traffic control, early warning of attack, weather alerting and monitoring, obstruction alerting, vehicular speed measurement, and satellite location monitoring. Elements In general, a search radar is composed of several distinct electronic elements having specific functions. These include a frequency synthesizer, a timing generator, a power amplifier, an antenna, an antenna platform, a microwave receiver, an intermediate frequency receiver, a signal processor, and a control computer. An operator display-and-control panel may be included. Figure 1 is a block diagram showing the interactions among these elements. The frequency synthesizer provides the basic radar signals necessary for carrier transmission and for local oscillator production. Each frequency conversion used in the radar receivers requires one local oscillator signal. An additional signal may be provided to serve as the basic system clock, which sets the timing intervals for pulse transmission, range gating, and digital sampling. Coherency is an absolute requirement in modern designs due to the need to reject clutter by Doppler processing. Therefore, all signals must be phase related. Often, coherency is controlled by starting with crystal-based oscillators operating at low frequency. Microwave signals are produced using frequency multiplier chains, offset modulators, and frequency dividers. A timing generator provides timing commands for the other radar elements. These commands include transmission pulse width and repetition frequency, range gate width, and digital converter sampling intervals. This element also provides start-and-stop events for the various algorithms involved in

731

signal processing. In designs using phase-steered antennas, the timing generator provides beam-steering commands. High-power microwave transmission is accomplished using a power amplifier. This element amplifies a low-level continuous-wave input from the synthesizer into a high-power pulsed signal for subsequent application to the radiating antenna. Peak power output may range from a few watts to several megawatts and pulse widths from tens of nanoseconds to several microseconds. In long- and medium-range applications, the amplifier will be a vacuum tube. Popular choices for this device are the klystron and the traveling wave tube. The latter is a broadband device suitable for frequency-agile operation. Solid-state devices are available for short-range designs where power levels below approximately 100 W are acceptable. Pulsed operation is provided by a modulator that controls the tube beam current in vacuum tube designs. The antenna functions to focus microwave energy into a narrow spatial beam. In the simpler designs, this device is a parabolic reflector having a feed horn mounted near its focal point. In Cassegrain designs, the feed is located behind the reflector and illuminates a subreflector mounted near the focal point. A more advanced approach uses a slotted waveguide array. Each slot is a sub-radiator and a beam is formed owing to the phase relationship among the slots. In the ultimate design, phase shifters are added to groups of slots to enable electronic beam steering. Advanced designs may add transmit-and-receive modules to slot groups. This approach obviates the necessity for a separate power amplifier. An antenna platform provides mechanical beam rotation and beam stabilization relative to the horizon. Land-based applications require only beam rotation. However, in shipbased or airborne designs, the three-dimensional motion of the host vehicle may dictate gimballing in one or two dimensions in order to maintain a constant geometrical reference. Beam stabilization is afforded using feedback control based on attitude sensors. The primary functions of the microwave receiver include receiver protection from the high-level transmission, frequency selectivity to minimize external interference, and at least one frequency conversion. Some designs also include radar frequency amplification to minimize the loss associated with frequency conversion and to establish a basic system noise figure. Usually, receiver protection is provided by a circulator device that depends on pathlength differentials to cancel the transmission while providing an unattenuated path for target reception. Passive or active limiters are also in use. In modern designs, the receiver is solid state and based on microwave integrated circuit (MIC) technology. The intermediate frequency (IF) receiver provides further amplification and frequency conversion of the target returns. It also applies frequency selectivity to pass the pulse spectrum while rejecting unwanted interference. Usually, the final conversion is to video where only a pulsed, Dopplershifted carrier remains. At that stage, the receiver implements an approximation to a matched filter that maximizes the signal-to-noise ratio. In modern designs, the last element is an analog-to-digital converter that outputs digitized words representing the voltages of incoming signals and noise. Word lengths may vary from eight bits in simpler designs to as many as sixteen bits in advanced applications. The receiver may employ an IF limiter to prevent large clutter returns from saturating the digital converter.

732

SEARCH RADAR

Antenna platform

Display

Microwave receiver

Antenna

Power amplifier

Intermediate frequency receiver

Frequency synthesizer

Signal processor

Control computer

Timing generator

Control panel

Figure 1. Generalized block diagram of a search radar showing interfaces between elements. Most modern search radars employ the generic elements shown in this diagram.

All decision-making algorithms such as noise estimation, target detection, clutter rejection, and parameter measurement are assigned to the signal processor. In modern designs, this element takes the form of a real-time, special-purpose digital computer. It is always solid state and may employ very large-scale integrated (VLSI) circuits to provide ultra-high-speed operation in a small volume. Often, algorithms are hard-wired to perform specific operations but, in some cases, more general purpose, programmable microprocessors are used. In automatic systems, all basic radar functions are controlled by a general-purpose digital computer. This control computer provides radar frequency, pulse width, and repetition frequency commands to the radar transmitter elements, beam steering and rotation rate commands to the antenna elements, and parameter boundaries to the receiver and signal processing elements. This element also receives direct outputs from the signal processor, which it uses to refine target correlation, location measurement, trajectory prediction, threat assessment, and, possibly, kill assessment. In modern designs, this computer is programmable in some higher order language (HOL) such as ADA to provide design flexibility and ease of modification. Systems dependent upon operator control use a computer of lesser complexity. These systems employ a control panel from which the operator enters basic commands to the radar. These may include radar frequency selection, pulse width, repetition frequency, beam control, scan rate, nonradiate sector cutouts, range boundaries, minimum velocity limits, and detection threshold. Radar outputs are sent to a display. Detected targets are shown as dots of various intensities located proportionally to target range and bearing. Codes may be attached to designate target type and/or priority. Separate digital readouts may be provided giving more accurate target location data and other pertinent information. Design Challenges The designers of a modern search radar system face a number of challenges that must be met to ensure specification compliance and cost effectiveness. Modern radar is dependent on the phenomenon of Doppler shift to provide a distinction between near-zero velocity clut-

ter and higher velocity targets. Advantage of this phenomenon can be taken only if the transmission and local oscillator signals are coherent. That is, their frequency and phase must remain nearly constant over time. Therefore, close attention must be given to short-term phase and frequency stability. In addition, the high population density of radar in operation forces a tight spectral allocation to each system. Legally, the designer cannot allow his frequency to drift outside the prescribed boundaries. This problem may be resolved by utilizing oscillators based on crystalline structures having a precise molecular structure that supports only a narrow vibration spectrum. Power amplifiers exhibiting low phase noise characteristics may also be required. Very-short-term instability is referred to as frequency modulation (FM) noise. This noise is usually broadband and is radiated in conjunction with the transmission carrier. Clutter-reflected returns of this noise can cause interference with the relatively low-strength returns from targets even though targets and clutter are separated by a Doppler shift. In typical applications, this noise must be maintained as low as 120 dB below the carrier level when measured in a bandwidth of 1 Hz. Control of this noise may be achieved using automatic phase and frequency control loops within the frequency synthesizer. A coherent radar is dependent upon the stability of the pulse repetition frequency. Deviations in this frequency, termed pulse jitter, can cause intermodulation products or clutter to spill over into the target spectrum and cause interference. Careful attention to the stability of the system clock can minimize this problem. Target detectability is a direct function of transmitted power level. Once a power amplifier device, modulator, and power supply have been selected, this parameter is set. However, care must be taken that its value does not deteriorate over time. In longer range applications, the transmission power is high enough to destroy receiver circuitry if allowed to enter. It is mandatory that provisions be made to prevent this occurrence. Common practice is to use high power circulator devices whose phase characteristics cause transmission cancellation at the receiver input. In reality, these devices provide

SEARCH RADAR

only approximately 20 dB of attenuation. If more isolation is required, active transmit–receive circuitry may be required. These devices are switches that turn off during pulse transmission and turn on afterward to receive signals. High-power passive limiters are also in use. These do not require external timing control. In any case, it is ordinarily not possible to process target returns during the time of pulse transmission. This must be accounted for in terms of the blind range effect on minimum detectable target range, periodic blind range zones, and range ambiguity resolution. Target detectability, apart from clutter effects, is a direct function of antenna gain. Gain is the ratio of the power density achieved on boresight to that given transmission isotropically into a sphere. Gain is directly proportional to antenna physical area. The design challenge is to find sufficient space to install a large physical item. Long range search antennas may have dimensions measured in meters. Where space limitations exist, trade-offs must be made among transmission power, gain, and receiver sensitivity to achieve an optimum design. Radar antennas exhibit spatial sidelobes. These lobes allow for leakage into every portion of the radiated sphere and allow for low-level reception from all objects within the sphere. Except for possible clutter degradation and large-target angular measurement, the real concerns are off-axis jamming and friendly interference. Some sidelobe control is always advisable. This is achieved by shaping the illumination current distribution across the antenna face. A uniformly illuminated aperture exhibits a first sidelobe only 13 dB below the main beam gain. Through careful design, this may be increased to 30 dB or more. The penalty for sidelobe control is decreased gain and increased beamwidth. Military, groundbased radars also commonly employ sidelobe cancellers, which are circuits capable of attenuating interference signals from discrete sources. Antenna beam scan rate is a parameter subject to optimization and trade-off when mechanically scanned antennas are employed. High scan rates yield short target beam dwell times that deprive the radar of available energy and reduce signal-to-noise ratio (SNR). Low scan rates increase the time interval between initial detection and verification detection. Thus, longer detection ranges may be required. Increased range also reduces SNR. The problem of scan rate optimization is compounded when phase-steered antennas are used. In these systems, the beam may be programmed to scan back to the angle of initial detection immediately for a second detection verification. Thermally induced receiver noise exists in all radars. Reliable detection and parameter measurement are possible only when target signal levels exceed noise by at least a factor of ten. The careful design will include only those electronic devices having a low noise figure. Noise figure is defined as the ratio of input SNR to output SNR and is a direct measure of degradation due to noise. SNR is also reduced by losses inherent in wave guide runs, coaxial cable runs, connectors, and discrete microwave components. There will also be a loss when the angle of the antenna beam center does not coincide with the actual target angle. Signal processing can also contribute to the loss structure. These include analog-to-digital converter quantization, imperfect thresholding, range gate straddling, and Doppler filter straddling. All of these losses must be kept to a minimum.

733

Dynamic range is a measure of the radar’s ability to process large signal returns simultaneously with very low signal returns without degradation. Generally, clutter returns are the largest signals present at the radar input. If a clutter echo arrives in time conjunction with a target echo and if the clutter level causes receiver saturation, then small signal suppression occurs and target detectability is degraded. Similarly, clutter may exceed the dynamic range of the analog-todigital converter. In this case, totally erroneous data may be produced. The careful designer will choose analog components having a high saturation level and will choose digital converters having a large number of bits in their output word length. Typically, word lengths of 14 to 16 bits are required to withstand worst-case clutter inputs. Even when clutter returns are processed linearly through the radar receiver and digital converter, they must be rejected within the signal processor to avoid interference with target returns. Classical analog filtering is not effective against clutter, because the receiver must pass the entire pulse spectrum of the target, typically several megahertz, while the Doppler offset is only on the order of a few kilohertz. Common practice is to employ digital clutter cancelers also known as moving target detectors (MTD). In its simplest form, the MTD algorithm provides subtraction of the current pulse sample from that received one or more repetition intervals earlier. Because clutter has a near-zero Doppler shift, it is canceled almost totally, whereas the Doppler shifted target return is passed with little attenuation. The result is a high-pass digital filter. Feed-forward and feedback multiplication factors may be used to tailor the frequency response to specific target requirements. Clutter may be received from the terrain, from the surface of the sea and from rainfall. Terrain reflections are usually the most intense, but, because their Doppler shift is very low, they are canceled easily. Sea clutter becomes a problem only under heavy-sea conditions. Rain clutter presents the greatest challenge, because it can be highly intense and that intensity increases with the fourth power of radar frequency. This problem is burdened by the Doppler shifts from wind-driven rain drops. Because wind velocity generally increases with altitude, the cancellation of higher altitude rainfall clutter becomes problematic. Microwave transmissions tend to resonate with molecules present in the atmosphere. This resonance causes attenuation with an attendant degradation to target detectability. Oxygen and water vapor are the primary contributors to this attenuation, although smoke, haze, and smog can also be factors. Even though there exist windows of decreased attenuation, the effect is more pronounced at higher radar frequencies. Rainfall also can cause severe attenuation. Over frequency, attenuation rates vary from a few tenths of a decibel per kilometer of transmission path to more than 1 dB. Microwave components contribute some degree of loss with attendant transmission strength reduction and target signal attenuation. Included are waveguide runs, waveguide joints, circulators, phase shifters, limiters, and filters. Without careful design, the summation of these losses can exceed 10 dB. There are a number of variable losses associated with the antenna beam and signal processing. A target may appear at an angle off the antenna elevation boresight. During the target dwell, the antenna gain will vary. Pulse returns may be sampled at other than peak response times. Noise level estimates may not be exact. The matched filter approximation

734

SEARCH RADAR

used will not yield the theoretical maximum SNR. These effects are always probabilistic but must be considered in detection calculations. In military applications, ECM must be considered. This is employed in an attempt to hide the echo from incoming aircraft or missiles or to decoy and confuse the search radar. The simplest form is barrage noise jamming in which high noise levels are radiated across the entire frequency band allocated to the radar. These systems require no knowledge of the exact transmission frequency of the victim radar. In those cases where the adversary is able to measure the transmission frequency, spot jamming may be effective. Here, the noise jamming is restricted to the instantaneous spectrum occupied by the radar transmission and may have much higher power density than that of barrage jamming. More advanced techniques are available in which the ECM system reradiates the transmission with variable delay or modulation in an attempt to create false targets. The effectiveness of ECM depends upon whether or not the jamming source can be located in angular coincidence with protected targets. If not, its effect may be minimized by careful control of antenna sidelobes. Worldwide, there exist a large number of radars of various types operating in a variety of frequency bands. It is likely that a given radar will be required to operate in an environment containing several other radars operating in proximity. In this case, reception of direct transmission and target reflected echoes can cause interference and false target production. When the various radars are of the same type, based on a common design, the designer may minimize interference using careful frequency channelization and repetition frequency selection. When disparate types are involved, techniques exist for editing out pulse returns at repetition frequencies other than that currently used by the victim radar. In all cases, careful antenna sidelobe control is mandatory. An emerging technology has been deployed that attempts to render aircraft, missiles, and ships invisible to radar. This stealth technology utilizes radar absorbent material (RAM) and geometric vehicle design to greatly reduce radar reflectivity. In fact, stealth techniques do not result in total invisibility but radar return levels from these targets are reduced appreciably. Unfortunately, from a design standpoint, little can be done to counter these threats except to increase transmitter power, frequency stability, antenna aperture, and/or receiver sensitivity. However, the stealth technology is not totally effective over an infinite bandwidth. Therefore, it may turn out that the next generation radar designs must be based on either very low or very high radar frequencies. These decisions will have far-reaching implications in terms of antenna beamwidth, Doppler resolution, and repetition frequency selection. Stealth is, probably, the most serious concern for future military radar development. Target detection decisions are based on signals exceeding some preset threshold. Ideally, that threshold is computed by multiplying an average measurement of the ambient receiver noise level by some constant. Thus, the threshold will vary as the noise level varies due to temperature changes and local clutter conditions. Since false alarms on noise are probabilistic events, this method ensures a constant average false alarm rate (CFAR). The problem is in determining the ambient noise level. This may be accomplished by averaging a number of samples taken from range cells next to the target cell under investigation. Use of a larger number of cells re-

duces the measurement error but introduces other errors due to the range dependency of clutter. Once an acceptable false alarm rate is established, the probability of target detection as a function of SNR is computed easily. However, the detection decision need not be based solely on the result of one received pulse. Rather, it should be based on the preponderance of evidence from a group of pulses representing an antenna beam dwell. This evaluation is termed pulse integration. It can yield a significant increase in effective SNR. The accuracy of range measurement is a function of transmitted pulse width. However, when peak power is fixed, short pulses result in lower energy and reduced SNR. Often, pulse compression techniques are employed in which phase coding is applied to subpulses within a longer pulse on transmission. Decoding on reception results in an output pulse width equal to the subpulse but with energy increased by the compression factor. This technique is especially useful in longer range applications where low repetition frequencies are required. Further accuracy may be obtained using split-range gating. In this approach, magnitudes of adjacent range samples are used to form a weighted average proportional to true range. The accuracy of this method improves as SNR increases. Bearing measurement accuracy is dictated by the azimuth beamwidth of the antenna pattern. In the simplest case, target bearing is assumed equal to the physical angle of boresight at the time of detection. Improved accuracy may be obtained by noting the beginning start and stop angles at the end of above threshold intervals. More advanced designs may use weighted averages of return samples and curve fitting to known antenna beam parameters to increase accuracy. Measurement errors decrease as SNR increases. In elevation, measurement accuracy is a sole function of elevation beamwidth. In electronically beam-steered applications, where the beam transitions the target vertically as well as horizontally, beam splitting can be used to improve accuracy. Monopulse techniques, commonly used in tracking radars, may also be used in search radars. In these systems, sum and difference beams are formed and processed in separate receiver channels. The voltage ratio between these channels is proportional to target angle in one or two planes. Except in applications using very low repetition frequencies, initial range measurement is always ambiguous. That is, the true target time delay is equal to the time delay between transmission and detection plus an integral number of repetition intervals. This ambiguity may be resolved by obtaining detections on two or more distinct repetition frequencies. Final radar computer outputs may include target trajectory prediction, threat assessment, and kill verification. All of these functions are dependent upon radar measurement accuracy. The number and types of algorithms possible are limitless. Development of these algorithms will occupy a large portion of the designer’s efforts. DESIGN CONSIDERATIONS This section addresses the methodology and mathematics currently used in the development of search radars. Radar Range Equation The design of any search radar begins with the radar range equation. This equation ties together all major radar parame-

SEARCH RADAR

ters into a single relation that enables the computation of SNR. The latter is the single parameter that dictates radar performance both in detectability and in measurement accuracy. Because matched filter theory shows that the maximum SNR achievable is equal to the pulse energy-to-noise power density ratio (ENR), it is usual to begin with the following equation. ENR =

Pt δt Gt σt Ar Np L (4π )2 R4 ηF

where Pt is peak transmitted power, 웃t is pulse width, Gt is transmit antenna gain, ␴t is target cross section, Ar is antenna capture area, Np is the number of pulses available during the target dwell, L is the system losses, R is target range, ␩ is receiver noise power density, and F is receiver noise figure. Skolnik (1) gives a derivation of the basic equation. The basic equation may be modified to reflect more germane parameters by substituting various relations among the parameters. The number of pulses available is given by Np =

Td Tr

where Td is the dwell time and Tr is the repetition interval. The transmit antenna gain may be expressed as Gt = 4π/θa θe where ␪a is the azimuth beamwidth and ␪e is the elevation beamwidth both expressed in radians. Beam dwell time is given by Td = θa /ω where 웆 is the antenna azimuth scan rate. The average transmitted power is Pa = Pt δt /Tr Scan rate is related to single revolution scan time, Ts, as ω = 2π/Ts Elevation beamwidth is given by θe = φe /Nb where ␾e is the required elevation angular coverage and Nb is the number of elevation beams used in the search pattern. Finally, the total volume search time is

735

system requirements, whereas noise power density must obey the laws of physics. Of the remainder, the designer may choose values of average power and antenna area while minimizing system losses and receiver noise figure. Of most importance is the product of average power and area. This product is termed the power-aperture product, and its value dictates directly the performance of the radar. Two parameters remain. These are volume search time and target range. These are subject to optimization. Optimum Scan Rate and Instrumented Range There will be a system requirement on the minimum acceptable detection range. In military applications, weapon response times dictate this range. In nonmilitary systems such as airport traffic control, there exists a desired minimum range so that traffic may be diverted or altered to ensure safe arrival of all aircraft. Let the minimum range be denoted Rd. The search radar requires a finite time to search its entire designated volume. It is possible that the last opportunity for detection occurs at the leading edge of this search time. The corresponding initial range is Ri = Rd + Vt Tv where Vt is target incoming velocity. If detection does not occur at that range, the next opportunity may be inside the minimum range. The ENR at initial range is given by ENR =

Tv Pa σt Ar L · 8π 2 φe ηF (Rd + Vt Tv )4

It seems reasonable to maximize ENR at the initial detection range. That occurs when the rightmost factor, involving Tv, is maximum. That maximum is found by differentiation. The result shows an optimum search time given by Tv = Rd /3Vt Substitution of this value gives the optimized initial range as Ri = 43 Rd This is the well known law of search radar design. Define k = 33 /44 8π 2

Tv = Nb Ts Then, maximized ENR at initial range is given by Insertion of these relations into the original range equation yields the following form, which highlights the critical radar parameters: ENR =

Pa σt Ar LTv 8π 2 φe ηFR4

Power-Aperture Product The designer has control over some parameters but not others. Target cross section and elevation coverage are fixed by

ENR =

kPa σt Ar L φe ηFVt R3d

This final form of the range equation shows the importance of power-aperture product and the impact of system design requirements. It is to be noted that, in this form of the range equation, ENR is inversely proportional to the third power of range, whereas the standard form suggests a fourth power dependency.

736

SEARCH RADAR

Preliminary Design Example

The resulting elevation beamwidth is

Manipulation of the radar range equation is such an integral part of the preliminary design process that an example is warranted. Solving the previous equation for power-aperture product gives

θe =

The antenna scan rate is

ENR · φe ηFVt R3d Pa Ar = kσt L

ω=

Suppose requirements are that the elevation coverage be 45 degrees (0.785 rad), the maximum target velocity is 1100 m/s, the desired declaration range is 10 km, and the minimum target cross section is 0.01 m2. Usually, range equation parameters are tabulated in terms of decibels. Power levels are expressed in decibels relative to one milliwatt (dB/mW). Furthermore, assume that the receiver noise figure is 3.0 dB and that the total system losses are 10.0 dB. Also, assume that an ENR of 13.0 dB provides adequate detectability and accuracy. Noise power density is fixed at ⫺174.0 dB relative to 1 mW/Hz and the constant, k, reduces to ⫺28.7 dB. These parameters are listed in Table 1 and the required power-aperture product is computed. Therefore, the required power-aperture product is 50 dB/mW ⭈ m⫺2. Suppose that a 1 m2 antenna fits the available volume constraints. Then, the required average transmitted power is 50 dB/mW or 100 watts. Now, the designer may work backward to find appropriate values for the other pertinent parameters. Assume that a pulse width of 200 ns is desired for range accuracy and that a repetition interval of 10 애s is acceptable from an ambiguity resolution standpoint. These decisions dictate a peak transmitted power of

Gt =

4πAr λ2

where ␭ is the wavelength. Wavelength is related to radar frequency, Fr, as λ = c/Fr where c is the velocity of propagation (3.0 ⫻ 108 m/s). Assume that this radar has been assigned to operate in the X-band at Fr ⫽ 10 GHz. Then, the wavelength is λ=

3.0 × 108 = 0.030 m 10 × 109

and the resulting antenna gain is Gt =

4π = 13, 963 (41.4 dB) (0.030)2

From the gain, azimuth beamwidth may be computed. Thus,

or 5 kW (67 dB/mW). The optimized volume search time is

θa =

Rd 10, 000 =3s = 3Vt 3 × 1100

Suppose that elevation accuracy requires four elevation beams. Then, the single revolution scan time is Ts =

2π 2π = 8.38 rad/s = Ts 0.75

Antenna gain is related to aperture area as

Pa Tr 105 × 10 × 10−6 Pt = = = 5 × 106 mW δt 200 × 10−9

Tv =

φe 45 = 11.25 deg (0.196 rad) = Nb 4

Tv 3 = = 0.75 s (80 RPM) Nb 4

Table 1. Example of Computation of Required Power-Aperture Product Parameter

dB

ENR ␾e ␩ F Vt R 3d k ␴t L P a Ar

13.0 ⫺1.0 ⫺174.0 3.0 30.4 120.0 28.7 ⫺20.0 10.0 50.1

4π 4π = 0.004459 rad (0.26 deg) = Gt θ e 13963 × 0.196

The azimuth dwell time is Td =

0.00459 θa = = 0.000548 s (0.548 ms) ω 8.38

and the number of pulses available during the dwell is Np =

Td 0.000548 = = 55 (17.4 dB) Tr 10 × 10−6

The range of initial detection is Ri =

4 4 × 10, 000 Rd = = 13737 3 3

Expressed in decibels, the fourth power of this range is 165.0 dB.

SEARCH RADAR Table 2. Example of Verification of Energy-to-Noise Ratio Parameter

dB

Pt 웃t Gt ␴t Ar Np L (4앟)2 R 4i ␩ F ENR

67.0 ⫺67.0 41.4 ⫺20.0 0.0 17.4 ⫺10.0 ⫺22.0 ⫺165.0 174.0 ⫺3.0 12.8

The previously derived parameters may be inserted into the original range equation to verify ENR at an initial detection range. Table 2 shows this calculation. The ENR obtained is very close to the required 13.0 dB. The discrepancy is in rounding errors. Essentially, the major radar parameters have been established and the preliminary design is complete. Admittedly, the foregoing exercise is idealized. In practice, several iterations would be required to firm the design.

The corresponding pulse repetition interval (PRI) is 136.4 애s. The two-way time delay to and from a target at range, R, is given by τ = 2R/c Therefore, the first blind range occurs at R=

3 × 108 × 136.4 × 10−6 c · PRI = = 20.5 km 2 2

A crucial factor in MTI design is the fraction, p, of PRF for which an adequate response is required. In this case, that fraction is given by p=

Doppler Shift

F =

2Vt Fr c

where Vt is the radial component of target velocity, Fr is the radar frequency, and c is the velocity of propagation. The impact of Doppler shift on radar design is illustrated best by example. Suppose that it is required to detect targets having incoming velocities ranging from 100 to 1000 m/s. For the first case, assume that the assigned radar frequency is in the L-band at Fr ⫽ 1.0 GHz. The minimum Doppler shift is

Fmin =

2 × 100 × 109 = 667 Hz 3.0 × 108

and the maximum is

Fmax =

2 × 1000 × 109 = 6667 Hz 3.0 × 108

6667 − 667 = 0.82 7334

This response is easily obtained using appropriate feedback and feedforward coefficients in a two-stage canceler. Thus, the design appears sound. Now, consider the effect of increasing the transmission frequency. Suppose the radar is to operate in Ka band at Fr ⫽ 35 GHz. In this case, the minimum target Doppler is

Fmin =

Modern radar takes advantage of Doppler shift to separate moving targets and stationary clutter. This phenomenon manifests itself as an increase in the carrier frequency received from an incoming, moving target relative to the frequency of transmission. The magnitude of the Doppler shift is given by

737

2 × 100 × 35 × 109 = 23, 333 Hz 3.0 × 108

and the maximum is

Fmax =

2 × 1000 × 35 × 109 = 233, 333 Hz 3.0 × 108

If a PRF of 7334 were used here, there would be 32 blind velocity zones across the target spectrum. This is probably unacceptable. If a PRF were selected to center the target spectrum, its value would be PRF = 233, 333 + 23, 333 = 256, 666 Hz The corresponding PRI is 3.9 애s with blind ranges occurring at a spacing of 585 m. Again, this value is probably unacceptable. Obviously, a compromise is warranted but, whatever the PRF choice, there will remain a large number of both blind velocity zones and blind range zones. This example shows the difficulty of designing at a high radar frequency. Coherency Effective utilization of the Doppler shift requires that the transmission and local oscillators bear a constant time-invariant, phase relation. This is termed coherency. An example best illustrates its importance. The carrier signal during a given transmission may be represented by

Suppose further that a clutter canceler such as MTD is employed. One characteristic of these cancelers is that their frequency response function repeats at intervals of the pulse repetition frequency (PRF). That is, nulls will appear at zero frequency and at integral multiples of the PRF. In this case, it seems reasonable to center the expected target spectrum between nulls. This would require that

where 웆1 is the radian frequency of transmission and ␾1 is an arbitrary phase term. The return from a clutter scatterer located at a time delay ␶ is given by

PRF = 6667 + 667 = 7334 Hz

f r1 = cos[ω1 (t − τ ) + φ1 ]

f t1 = cos(ω1t + φ1 )

738

SEARCH RADAR

In the noncoherent case, the local oscillator signal used to convert the return to video may be written as Ilo = cos ω2t There are a number of procedures for processing the video output. A preferred approach is to utilize a quadrature local oscillator given by Qlo = sin ω2t It is assumed that the two local oscillators are free-running at a constant frequency. During transmission of the second pulse, assume that the transmission has changed frequency and phase. Its form is now f t2 = cos(ω3t + φ3 ) In the simplest of clutter cancelers, the two samples in each channel are subtracted and a magnitude formed as the square root of the sum of the squares. Simple trigonometric manipulation shows that the output magnitude has the following form. M = {2 − 2 cos[φ1 − φ3 + (ω3 − ω1 )τ + (ω2 − ω3 )T]}1/2 In the case where all signals are coherent, all frequencies and phases are equal, the argument of the cosine function is zero and the output magnitude is zero as desired. However, when coherency is not maintained, that argument may range anywhere from zero to 2앟. In the worst case, it will be 앟. Then, the output will be twice the input level and clutter will have been amplified rather than canceled. Clearly, coherency is mandatory if clutter is to be rejected. Signal Synthesis To maintain coherency, it is usual to design the frequency synthesizer based on frequency multiplication of crystal controlled oscillators and offset modulation of these signals. There are numerous techniques for achieving coherency. As an example, one technique is depicted in Fig. 2. This example shows a typical interface between the signal synthesizer and the microwave and intermediate frequency receivers. Based on this block diagram example, an example of the methodology used to select synthesizer parameters is pre-

ff1

ft

Antenna ft

fl1

xN3

xN4

ff2

Video

fl2 xN5

f4 f2 f3

S

f0

xN2 f1 S

xN1

Figure 2. Signal flow diagram using typical frequency synthesizer and radar receiver. Coherent reception is ensured by careful selection of local oscillator and intermediate frequencies.

sented. For this example, it is assumed that the system is to operate at 10,000 MHz. The first decision is the selection of oscillator frequency, f 1. This signal will serve as the system clock. Among other functions, it will provide timing for pulse width and pulse repetition interval events. Obviously, its selection also sets the second intermediate frequency (IF) so that final frequency conversion is to direct current or video. A good choice is f 1 ⫽ 100 MHz. The period of this signal is 10 ns, which allows for timing events in integral multiples of this period. It is also a good choice for the second IF where filtering must be provided to pass only the received pulse spectrum while rejecting interference outside that spectrum. Filter design is relatively straightforward when the required bandwidth is, roughly, one-tenth of the center frequency. Thus, the selected IF will support a pulse spectrum of 10 MHz which corresponds to a pulse width of 100 ns. The next selection is that of the first IF. The transmission frequency is f t = N1 N3 f 0 and, assuming that the sum frequency is taken at the offset modulator output, the first local oscillator signal is f l1 = (N1 f 0 + N2 f 1 )N4 = N1 N4 f 0 + N2 N4 f 1 Assuming that the local oscillator (LO) frequency is higher than the transmission frequency, the first mixer output or first IF is given by f f1 = f l1 − f t = N1(N4 − N3 ) f 0 + N2 N4 f 1 The oscillator f 0 is varied to provide tuning of the transmission frequency. Therefore, it must not be allowed to vary the first IF. This dictates that N4 ⫽ N3 and the resultant first IF is f f1 = N2 N4 f 1 It is always good design practice to avoid image responses in the first IF. When the first LO is spaced by the IF above the transmission, it will transfer the transmission to the IF. However, it will transfer also any signal spaced by the IF above the LO back to IF. The latter is termed an image response. Normally, the microwave receiver provides filtering to pass only those RF channels assigned to that system and to attenuate other signals outside that band. In the example, suppose that the allocated band ranges from 10,000 MHz to 11,000 MHz. Then, if an IF of 500 MHz was chosen and the target frequency was 10,000 MHz, the image would lie at 11,000 MHz within the receiver passband. A higher IF would force the image down the slope of the receiver filter and provide some image attenuation. Thus, it is reasonable to require the first IF to be at least one-half of the allocated system bandwidth. Even higher values are encouraged to provide more image attenuation. In the example, it is seen that the first IF must be an integral multiple of 100 MHz. Furthermore, that integer must be factorable into two smaller integers. This means that a value of 700 MHz is not allowable because 7 is a prime integer. Similarly, 1100 and 1300 MHz are disallowed. A reasonable choice is 1500 MHz. Then, the multiplication integers might be chosen as N2 ⫽ 3 and N3 ⫽ N4 ⫽ 5.

SEARCH RADAR

The next choice is the selection of frequency for the other basic oscillator, f 0. By requirement, it must follow that N1 N3 f 0 = 10,000 However, since N3 ⫽ 5 has been chosen, the requirement reduces to N1 f 0 = 2000 Here, a good choice might be f 0 ⫽ 200 MHz. This would require N1 ⫽ 10 which decomposes nicely into 5 ⫻ 2. From an implementation standpoint, smaller integer multipliers are to be preferred. The final selection is the second LO. The only choices are 1400 and 1600 MHz. The former requires N5 ⫽ 7 ⫻ 2, whereas the latter requires N5 ⫽ 4 ⫻ 4. Clearly, the better choice is f l2 ⫽ 1600 MHz. Frequency multiplication is achieved by applying a given signal to a nonlinear device. This device produces harmonics of the input frequency. Subsequent filtering removes the undesired harmonics with the residual being the multiplied frequency. The simplest case is the frequency doubler. Here, the nonlinear characteristic is y ⫽ x2. Assume that the input signal is given by x = cos[ω1t + φ(t)] where 웆1 is the input radian frequency and ␾ is a time varying noise process. Then, the output signal is y = 12 {cos[2ω1t + 2φ(t)] + 1} Thus, the output frequency is twice that of the input. However, note that the noise function has been doubled also. It is a characteristic of frequency multipliers that frequency modulated noise will be multiplied also by the multiplication factor. This disadvantage is offset by the basic coherency afforded by these devices. By simple signal tracing, the reader may verify that the various frequencies produced in the given multiplication and heterodyning processes are reduced to direct current at the video output. By analogy, all noise processes are canceled by the time they reach video. This is the case only because no time delays were postulated in any of the circuit paths. In particular, the time delay to and from a clutter scatterer was ignored. In reality, the FM noise associated with oscillator f 0 and reflected from clutter will not be canceled totally and will appear at the video output. This noise is of serious concern to the designer and its level at the source must be controlled carefully. Power Amplifier Once the system designer has chosen average power, pulse width, and peak power, little else is required for the power amplifier element. The remainder of the design becomes a circuit design problem. However, the designer must be aware that this element represents the best candidate for cost effectiveness. Power amplifiers and their attendant modulators and power supplies are expensive, large, and heavy. Highpower amplifiers are highly expensive, very large, and very heavy. Moreover, they represent a hazard to operating per-

739

sonnel owing to the high voltages involved and the high radiation levels required. Failure rates are often high due to the large component stress induced by generated heat within internal circuitry. Finally, periodic maintenance is complicated by the weight of components and the fact that most components must be submerged in heat-dissipating liquids. All of these disadvantages are balanced only by the single gain factor offered by this element. In making system level trade-offs, the careful designer is advised to select the largest antenna possible to reduce the requirements on the power amplifier. With the possible exception of exotic phased arrays, antennas are, generally, less expensive, more reliable, less hazardous, and more maintainable than their power amplifier counterparts. In addition, the larger antenna offers narrower beamwidths and increased angular accuracy. Power output notwithstanding, radar frequency bandwidth may be an important design consideration. The first requirement is the ability to pass, with high fidelity, a narrow pulse spectrum. Pulse widths on the order of 100 ns having bandwidths of 10 MHz are not uncommon. The klystron vacuum tube provides an instantaneous bandwidth capable of supporting most pulse applications. However, when rapidly tuned frequency agility is required, amplifier bandwidth becomes very critical. In applications designed to thwart narrowband spot jamming, agility bandwidths on the order of 1000 MHz may be required. Tuning intervals of only a few milliseconds may be dictated. These parameters are supported easily by the frequency synthesizer, but, for example, a mechanically tuned klystron will not meet the tuning speed desired. In that case, wideband devices such as the traveling wave tube (TWT) must be used. A potentially serious problem arises when broadband power amplifiers are utilized. This is radiation of broadband noise. The amplifier not only provides gain to the transmission signal but to thermal noise as well. When this energy impinges another friendly radar in the vicinity, the noise can cause sensitivity degradation in the victim radar. Selectivity in the victim radar can be used to reject the signal spectrum of the interference but will not reject the broadband noise. Table 3 depicts a typical scenario. In the table, ␩ is the inherent noise power density at the power amplifier input, Ga is the amplifier gain, F is its noise figure, Gt is transmit antenna gain, R is a typical separation between interferer and victim, Ar is the victim antenna capture area, Gsl is a typical victim sidelobe rejection, and Pr is the resultant noise density at the victim receiver. Suppose that the victim receiver noise figure is 3 dB. Then, the reTable 3. Computation of Received Noise in a Typical Interference Scenario Parameter

Value

Units

␩ Ga F Gt 4앟 R2 Ar Gsl Pr

⫺174 60 20 40 ⫺11 ⫺66 0 ⫺20 ⫺151

dB/mW · Hz⫺1 dB dB dB dB dB/m2 dB/m2 dB dB/mW · Hz⫺1

SEARCH RADAR

ceiver noise density in the absence of interference is ⫺174 ⫹ 3 ⫽ ⫺171 dB/mW ⭈ Hz. Thus, the interference causes a 20 dB sensitivity degradation each time the main beam of the interferer passes the victim. This unacceptable situation may be alleviated by reducing the amplifier gain and, consequently, its output power. However, that would require an increase in antenna gain with no overall benefit. Reduction of the victim sidelobe level would help but that antenna may not be within the purvey of the designer of the interfering radar. The ultimate solution might be in incorporating a low-noise high-gain, preamplifier to drive the main power amplifier. This has the potential for reducing the noise output by 17 dB but the problem remains. Scenarios of this type require careful consideration.

0 –5 Antenna power gain (dB)

740

–10 –15 – 20 –25 – 30 – 35 – 40 –8

Antenna Once the basic antenna area has been selected based on power-aperture product requirements, the designer is free to choose the horizontal and vertical dimensions. Longer dimensions produce narrower beamwidths in that plane. Thus, the antenna might be long horizontally and short vertically. This would produce a fan beam narrow in azimuth and broad in elevation. However, certain applications might require a rotated fan beam with narrow elevation and wide azimuth beamwidths. In some cases, a square or circular aperture might be selected to give comparable beamwidths in both planes. In all cases, the antenna gain should be constant to satisfy range equation requirements. The antenna radiation pattern is, by reciprocity, identical to the reception pattern. The far-field electrical field intensity in a given plane, analogous to voltage in an electronic circuit, is determined by taking the Fourier transform of the current distribution across the antenna face in that dimension. Thus, a/2 z E(φ) = A(z)exp j2π sin φ dz λ −a/2 where ␾ is the angle off boresight, a is the linear dimension of the antenna, A(z) is the current distribution, and ␭ is the wavelength. When the current distribution is constant or uniform, the gain is maximized and the normalized pattern is given by E(φ) =

sin[π(a/λ)sin φ] π(a/λ)sin φ

The power gain, in decibels, is Gt = 20 log E(φ) This pattern is plotted in Fig. 3 for the case where a ⫽ 1.0 m and ␭ ⫽ 0.03 m. Sidelobe levels may be depressed by weighting the distribution as a function of length. An example is the cosine function for which A(z) = cos

πz a

The corresponding normalized intensity pattern is π sin(ψ + π/2) sin(ψ − π/2) + E(φ) = 4 ψ + π/2 ψ − π/2

–6

–4

–2 0 2 Off-axis angle (deg)

4

6

8

Figure 3. Comparison of antenna patterns using uniform and cosine aperture illumination. A cosine illumination gives a significant reduction in unwanted sidelobes.

where ␺ ⫽ 앟(a/ ␭) sin ␾. In this case, the aperture is not illuminated fully and the main beam gain is reduced to 0.81 (⫺0.9 dB) relative to the uniformly illuminated case. This pattern is shown in Fig. 3 also. The uniform distribution yields a first sidelobe level of 13 dB relative to mainlobe gain while the cosine distribution gives a 23 dB first sidelobe. The respective beamwidths are 1.53⬚ and 2.07⬚. Skolnik (2) has a definitive dissertation on antenna design. Microwave Receiver The most important parameter associated with the microwave receiver is the noise figure. That parameter sets the basic sensitivity of the radar. The noise figure of any electronic device is defined as the ratio of input SNR to output SNR. It is a measure of the contribution of that device to overall system noise level. If the first component in the receiver chain is the mixer used to convert to first IF, then the loss in that mixer dictates the noise figure. This loss could be as high as 10 dB even for a well-designed mixer. If this noise figure is unacceptably high, it may be reduced by incorporating a low noise radar frequency amplifier ahead of the first mixer. The noise figure of the combined amplifier-mixer is given by F0 = F1 +

F2 − 1 G1

where F1 is the noise figure of the amplifier, F2 is the mixer noise figure, and G1 is the amplifier gain. In a typical design, F1 might be a factor of two (3 dB) and the gain might be a factor of 100 (20 dB). Then, F0 = 2 +

10 − 1 = 2.09 (3.2 dB) 100

This improvement of almost 7 dB equates to a reduction in required transmission power or antenna aperture of approximately a factor five.

SEARCH RADAR

A second important feature of the microwave receiver is component protection from the high power level of the transmission. Often, this protection is provided by a circulator. Figure 4 shows a schematic representation of this device. The circulator may be viewed as a circular racetrack where the line length between the transmitter port and the receiver port is one-quarter wavelength, the length between transmitter port and antenna port is also one-quarter wavelength and that between antenna port and receiver port is one-half wavelength. At the transmitter port, the signal is divided with onehalf traveling one-quarter wavelength counter-clockwise to the receiver port and one-half traveling three-quarter wavelengths clockwise to the receiver. Thus, the two signals arrive out of phase at the receiver port and cancel. On the other hand, the received signal at the antenna port is divided also but, on both the counter-clockwise and clockwise paths, the signals traverse the same one-half wavelength. Thus, they arrive in-phase and recombine to preserve the target return. Theoretically, the transmission will be canceled exactly with no residue left at the receiver input. However, the circulator line lengths can be designed only for one particular frequency. When the radar must operate over a band of frequencies, the line lengths may not match the actual frequency. The result is imperfect cancellation. Circuit tolerances may also alter line lengths. For these reasons, only isolation on the order of 20 dB can be guaranteed. A final function provided by the microwave receiver is frequency selectivity. The receiver need pass only the pulse spectra of transmissions within it’s allocated operating band. Signals received from radars operating outside this band should be rejected to avoid potential interference. When the radar frequency is much higher than that of expected interferers, then the natural cut-off frequency of wave guide may provide adequate selectivity. However, in most cases, special filtering circuitry must be employed. This problem is compounded further by the fact that the pulse emission spectral width essentially, is, infinite. The spectral density is given by H( f ) =

sin(π fτ ) π fτ

2

where f is the frequency away from the carrier and ␶ is the pulse width. The envelope of the spectrum is (in decibels), He ( f ) = −20 log(π fτ ) Suppose that the frequency separation between victim and interferer is 5 GHz and the interference pulse width is 200

λ /2

Antenna

Receiver

λ /4

λ /4

Transmitter Figure 4. Circuit schematic of a circulator used for receiver protection. The circulator utilizes path length differences to cancel the transmission of the receiver port.

741

Table 4. Computation of Received Power Level in a Typical Interference Scenario Parameter

Value

Units

Pt Gt 4앟 R2 Ar Gsl Ls Pi

70 40 ⫺11 ⫺86 0 ⫺20 ⫺70 ⫺77

dB/mW dB dB dB/m2 dB/m2 dB dB dB/mW

ns. Then, the spectral attenuation at the victim frequency is Ls = −20 log(π × 5 × 109 × 200 × 10−9 ) = −70 dB A typical interference scenario is depicted in Table 4 where the applicable range equation is tabulated. In the table, Pt is the assumed interferer peak power level (10 kW), Gt is its antenna gain, R is the range separation (20 km), Ar is the victim capture area, Gsl is the victim maximum sidelobe level, Ls is the spectral attenuation, and Pi is the resultant interference level received by the victim radar. Assume that the noise figure of the victim radar is 3 dB and that its ultimate processing bandwidth is 5 MHz. Then, the victim receiver noise level is Pv = −174 + 67 + 3 = −104 dB/mW Thus, the interference SNR is 27 dB, which guarantees unwanted interference detection in the victim radar. Intermediate Frequency Receiver Attributes of the IF receiver that are of concern to the system designer are overall gain, IF filtering, video filtering, and digital converter saturation protection. The receiver must provide sufficient gain to place minimum signals and noise within the digital converter range. Typically, the receiver noise is set at about 2 mV at the converter input. Assuming a load of 50 ⍀, this represents a power level of ⫺41 dB/mW. The receiver input noise level is dependent on receiver bandwidth and noise figure. Assume a bandwidth of 5 MHz and a noise figure of 3 dB. Then, the input noise level is ⫺104 dB/mW. The required gain is then 63 dB. This is a net figure and must account for losses in frequency conversion and filtering. Usually, IF filtering is provided to ensure the passage of the return pulse with adequate fidelity but to reject unwanted interference signals. Fidelity will require an IF bandwidth on the order of ten times that of the received pulses. Thus, a pulse width of 1 애s might require an IF bandwidth of 10 MHz. The objective is to provide only sufficient filtering to reject interference but to not compromise the design of the matched filter at video. A very important component of the receiver design is the filter that immediately precedes the digital converter. Commonly this is referred to as the matched filter. Here, an attempt is made to maximize the SNR available at the instant when the pulsed return is sampled. Choices for the architecture of this filter are limitless. The simplest case is the first-

742

SEARCH RADAR

order low-pass filter. In the Laplace transform domain, this filter has a transfer function given by H(s) =

ωc s + ωc

where 웆c is a radian corner frequency defining the filter bandwidth. The magnitude squared response of this filter is H( f ) =

f2

f c2 + f c2

where 웆c ⫽ 2앟f c and f is the frequency variable. The noise equivalent bandwidth of this filter is Beq =

1 2

∞ −∞

f2

π fc f c2 df = 2 + fc 2

For a rectangular input pulse of width ␶, the peak response occurs at t ⫽ ␶ and is given by Vmax = 1 − exp(−2π f c τ ) At the peak, the SNR is SNR =

[1 − exp(−2π f cτ )]2 η(π f c /2)

where ␩ is the receiver noise power density. A plot of this function will show that its maximum occurs when f c␶ ⫽ 0.2. Matched filter theory states that the maximum SNR attainable is equal to twice the ratio of pulse energy to noise power density. That is, Smax = 2τ /η In this case, the ratio of optimized SNR to maximum achievable may be shown to be ⫺0.9 dB. This loss may be reduced by using higher order filters at the expense of increased circuit complexity. Schwartz (3) derives the matched filter theorem and gives detailed examples. It is imperative to ensure that input signals do not exceed the dynamic range of the digital converter. Excess signals will not be decoded properly and the output data may cause interference. It is usual to provide converter protection by implementing a limiter in the final IF stage. The limit level is set somewhat below the actual converter peak capability to account for receiver gain fluctuations. This offset, effectively, reduces receiver dynamic range. In advanced designs, level sensors and feedback may be used to adjust receiver gain so that the limit level may be set very close to the peak of the converter range. The limiting should not be applied at video. The nonlinearity of the limiter creates harmonics of the input signal. If that signal is clutter, these harmonics will not be canceled in the MTD and undue interference will result. Waveform The term waveform refers to the pulse width and pulse repetition frequency (PRF) of the transmitted signal. However, other characteristics such as frequency, phase, and amplitude modulations may be included in its definition.

Generally, pulse width is selected to yield the required range measurement accuracy. However, a longer pulse may be required to achieve the energy necessary for reliable detectability. Pulse widths ranging from a few tens of nanoseconds to a few microseconds are not uncommon. In those cases where available peak power is limited, it may be advisable to incorporate pulse compression. In this technique, modulations are applied to a relatively long pulse on transmission. Appropriate circuitry within the receiver decodes these modulations to produce a narrow pulse. The most common modulation is binary phase coding. This consists of applying either a zero or 180 deg phase shift to each subpulse within the longer pulse. Decoding is achieved by applying the received pulse to a series of delay lines spaced by multiples of the subpulse width. The output of each delay is multiplied by the inverse code and the outputs summed to produce the compressed pulse. The drawback to this technique is that it produces unwanted responses termed time sidelobes which are a source of potential interference and confusion. There exists a class, Barker codes, which yield alternating sidelobes of zero and unity while producing a compressed amplitude equal to the code length. These are in popular use but the maximum known length is 13. Pulse repetition frequency may be used to classify a given radar into one of three categories. In a low PRF system, the pulse repetition interval (PRI) is chosen such that no range blinding occurs over expected target ranges. However, there may be blind velocities produced by the MTI. A high PRF design dictates no velocity blinding over expected target velocities but blind ranges may occur. A compromise waveform is medium PRF which has both range and velocity blinding. The choice of category is application specific, with no particular choice being universally superior. In practice, targets may be detected at extremely long range and may have very high velocity. Thus, pure low or high PRF systems, probably, do not exist. All waveforms may be viewed as medium PRF. The choice of waveform toward low or high PRF depends upon the acceptable number of blind range zones and blind velocity zones. Total range and velocity blinding may be avoided by choosing a waveform set having multiple repetition frequencies. This multiplicity also allows for range and velocity ambiguity resolution. The PRF may be changed on a scan-to-scan basis or during each beam dwell. The disadvantage of the latter approach is that clutter transients may follow a PRF change and reduce the effectiveness of clutter cancellation. This may require receiver blanking during the clutter settling interval, which wastes valuable transmission energy and reduces detectability. Conversely, the former approach complicates ambiguity resolution because the target will probably change range between scans. Range ambiguity resolution is illustrated best by assuming PRF transitions during a single beam dwell. In this case, it may be assumed that target range remains constant while both frequencies are processed. Let the repetition interval on the first PRF be T1 and on the second be T2. Assume that the measured time delay between transmit pulse and detected pulse is ␶1 on the first PRF and ␶2 on the second. These are the ambiguous range measurements. Assume that the unknown unambiguous time delay is T. That value must satisfy both of the following equalities.

SEARCH RADAR

T = τ1 + N1 T1 T = τ2 + N2 T2 Usually, determination of the two integers, N1 and N2, is accomplished by iteration or trial and error. Either integer may be used to compute true range. Because the combined pattern of two frequencies will repeat at some long range, the resolution procedure, in reality, only resolves the ambiguity to within some other ambiguity. The pattern repeats, in general, at the product of the two repetition intervals when those intervals are expressed as integral multiples of the sampling interval unless there are common factors involved. As an example, assume that T1 ⫽ 29.0 microseconds (애s) and that T2 ⫽ 37.0 애s. Furthermore assume a sampling interval of Ts ⫽ 0.2 애s. Then T1 ⫽ 145 Ts and T2 ⫽ 185 Ts. The product is P ⫽ 145 ⫻ 185 ⫻ 0.2 ⫽ 5365 애s. However, the common factor of five reduces the pattern repetition to 1073 애s (161 km). A better choice might be T1 ⫽ 29.8 애s and T2 ⫽ 36.2 애s. Then, the product is P ⫽ 149 ⫻ 181 ⫻ 0.2 ⫽ 5393.8 애s (809 km) and, since the integers are prime, that is the maximum resolved ambiguity. Dynamic Range The ratio of maximum to minimum signal levels that may be processed linearly without concern for harmonic production or small signal suppression is termed the dynamic range of the system. This parameter is set either by the characteristics of the microwave and IF amplifiers or those of the analog-todigital converter (ADC). For low-level input signals, electronic amplifiers are linear and the output level is proportional to the input level. However, as the input level increases, the output begins to approach a constant independent of input level. This condition is termed saturation. When a high-level clutter signal and a low-level target signal are processed simultaneously through a saturated amplifier, the target signal is suppressed. Although the receiver noise is suppressed also, the FM noise carried by clutter is not affected. The overall result is a decrease in target SNR. The ADC usually determines the system dynamic range. The theoretical noise level due to quantization in the ADC is of one least significant bit (LSB). This is equivalent to a power level of ⫺11 dB relative to the LSB. However, practical ADC devices exhibit noise levels well above theoretical. Typically, a noise level of ⫺1 dB can be expected. The maximum signal that may be processed linearly is dependent upon the number of bits available at the ADC output. Consider a 12bit device. Because one bit must be assigned to the sign of the output, the maximum peak signal is 20 log (211) ⫽ 66 dB. The corresponding rms level is 63 dB. Thus, the apparent dynamic range is 64 dB. The ADC noise will add to the input receiver noise and reduce sensitivity. This reduction may be minimized by setting the receiver noise somewhat higher than that of the ADC. For example, a receiver noise level 6 dB above ADC noise yields a 1 dB degradation. Now, the dynamic range is reduced to 58 dB. Compounding the problem are fluctuations in receiver gain over ambient temperature. Should that gain drift downward by 6 dB, the sensitivity degradation would be 3 dB. This unwanted circumstance is avoided by elevating the nominal re-

743

ceiver noise level an additional 6 dB. If the gain should increase by 6 dB and if the IF limit level were set exactly at the maximum ADC level, then the higher level would overload the converter. This is avoided by setting the limit level 6 dB below that of the ADC. The overall result of these adjustments is a 12 dB decrease in dynamic range to a value of 46 dB. Additional dynamic range may be obtained by increasing the number of bits available from the ADC. Each added bit represents an increase of 6 dB in the dynamic range. However, adding bits reduces processing speed. Currently, devices are available that output 14 bits at rates of 10 MHz. In the near future, it is expected that 16 bits or more may be achieved at rates up to 30 MHz or higher. It is possible to monitor receiver noise level and, using feedback, to control the receiver gain. This would enable the receiver noise and IF limit level to be maintained constant relative to the ADC parameters. The result would be a 12 dB increase in dynamic range. Use of this technique with a 14 bit ADC enables a dynamic range of 70 dB. Target Detection A primary function of the search radar is to provide detection of targets. Usually, this is achieved by comparing received signals to a preset threshold. For this function, the operative measure of performance is probability of detection. This probability depends upon input SNR and allowable false alarm rates. Detection theory is couched in Rician statistics. This theory treats the problem of the probability that the magnitude of a sinusoidal signal imbedded in Gausian noise will exceed a given threshold. [See Schwartz (4) for a detailed derivation.] The correct threshold is determined by electing an allowable probability of false alarm, 애. That probability is given by µ = exp(−Vt2 /2σ 2 ) where Vt is the threshold voltage and ␴ is the root mean square (rms) noise voltage. It is usual to set this threshold based on a measured value of the average noise magnitude. That average is π m1 = σ 2 The threshold is set at a constant, k, times the measured average. The proper selection of the constant is 1/2 4 k = − ln µ π The threshold is related to the rms noise level and probability of false alarm as 1/2 1/2 − ln µ − ln µ Vt = = 2σ 2 2 where the assumption that the rms noise level is unity does not result in loss of generality. Under that assumption, the probability of target detection is Vt Pd = 1 − q(r) dr 0

744

SEARCH RADAR

where √ q(r) = exp(−s )r exp(−r /2)I0 (rs 2) 2

2

Here, s2 is the target SNR and I0 is the modified Bessel function of the first kind and zero order. The integral cannot be evaluated in closed form. However, it is solved easily by numerical integration on a digital computer. This evaluation is aided by the following power series expansion of the modified Bessel function. I0 (z) =

∞

z2n

n=0

22n (n!)2

A typical application of this technique yields the curves shown in Fig. 5. In the figure, it is noted that reliable detection requires an SNR in excess of approximately 13 dB. The penalty for reducing false alarm rate by one order of magnitude is about 1 dB in terms of required SNR. Seldom is target detection based on the result due to a single received pulse. Rather, integration of some form is employed to reduce the SNR required on a single pulse. Pulse integration is possible since there are, generally, multiple returns available during a beam dwell. Suppose that the antenna azimuth beamwidth is 2 deg, the scan rate is 60 RPM and the repetition interval is 30 애s. These parameters result in the availability of 185 target return pulses over the beam dwell. For simplicity, assume that these pulses are of equal magnitude. Furthermore assume that an allowable overall false alarm interval from the signal processor is 1 s and that the pulse width and attendant sampling interval is 0.2 애s. Then, the number of range cells is 150 and the allowable false alarm interval from each is 150 s. The simplest integration technique is, essentially, to do nothing. Since there are 185 equally probable opportunities for detection, the overall probability of detection is Po = 1 − (1 − Pd )185

Probability of detection

1.0

0.8 u=1.0E– 5 0.6 u=1.0E– 6

µ=

30 × 10−6 = 2.0 × 10−7 150.0

This combination of parameters results in a required, perpulse SNR of 7.0 dB. At the other extreme is completely coherent integration. In this case, voltages from all 185 pulses are added before comparison to a threshold. The result is an increase in SNR of 10 log 185 ⫽ 22.7 dB. On the other hand, the decision interval has been increased by a factor of 185 so that the allowable false alarm probability is now µ=

185 × 30 × 10−6 = 3.7 × 10−5 150.0

At the corresponding threshold, the required integrated SNR for a 0.9 probability of detection is 12.1 dB. The per-pulse requirement is only ⫺10.6 dB. Therefore, the improvement afforded by coherent integration is 17.6 dB. This is a very significant improvement, because it reduces the required power-aperture product by almost two orders of magnitude. Truly coherent integration may be obtained only through the use of fairly sophisticated processes, such as the discrete Fourier transform discussed later. However, there are available simpler approaches that achieve notable improvements. One such technique is the binary, or Markov chain, integrator. The binary integrator is implemented by assigning a digital counter to each range cell in ambiguous range space. If a given sample exceeds a preset threshold, the counter is incremented by one. If not, the counter is decremented. The counter is not allowed to go negative nor to exceed some positive value. A detection is declared when the counter achieves some predetermined threshold. The design of a binary integrator is an interesting exercise. Counter threshold is the parameter subject to trade-off. The first step is determining the proper per-pulse threshold necessary to hold the overall false alarm interval to the allowable level. It may be shown (5) that the expected number of samples required to transition from a counter state of zero to a state of N is i 1 N−1 q EoN = (N − i) p i=0 p where p is the per-pulse probability of false alarm and q ⫽ 1 ⫺ p. For a given counter threshold and allowable false alarm interval, it is straightforward to determine the required value of p and this sets the per-pulse threshold. The basic equation of the Markov chain is

0.4

0.2

0.0

where Pd is the single sample probability. Suppose that an overall detection probability of 0.9 is desired. Then, on a persample basis, the required probability is Pd ⫽ 0.012. The allowable probability of false alarm is the ratio of the repetition interval to the allowable false alarm interval. Thus,

3

6 9 12 Signal-to-noise ratio (dB)

15

Figure 5. Detection performance as a function of signal-to-noise ratio. Curves represent two different false alarm rates. Probability of detection increases as SNR increases but decreases as false alarm rate is decreased.

P(m,n) = qP(m + 1, n − 1) + pP(m − 1, n − 1) where p and q are now target detection probabilities on a perpulse basis and P(m,n) is the probability of being in state m after the nth step. The result is a set of equations that may be exercised iteratively to yield overall probability of detection, after a given number of received pulses, given per-pulse

SEARCH RADAR

probability. Iteration over input probability yields that value which achieves the required overall probability of detection. In the example, that value is 0.9. Finally, interpolation of tables of detection probability as a function of probability of false alarm and SNR yields a required SNR for the given overall probability as a function of the Markov threshold. An iteration over Markov threshold will show that there is an optimum threshold that minimizes required SNR while maintaining the desired false alarm rate. Using the parameters of the current example, the described procedure was executed by computer simulation. The result shows that the optimum Markov threshold is 30 and, at that value, the required per-pulse SNR is ⫺1.4 dB. This represents an improvement of 8.4 dB over the case having no integration. Although this remains 9.2 dB lower than the improvement afforded by coherent integration, it does represent a major improvement because it reduces the required power-aperture product by almost an order of magnitude. Accurate threshold setting depends on an accurate measurement of the ambient noise level. Usually, this measurement is obtained by averaging the magnitudes of several range cells adjacent to a range cell under test. This average should slide as successive range cells are tested for detection. The variance of this measurement may be shown to be V (Na ) =

V (Ni ) N

where V(Ni) is the variance of an individual noise cell and N is the number of samples averaged. Thus, the measurement error decreases as the number of samples increases. Typically, the number of samples used is on the order of 30 or more. It may be shown that the use of averaging contributes approximately 0.5 dB to the total processing loss. This loss is tolerable because averaging serves to maintain a constant average false alarm rate even when receiver gain fluctuates. Target Measurement Accuracy A primary function of the radar is to provide a measure of target location in either two- or three-dimensional space. Thus, the radar must measure target range and target bearing as a minimum. Additional measurements of elevation angle, velocity, and target cross section may be taken in advanced designs. In the simplest design, range is determined by noting the time delay between transmitted pulse and received pulse. In this case, the accuracy can be no better than the pulse width. When improved accuracy is required, further processing of the received signals must be performed. One approach is to use a weighted average of successive range cells. The estimated range is computed as re =

M 1 r1 + M 2 r2 M1 + M2

where M1 and M2 are the magnitudes and r1 and r2 are the measured ranges in two adjacent range cells. This technique is referred to as split gating. The utility of this technique is based on the fact that the video filter output pulse is rounded rather than square so that pulse magnitude varies predictably with time delay.

745

The accuracy of any split gate technique depends upon pulse width and SNR. In general, the rms error is given by √ σr = k r δr / S where kr is a constant depending upon implementation parameters, 웃r is the pulse width measured in units of range, and S is the SNR. At higher SNR, the error may be reduced to a fraction of the pulse width. Note that in systems using pulse integration split gating may be used only if sums of pulse magnitudes are obtained or true coherent integration is employed. It is of no use with binary integration. In simple applications, target bearing is determined by noting the angular orientation of the antenna at the time of detection. Thus, the accuracy can be no better than the azimuth beamwidth of the antenna. When improved accuracy is required, there are a number of beam splitting techniques available. A very sophisticated approach takes advantage of the known shape of the antenna pattern to form a curve fit of the received data. Solution of the resulting set of equations yields a very exact measure of azimuth. In all beam splitting techniques, the rms error will be given by √ σa = k a θa / S where ka is a constant related to antenna parameters and ␪a is the azimuth beamwidth. Higher values of SNR yield errors of a fraction of beamwidth. Skolnik (6) derived theoretical boundaries for measurement accuracy. Except for some advanced designs where the antenna is scanned past the target in elevation as well as azimuth, elevation beam splitting is not possible. Therefore, the error remains the elevation beamwidth and this value may be quite large. This error can have a profound impact on scan-to-scan correlation algorithms. In some applications, it is required that a measure of target cross section be obtained. This parameter might be used, for example, to distinguish bird and insect returns from aircraft. This measurement depends upon a priori knowledge of all the parameters comprising the radar range equation. Since these parameters may fluctuate, especially over time, the accuracy of this measurement probably is limited to ⫾3 dB. Target Acquisition In automatic systems, target acquisition is a requirement. Through this process, a target file is maintained for each detected target. This file contains a record of target range and bearing and, possibly, elevation, velocity, and cross section. Each file is updated after each antenna scan with the latest measurement data. File updates are accomplished by performing scan-to-scan correlation. This process is effected by associating the latest detection data with earlier data using the concept of correlation windows. For each target, a window in space is established into which it may be expected that the next measured parameters from that target may fall. If a given set of parameters falls outside all established windows, then a new target is declared and a new file initiated. Correlation windows are, generally, rather large after initial detection but are allowed to decrease in size as the target history matures. An example of the use of windows may be helpful. Assume a two-dimensional radar having a single elevation beam that

746

SEARCH RADAR

rotates at 60 RPM. Then, the data interval is 1 s. Suppose an initial target detection occurs at a range of 10,000 m and at a bearing of 0⬚ (due north). Assume further that the maximum expected target velocity is 1000 m/s. Then, if the target is traveling tangentially, it could change bearing by 5.7⬚. If it travels radially, it could change range by ⫾1000 m. Thus, an appropriate correlation window might be ⫾6 deg in bearing by ⫾1000 m in range. Suppose, on the subsequent scan, that a target is reported at range of 9500 m and at bearing 0⬚. These parameters are well within the assumed correlation window and these data would be associated with the first detection. In this case, the apparent radial velocity is 500 m/s and the angular velocity is 0. It might be expected, therefore, that the third scan would reveal a target at 9000 m and 0⬚ bearing. Then, the third scan correlation window could be reduced to zero width in both dimensions. In practice, that window must be increased to account for measurement error and possible target maneuver. However, its dimensions would remain much smaller than that for the second detection. The concept of target file may be extended to target trajectory prediction. That prediction has a practical benefit in a variety of applications. For example, a ship defense radar might predict whether a given ballistic missile target will impact the ship or fall harmlessly into the sea. The latter determination would save the waste of one or more defensive weapon rounds. Another example is in airport surveillance where it is desirable to determine if two given targets are on a collision course. Trajectory prediction may be predicated on polynomial curve fitting. Its accuracy improves with the number of detections available and with the accuracy of the individual measurements. A final application, in military designs, is kill assessment. Here, a rapid change in trajectory parameters may indicate weapon damage to the target and may indicate that engagement with further rounds is unnecessary. This ammunition saving is critical since the initial load is finite and there may be other threats following the first. Clutter Rejection It is mandatory that clutter rejection be given a high priority. Exceptions are those systems specifically designed to provide terrain mapping or weather detection. Clutter returns include backscatter from terrain, islands, the sea surface, and rainfall. In military applications, it also may be necessary to consider enemy dispensed chaff clouds. Except for wind driven rainfall and chaff, the clutter Doppler shift for a stationary radar is normally 0. However, its spectral width may be nonzero depending on local motion of the various scatterers. In airborne applications, the clutter spectral bandwidth will be twice the platform velocity. This relatively wide band presents an interesting exercise in the design of clutter cancellation algorithms. Clutter-to-noise ratio at the radar receiver is computed using the radar range equation except that clutter radar cross section is used instead of target cross section. Ideally, clutter levels would be calculated by performing a triple integral over the antenna pattern in two dimensions and over the return pulse shape in range. In practice, that integral is approximated. For surface reflections such as those from terrain and sea, the approximate clutter cross section is σc = Rθa δr σ0

where R is clutter range, ␪a is azimuth beamwidth, 웃r is range cell width, and ␴0 is clutter cross section density. In a metric system, the dimensions of the latter parameter are square meters per square meter of surface area. For volume scattering from rain and chaff, the approximate cross section is σ c = R2 θ a θ e δ r σ 0 where ␪e is the elevation beamwidth and ␴0 is the clutter density in square meters per cubic meter of clutter volume. Over the history of radar, there have been extensive efforts to measure and to model clutter cross section density. These efforts have resulted in a number of empirical models (7). The most precise of these is for rainfall. There, it is shown that the clutter density is proportional to the rainfall rate raised to a power of 1.6 and is proportional to the fourth power of radar frequency. For sea clutter, the models are less precise. However, the trend seems to indicate an increase with wave height, a linear increase with frequency, and a nonlinear increase with grazing angle. Terrain clutter is the least well defined because there are wide fluctuations in terrain type, local ground slope, type and depth of vegetation, presence of manmade structures, and soil composition. Lacking definitive models, the designer must be content with ensuring that all clutter levels within the dynamic range of the system are rejected to an acceptable level. In modern, digital search radar, the most common defense against clutter is the use of MTD. These also may be referred to as clutter cancelers. In essence, these devices subtract the value of one received signal from that received one pulse repetition interval earlier. Since clutter has a zero-frequency Doppler shift, its value is 0 at the canceler output. A Dopplershifted target return, depending on phasing, may be enhanced at the output. When both feed-forward and feedback are employed, the MTD may be viewed as a digital filter. A block diagram of a single stage MTD is given in Fig. 6. A general treatment may be found in Skolnik (8). In this example, one feed-forward coefficient, A1, and one feedback coefficient, B1, are used. The term, 1/z, is the ztransform representation of a delay of one repetition interval.

A1 x

fi

1/z

f0

x

B1 Figure 6. Single stage moving target indicator using both feed-forward and feedback. Frequency response may be shaped by varying the multiplier coefficients.

SEARCH RADAR

66 dB below the clutter carrier. If the noise spectrum were flat, this would translate to a requirement that the noise density, as measured in a 1 Hz bandwidth, be 126 dB below the carrier.

10

Magnitude response (dB)

No feedback 0

747

With feedback

Electronic Countermeasures –10

– 20

– 30 0.0

0.2

0.4 0.6 0.8 Frequency relative to PRF

1.0

Figure 7. Frequency response of a single stage moving target indicator with and without feedback. Note the widening of the response when feedback is used.

A difference equation may be derived for this filter relating output to input. That equation is f o (n) = f i (n − 1) + B1 f o (n − 1) + A1 f i (n) Note that, when A1 ⫽ ⫺1 and B1 ⫽ 0, the filter reduces to a simple subtractor. To avoid blind phases, it is usual to implement identical MTI filters in the in-phase and quadrature channels of the radar Signal Processor. The output magnitude is taken as the square root of the sum of the squares of the two channels. The frequency response of the filter is obtained by plotting the steady-state output magnitude as a function of input signal Doppler shift. Typical results are given in Fig. 7. In Fig. 7, the curve labeled ‘‘no feedback’’ represents the response for the simple subtractor. Its maximum response, 6 dB, occurs for a frequency equal to one-half PRF. However, at low frequencies, the response is poor. At 10% of PRF, a loss of almost 5 dB is incurred. This wide fluctuation is controlled using feedback. The curve marked ‘‘with feedback’’ shows the response for A1 ⫽ ⫺1.0 and B1 ⫽ 0.81. In this case, the response is relatively flat at about 0.5 dB over 80% of the PRF. This response may be improved further by adding more delay lines and coefficients. Adjustment of pole and zero locations in the z-domain transfer function of higher ordered filters is helpful in tailoring the response to specific requirements. Filter designs using four delays and six coefficients are not uncommon. In addition to clutter itself, clutter-carried FM noise can cause serious sensitivity degradation if it is not carefully controlled. The FM noise spectrum is generally wideband with significant levels extending for several megahertz. The total noise at the digital converter input is the integral of the product of the FM noise spectrum and the transfer function of the video filter. This total FM noise must be specified small compared to the total receiver noise level. As an example, assume that the digital converter dynamic range is 60 dB and that the video filter noise equivalent bandwidth is 1.0 MHz. To maintain the degradation due to FM noise at less than 1 dB, it would be necessary to require the total FM noise to be

In military applications, the use of electronic countermeasures (ECM) by the enemy is, practically, a given condition. A prevalent form of ECM is broadband noise jamming. If the enemy can muster sufficient power levels and place this noise sufficiently close to the victim radar, then it can always defeat any radar. In those cases, the radar designer can do little to prevent degradation directly. However, there are scenarios in which jammer identification and/or frequency agility may be used to mitigate this degradation. The total noise power received from a jammer is given by

Pr =

Pj Gj Br Ar Gsl 4πR2j

where Pj is the jammer noise power density, Gj is the jammer antenna gain in the direction of the victim radar, Br is the radar receiver noise equivalent bandwidth, Ar is the radar antenna capture area, Gsl is the sidelobe antenna gain in the direction of the jammer, and Rj is the jammer range. If the existence of jamming can be determined and the bearing angle of its source measured, then various weapons, such as home-on-jam missiles, can be brought to bear to destroy the jamming source. This may or may not be in time to prevent the ingress of hostile targets. Identification and location of a jammer are implemented by maintaining a history of the ambient noise level. When that level increases significantly at a given bearing, it may be assumed that a jammer exists at that bearing. Thresholding and integration length for this technique are dependent upon antenna pattern characteristics and antenna scan rate. However, if the jammer level is sufficient to cause degradation, it will always be detectable by this technique. In the simplest scenario, the enemy is ignorant of the instantaneous frequency used by the victim radar. However, from intelligence data, the basic band of operation is known almost always. In this case, the jammer has little choice but to spread its energy over this entire band. This frequency dilution requires fairly high power levels. In advanced scenarios, the enemy must be given credit for the ability to measure the instantaneous frequency radiated by the victim. Then, jamming energy may be concentrated at that frequency with a consequent reduction in required power level. This is referred to as spot jamming. Suppose that the victim radar operates somewhere within an allocated band of 500 MHz but that the instantaneous radiated and received bandwidth is only 5 MHz. The barrage jammer would have to radiate noise over the entire 500 MHz but the spot jammer bandwidth would need be only 5 MHz. The advantage would be a factor of 100 or 20 dB. Spot jamming may be countered by frequency agility within the victim radar. Because there is a fairly large range separation between radar and ECM system, there will be a finite time delay between radiation initiation from the victim and sensing of that radiation at the ECM system. An additional delay results from the finite tuning time required for

748

SEARCH RADAR

the spot ECM system to change frequency. During the sum of these delays, the victim radar enjoys a noise-free target dwell. If the victim changes frequency after a given delay, the ECM system cannot keep up, and the victim is free to detect targets. A final ECM technique is termed deception. Here, the ECM system receives the victim-radiated pulse and reradiates that pulse with a time delay or Doppler shift. The victim radar thus receives two signals. One is the true target reflection and one is a false target return. Through clever manipulation of the time delay or Doppler shift, the ECM may be successful in decoying the radar to assign a weapon to the false target and, thus, to waste ammunition. However, because this technique does not hide the true target return, its benefit is highly questionable. This rather complex ECM method is not in common practice against search radars due to its lack of real costeffectiveness. It is much more useful against dedicated tracking radars where it may effect stealing of range and velocity tracking loops. Radar Frequency Interference There are a large number of radars in operation worldwide. It is likely that several of these systems may be located in close proximity. Because the one-way propagation loss between two systems falls off at only R2, one radar may detect the direct radiation from another. The result is termed radar frequency interference (RFI). Radar frequency interference may be categorized into two classes. The first class involves two or more radars of the same type. These are designed for the same function, by the same design team, and produced by the same manufacturer. An example is a fleet of ships where each ship carries its own search radar. The second class involves radars of dissimilar type. These have different functions and, probably, have been produced by different companies. An example might be an airport installation using one high-power general surveillance radar and one or more lower power radars to monitor takeoffs and landings. To minimize RFI, the federal government issues licenses, establishes regulations, and allocates frequency bands to every radio, television, and radar system deployed within the United States. Other nations have similar functions. Control is extended to all radiating systems including experimental and developmental models. When two radars in close proximity are of the same type, the designer has some control over RFI. The first defense is channelization. The radar system may be allocated, for example, a total band of 1000 MHz but its instantaneous radiated bandwidth might be only 10 MHz. Assignment of separate channels spaced by 10 MHz to each particular radar would allow 100 systems to operate simultaneously without any two being in the same channel. Unfortunately, channelization does not completely solve the RFI problem. The radiated spectrum of a rectangular pulse has very significant energy many bandwidths removed from the carrier frequency. This energy can enter the victim receiver and cause interference. A second defense is PRF diversity. When two radars operate at the same PRF, the received pulses from the interferer integrate normally and, probably, create a detection. When two different PRFs are used, the integration is negated and the RFI effect is reduced by a factor equal to the integration

length. This is not a total panacea. Selection of PRF is a delicate process involving questions of blind ranges, blind velocities, and range ambiguity resolution. In a given application, there may not be a sufficient band of usable PRFs to serve multiple radars. Moreover, diversity may even compound the problem. If an interference pulse is sufficiently strong, it may be detectable without integration. This pulse will migrate over many range cells and produce multiple detections. This could result in a swarm of false targets that might overcome the correlation capability of the system computer. Neither of the techniques discussed above is applicable to the class involving radars of different types. In this case, the only viable defense appears to be asynchronous pulse detection (APD). In the APD technique, successive pulses in a given range cell are compared. If the later pulse is much larger than the earlier pulse, it is edited from the data stream and replaced by either the earlier sample or a random number. Of course, this approach works only when two different PRFs are involved. Asynchronous pulse detection can be very effective in minimizing RFI. However, depending upon the thresholding scheme employed, it can cause degradation in the detectability of real targets. ADVANCED DESIGN In this section, topics are discussed that represent potential performance improvements to future designs. However, none of these are new innovations. The concepts have been known and understood for many years. As technology evolves, though, these techniques become more and more cost effective and attractive to the designer of advanced systems. Phased Array Antennas Phased array antennas are planar arrays of waveguide slots. Variable phase shifters are used to drive a group of slots and, thus, to effect electronic steering of the antenna beam. This technique can eliminate the necessity for bulky mechanical devices such as motors and gimbaled platforms. The basic theory of phased arrays is described best by considering the simplest case, which is a two-element array. The radiated fields from two adjacent sources combine in space to form a radiation pattern. When a phase shift is applied to one element, the directivity of that pattern may be altered. In this simple case, it may be shown that the relative gain of the array is G=

2 + 2 cos(π sin α − φ) 4

where 움 is the space angle relative to perpendicular and ␾ is the introduced phase shift. Maximum gain occurs when α = sin−1 (φ/π ) Thus, for small angles, the ratio of phase shift to steered angle is approximately a factor of three. In no way does this example present the design equations for a complex array. It simply shows the effect of phase shift on boresight shift. The concept is extended easily to linear arrays of any length, N, and to two-dimensional arrays having N ⫻ M elements (9). In practice, the pattern produced by an array will

SEARCH RADAR

be the product of the pattern from each element and the array factor, which is determined by the element spacing. The optimum spacing is one-half wavelength. At wider spacing, the pattern begins to develop unwanted sidelobes called grating lobes. These can be as large as the main lobe and may cause confusion or interference. In addition, coupling between elements can alter the actual antenna pattern. Phased array design is a complex process. When the array beam is steered off-axis, the beam will broaden and the gain will decrease. In general, this effect is in proportion to the cosine of the steered angle. For example, a steering angle of 45⬚ may result in a loss of 1.5 dB relative to on-axis gain. When the element spacing is one-half wavelength, the number of elements and required phase shifters can become quite large. For example, consider a design at X band where one wavelength is 0.03 m. An antenna 1 m on a side would require 4356 phase shifters to enable steering in both planes. The sheer cost and weight of this system might be prohibitive. A good compromise design is one in which the antenna array is rotated mechanically in azimuth while being steered electronically in elevation. The example given previously would then require only approximately 66 phase shifters to provide elevation-only steering. This approach also allows for raster scanning. Rather than holding the elevation position constant over a full 360⬚ rotation, the beam could be directed to visit several elevation positions during one scan. This not only reduces the time required to illuminate a given volume but, since the beam traverses the target in both dimensions, beam splitting in elevation and azimuth could be implemented. An ultimate phased array design is the conformal array. Here, the array is designed as an integral part of an existing geometry. That geometry might be the fuselage or wing of an aircraft or the hull of a ship. The ideal conformal array would be a sphere or hemisphere. This design could eliminate offaxis steering loss, because the beam would always be perpendicular to the array surface. Another advantage of phased arrays is their capability for instant target verification. An initial detection could be followed by freezing the beam in the direction of the target detection. Then, a longer dwell could be chosen to both reduce measurement error and increase confidence level. The time savings relative to scan-to-scan verification could be significant. A final application of phased arrays is in platform motion compensation. When the radar is carried by an aircraft or ship, it is desirable that the beam position be maintained relative to earth coordinates independent of platform motion. This is implemented easily using beam steering and its use eliminates the necessity for complex motor-gimbal apparatus.

the same coherent signal from a frequency synthesizer. That element also provides one or more local oscillator signals for frequency conversion to IF. The IF signals are combined to form one channel for signal processing. A chief advantage of the modular array is the elimination of a large, expensive, hard-tube transmitter. Its disadvantage is in the huge number of modules required. If 4356 elements are required, then that number of modules is also required. If the cost of each module were only $100, then the total array cost would approach $500,000. For this reason, the modular array has not found a large application. However, as technology improves and costs decrease, this approach may become very popular in future designs. Discrete Fourier Transform The discrete Fourier transform (DFT) is a digital signal processing technique that provides perfect, coherent integration of incoming pulse trains and filter discrimination among target and clutter Doppler shifts. Usually, it is implemented as a bank of equally spaced digital filters that spans a frequency band equal to the PRF. The response of a given filter repeats at integral multiples of the PRF. These filters are generated by summing the product of the incoming samples with a time-varying sequence of coefficients. The latter is designed to produce a narrowband filter at some prescribed frequency. The output of a given filter is

F=

N−1

A[cos 2πfnT r + j sin 2πfnT r ]

n=0

× [cos 2π f c nT r − j sin 2π f c nT r ] where A is the peak amplitude of the input signal, f is the frequency of the input signal, f c is the center frequency of the filter and Tr is the pulse repetition interval. N is the number of pulses integrated. Note that the input and coefficient sequences are expressed as complex quantities. The real part of the input is taken from the in-phase video channel, whereas the imaginary part is taken from the quadrature. The output is also complex valued. The real, or in-phase, output may be shown to be

Fi =

N−1

A cos 2π( f − f c )nTr

n=0

and the imaginary, or quadrature, output is

Fq =

N−1

A sin 2π( f − f c )nT r

n=0

When the input signal frequency coincides with the filter center frequency, the outputs are Fi = NA

Active Arrays The next generation of search radar may well use the concept of the active array. This is a natural extension of the phased array. In that design, each array element is provided with its own transmitter and receiver module. This solid-state module contains a low-power transmission amplifier, receiver protection, a low-noise RF preamplifier, filtering, and frequency conversion. It also contains a digitally controlled phase shifter for beam control. On transmission, all modules are driven by

749

and Fq = 0 The square of the output magnitude, which is output power, is, then, M 2 = Fi2 + Fq2 = N 2 A2

750

SEARCH RADAR

The filter response to noise is given by N−1

–5

[xn + jyn ][cos 2π f c nT r − j sin 2π f c nT r ]

Magnitude response (dB)

F=

0

n=0

where xn and yn are noise samples taken from the in-phase and quadrature channels, respectively. These samples are assumed to be independent from channel to channel and sample to sample. They have a Gaussian probability distribution with zero mean and variance ␴2. The in-phase output is

Fi =

N−1

–10 –15 – 20 – 25

xn cos 2π f c nT r + yn sin 2π f c nT r

n=0

– 30 0.0

The variance of that output is

V(Fi ) =

N−1

V(xn )cos2 2π f c nT r + V(yn )sin2 2π f c nTr

n=0

=

N−1

F=

N−1

AW(n)[cos 2πfnT r + j sin 2πfnT r ]

n=0

It may be shown also that

[cos 2π f c nT r − j sin 2π f c nT r ]

V(Fq ) = Nσ 2 Then, the output SNR is SNR0 =

1.0

Figure 8. Frequency response of a discrete Fourier transform filter with uniform window. The DFT response is relatively narrow but exhibits sidelobes, which may cause extraneous detections.

σ 2 [cos2 2π f c nT r + sin2 2π f c nT r ] = Nσ 2

n=0

0.2 0.4 0.6 0.8 Frequency normalization factor (k)

N 2 A2 A2 = N 2 = N · SNRi 2 2Nσ 2σ

where SNRi is the input, per-pulse SNR. Thus, the DFT provides an SNR gain equal to the length of integration. For example, if that length was 100, the gain would be a very impressive 20 dB. No other known technique yields a higher SNR gain than the DFT. The normalized frequency response of the DFT filter may be shown to be

sin Nφ Gnorm = N sin φ

2

where W ( ) is the window function. A typical function is πn W (n) = sin2 N which reduces the first sidelobe to a level approximately 32 dB below the main lobe response. There is almost an infinite number of windows that might be applied in a given application. Harris (10) presents an entire catalog of windows and lists advantages and disadvantages. The use of windows is not without penalty. Invariably, the main lobe gain will be reduced and the bandwidth will be increased. Thus, a trade-off exists between sidelobe response and main lobe performance. The DFT dwell length is Td = NT r and the filter bandwidth is, roughly,

where φ = π f/f r f is frequency relative to filter center frequency and f r is the PRF. This response is plotted in Fig. 8 for the case N ⫽ 10. In this plot, a frequency normalization factor, k, is used. When k ⫽ 0, f ⫽ 0 and when k ⫽ 1, f ⫽ f r. It will be noted that the response exhibits sidelobes. The first sidelobe is 13 dB below the main lobe response. Also, note that the response repeats at integral multiples of the PRF. If the relatively high sidelobes are not sufficient to provide desired clutter attenuation, then window functions may be applied to suppress these sidelobes. A window function is a real-valued, time-varying sequence applied to the input data for all filters. With a window, the DFT response is given by

B=

1 fr = N Td

Normally, the nominal number of filters implemented is equal to the integration length, N. Then, the filter spacing is almost equal to the bandwidth. This is not an inviolable rule. When the number of filters exceeds N, the loss to frequencies between filter centers is reduced but the filter outputs are correlated somewhat. Fewer filters result in more loss. A very interesting signal processing scheme is the one in which a nonrecursive MTD is used preceding the DFT. This type of MTD, which does not employ feedback, has the very attractive feature that no transients are produced when radar frequency is changed, beams are transitioned, or PRF is changed. Thus, wasteful receiver blanking is avoided. This feature is achieved because the combined MTD–DFT re-

SEARCH RADAR

sponse is constant over most of the PRF bandwidth. Although the MTD attenuates low-frequency signals, it also attenuates noise in their vicinity within the narrow DFT bandwidth. The result is an almost constant SNR over most of the band covered. Thus, the combined response can be made to approximate that of a well-designed MTD with feedback, but transients will be absent. Wilson (11) presents a derivation of SNR loss using this combination. Clutter Tracking In many applications, shipboard and airborne, the radar platform is in motion. In these cases, the received clutter Doppler may be nonzero and its spectrum may not be canceled sufficiently in the relatively narrow null of the MTD response. Improved clutter rejection may be obtained by incorporating a clutter tracker. This subsystem measures the clutter Doppler and, through various tuning schemes, positions the clutter spectrum exactly at the MTD null for maximum suppression. The clutter tracker is a special purpose, automatic frequency control (AFC) loop. Archaic designs might use analog frequency discriminators for clutter frequency sensing. However, a preferred approach is to use a pair of DFT filters. One filter is centered at some frequency below clutter and the other is centered above. An almost linear error function is obtained by forming the following ratio:

=

Fl ( f ) − Fh ( f ) Fl ( f ) + Fh ( f )

where Fl and Fh are the frequency responses of the lower and higher filters, respectively. The spacing of the filters and the required integration length are dictated by the clutter scenario and geometry. The AFC loop is used to drive this error to 0 where maximum cancellation is achieved. Tuning of the clutter spectrum may be achieved using a variable frequency oscillator as the final local oscillator. That approach is discouraged, because free-running oscillators tend to be noisy and, invariably, produce unwanted intermodulation products within the video bandwidth. A more attractive option is to tune the MTD itself using digital control. As an example, consider the first-order canceler. Without tuning, its output is given by F0 = g(n) − g(n − 1) where g(n) ⫽ cos(2앟fTr) ⫹ j sin(2앟fTr), f is the Doppler frequency, and Tr is the repetition interval. This function may be shown to have a null at zero frequency. Tuning is accomplished by multiplying the delayed pulse by a complex constant. Thus, F0 = g(n) − [cos φ1 + j sin φ1 ] g(n − 1) where ␾1 ⫽ 2앟f 1Tr. It may be shown that the null of this function has been shifted to f 1. When using cancelers having multiple delays, two or more separate nulls may be obtained. This would be useful, for example, in placing sea clutter and rain clutter, which are separated in Doppler, into their own nulls. Of course, the design of a dual-frequency AFC loop does present a challenge. Because the clutter Doppler oscillates as the antenna bearing

751

angle rotates relative to the platform velocity vector, tracking may be aided by introducing antenna angle into the AFC loop. With this aid, the tracking loop can be a simple, first-order implementation which is not required to compensate for Doppler accelerations introduced by rotation. Wilson (12) suggests an adaptive clutter tracking algorithm. Post Discrete Fourier Transform Integration In the usual implementation of the DFT, one electronic module is assigned to each center frequency, and that module processes all range cells within the repetition interval. Current technology allows these modules to be designed as integrated circuits using VLSI. Although each module is very small, the sheer number required can place a severe strain on available signal processor real estate. Also, there is, a cost impact. Thus, there will be a limitation on the number of filters available. Given that the number of filters is limited to Nf , then the spacing between filters is PRF/Nf . However, a particular target Doppler may fall midway between centers. As the integration length increases, the bandwidth of each filter decreases and a loss is imparted to the midway signal. Thus, there is an optimum integration length that maximizes the SNR gain for that signal. It may be shown that the optimum length is 0.74 Nf for an unweighted DFT. In many applications, the available pulses during a beam dwell will far exceed the optimum integration length. For example, with a beamwidth of 5⬚, a scan rate of 60 RPM and a PRI of 30 애s, the number of pulses over the beam is approximately 463. If only 50 filters were implemented, the optimum DFT length would be 37. Thus, there would be 12 or 13 DFT outputs per beam position. The careful designer should try to combine these outputs into a single decision. This combining requires post-DFT integration. The post-DFT integrator might be a Markov chain, which is a simple and effective approach. However, the magnitude integrator has superior performance. In that approach, individual DFT output magnitudes are summed and the result applied to a threshold test. The only disadvantage to this technique is that probability of false alarm is difficult to predict. Apparently, the only viable means for predicting false alarm characteristics is by Monte Carlo simulation. Another aspect of magnitude integration is block versus sliding integration. In the block integrator, several DFT outputs are allowed to accumulate before thresholding. Then, another block is collected. In this approach, it may turn out that both blocks lie on the beam skirts and full benefit of main bean gain is not achieved. A preferred approach is to allow the DFT summation to slide across the beam where a new sum is produced after each DFT on a first-in, first-out basis. This requires more computer memory but ensures that at least one block will encompass the maximum antenna gain. BIBLIOGRAPHY 1. M. I. Skolnik, Introduction to Radar Systems, New York: McGrawHill, 1962, pp. 3–5. 2. M. I. Skolnik (ed.), Radar Handbook, New York: McGraw-Hill, 1970, pp. 9-1, 9-40. 3. M. Schwartz, Information Transmission, Modulation, and Noise, New York: McGraw-Hill, 1959, pp. 282–291.

752

SEISMIC SIGNAL PROCESSING

4. M. Schwartz, Information Transmission, Modulation, and Noise, New York: McGraw-Hill, 1959, pp. 392–413. 5. D. D. Monahan, Average Time Between False Alarms for Phalanx Block I in Low PRF, TM#6-257-851, Pomona, CA: General Dynamics Corporation, 1985. 6. M. I. Skolnik, Introduction to Radar Systems, New York: McGrawHill, 1962, pp. 462–482. 7. F. E. Nathanson, Radar Design Principles, New York: McGrawHill, 1969, pp. 228–275. 8. M. I. Skolnik (ed.), Radar Handbook, New York: McGraw-Hill, 1970, pp. 17-1, 17-60. 9. M. I. Skolnik, Introduction to Radar Systems, New York: McGrawHill, 1962, pp. 294–320. 10. F. J. Harris, On the use of windows for harmonic analysis with the discrete Fourier transform, Proc. IEEE, 66: 51–83, 1978. 11. D. J. Wilson, Signal-to-Noise Ratio Loss in a Combined Clutter Canceler-DFT Signal Processor, IC#5G41.10/003/DJW, Tucson, AZ: Hughes Missile Systems Company, 1994. 12. D. J. Wilson, A Clutter Adaptive Algorithm for the Phalanx Upgrade Search Radar, IC#5G41.10/010/DW, Tucson, AZ: Hughes Missile Systems Company, 1995.

DUSTIN J. WILSON Hughes Missile Systems Company (retired)

SECOND-ORDER CIRCUITS. See TRANSIENT ANALYSIS. SEEBECK EFFECT. See PELTIER EFFECT. SEEBECK EMF. See THERMOCOUPLES. SEGMENTATION, IMAGE. See IMAGE SEGMENTATION. SEISMIC DATA PROCESSING. See GEOPHYSICAL SIGNAL PROCESSING.

01 Aerospace & Electronic Systems

Comments

Content

Sponsor Documents

Recommended