Automotive Systems
Content
Control Engineering Practice 11 (2003) 179–190
Modelling and simulation for mechatronic design in
automotive systems
T. Bertram, F. Bekes, R. Greul, O. Hanke, C. Ha, J. Hilgert, M. Hiller*, O.
.
Ottgen,
P. Opgen-Rhein, M. Torlo, D. Ward
Faculty of Engineering Sciences, Gerhard Mercator University of Duisburg, Institute for Mechatronics and System Dynamics,
Mechatronics Laboratory, Lotharstr. 1, D-47057 Duisburg, Germany
Received 19 October 2001; accepted 17 January 2002
Abstract
This paper gives an overview of current industry based projects in the field of vehicle modelling and simulation for the
mechatronic design of automotive systems. It shows the wide range of applications for analysis and synthesis during the
development process, including vehicle systems, vehicle dynamics, occupant safety, adaptive cruise control, hardware-in-the-loop
and fault tolerant real-time systems. r 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Modelling; Simulation; Control of multibody kinematics and dynamics; Active and passive safety in automotive systems
1. Introduction
Something historic is happening in the automobile
business. It will affect the transportation world in the
same way that the invention of the internal combustion
engine affected personal transportation. The process is
happening overnight and is influencing the whole
industry. It involves an entirely new automobile devel-
opment and manufacturing process, and requires a
fundamental shift in thinking. From thinking of the
car as a mechanical device that carries some electronic
controls to thinking of the car as a mechatronic device
(Fig. 1). This means a device where the mechanical,
electrical, and software parts are fully integrated
(Dickinson, 1996; DesJardin, 1996).
The main driving forces for this shift in thinking are the
expectations of the consumer. Consumers already expect
the same things from their automobiles as they do from
their other consumer electronics products. They expect
safety, security, reliability, ease of operation, comfort,
entertainment, and value for money. Furthermore they
expect what they can get elsewhere, for example, in their
home or in their office, to be available in their
automobiles. The automobile in the mind of the younger
generation is no more than a powerful computer. It is
important to keep in perspective the fact that auto-
mobiles are primarily mechanical products with me-
chanical functionality. Electrical assemblies and the
embedded software are only enabling technologies, and
not the critical vehicle functions themselves. However,
sophisticated functions such as engine management,
traction control, and active vehicle dynamics can only be
implemented today by the judicious combination of
mechatronic technologies.
Among the various current developments in the
electronics field, the trend towards networking existing
and newly developed systems is playing a prominent
role. While linking control systems for active safety has
already been employed for some years, the next step in
this evolution is the integration of systems, aimed at the
user’s wish for increased safety, improved security
systems, reduced power consumption, responsible eco-
logical friendliness, comfort, and growing multimedia
capabilities. Thus, electronic systems that were up until
now essentially autonomous are now growing together.
This process is mainly driven by demand for improved
functionality and the need to limit costs. Extended
system interaction helps to make more intelligent use of
what is already installed and can even simplify present
*Corresponding author. Tel.: +49-203-379-2199; fax: +49-203-379-
4143.
E-mail address: [email protected] (M. Hiller).
0967-0661/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved.
PII: S 0 9 6 7 - 0 6 6 1 ( 0 2 ) 0 0 0 7 6 - X
installations. The traditional example of interaction in
the area of active safety is the link between the traction
control and engine management systems for torque
control.
The engineering of such a new, interconnected system
poses great challenges—in particular for guarantying its
reliability, safety, and acceptance by the car user. The
network has to be set up systematically to achieve
advantages going beyond the sum of the components,
and to avoid mutual disturbance. On top of that, each
network component must be able to work in a wide
variety of configurations where varying contributions
from different sources come together. Therefore, the
complete network must be scalable from a low level of
functionality and cost, via numerous customer-oriented
variants, up to the future state-of-the-art in automotive
electronics. To deal with these challenges, and the long
and complex supply chain associated with them, the
automotive industry has been converging on develop-
ment processes where systematic modelling and simula-
tion play a major role.
This contribution is divided into nine main sec-
tions. Section 2 briefly describes modelling vehicles in
Fasim C++, and how they are modelled as mecha-
tronic systems, incorporating multibody kinematics and
dynamics, hydraulics, controllers, sensors, data manage-
ment systems, and environmental conditions. Fasim C++
is the multibody vehicle simulation package used (and
developed) in the Mechatronics Laboratory at the
University of Duisburg (Hiller, Schuster, & Adamski,
1997). The next six sections illustrate application of
complex vehicle dynamics simulation control (Section
3), rollover simulation (Section 4), crosswind compensa-
tion (Section 5), dynamic headlamp levelling control
(Section 6), semi-autonomous driving (Section 7) and
hardware-in-the-loop simulation with the complex
vehicle model (Section 8) and makes some remarks with
regard to the fault tolerant integration of a decoupled
control system into the vehicle. Section 9 presents some
conclusions.
2. Vehicle dynamics
Development of vehicle controllers requires an
appropriate model of the vehicle dynamics built into a
versatile simulation environment. This simulation en-
vironment has to be able to simulate different vehicle
types or models without any recompilation. The vehicle
model has to have a modular form so that single
component of the vehicle may be exchanged, depending
on the simulation task. Thus, models of the vehicle
dynamics with differing levels of complexity can be
defined covering correspondent physical effects with the
desired accuracy. The modular structure of a vehicle
model in Fasim C++ is shown in Fig. 2 using the
example of a passenger car.
The structure presented does not show the construc-
tion details of the modules, e.g. which kind of front
suspension is used. During initialization this is not
important, because the required information for gen-
erating the equations of motion is part of the modules
and only at the beginning of simulation is it evaluated.
The topology of the vehicle, which describes the
kinematic topology of the individual modules, is shown
in Fig. 3. For reasons of clarity the modules engine
hydraulics (braking system), driver and environment are
not shown. Using this modelling technique it is possible
to decide during runtime which configuration of a
Fig. 1. The automobile as a mechatronic device.
Car Body / Chassis
Front Suspension
Rear Suspension
Driver
wheel wheel
wheel wheel
Engine
Drive Train Hydraulics
Fig. 2. Modular structure of a passenger car in Fasim C++.
T. Bertram et al. / Control Engineering Practice 11 (2003) 179–190 180
vehicle is used without any recompilation of the
program.
Fasim C++ contains a large library of different
vehicle modules such as suspensions, tire models, drive
trains, engines, engine mounts, controllers, sensors,
elasticities, a rigid or flexible car body, several hydraulic
braking systems, a driver and an environment model.
The structure of the modules makes it easy to expand
the library by adding new modules. The equations of
motion are based on D’Alembert’s principle:
X
n
B
i¼1
ðm
i
. r
S
i
À F
i
Þ Á dr
S
i
þ ðH
S
i
’ x
i
þx
i
ÂH
S
i
x
i
À T
S
i
Þ Á d/
i
¼ 0; ð1Þ
where n
B
is the number of mass-endowed bodies, m
i
; H
S
i
are the mass and inertia tensor of body i, . r
S
i
the
acceleration of c.o.g., F
i
; T
S
i
the applied force and
torque, and dr
S
i
; du
i
are the virtual linear and angular
displacement.
Due to the constraints in the system, the virtual
displacements are not independent. To generate the
equations of motion in minimal coordinates the choice
of f independent generalized coordinates q
1
; q
2
; y; q
f
; is
necessary, corresponding to the number of degrees of
freedom in the system. The equations of motion of the
mechanical system in minimal coordinates can then be
written as:
MðqÞ. q þ bðq; ’ qÞ ¼ Qðq; ’ q; tÞ; ð2Þ
where M is the generalized mass matrix, b the general-
ized gyroscopic forces, q the generalized coordinates,
and Q the generalized applied forces.
Applying the principle of kinematic differentials, the
elements of the equations of motion are calculated
expressing partial derivatives using kinematic terms.
Due to the modular structure of the matrices and
vectors, their elements can easily be calculated from the
corresponding modules. For this reason they are
subdivided into an inner sum, inside the module l
considering all its bodies n
B
and in an outer sum
considering all modules n
M
:
M
j; k
¼
X
n
M
l¼1
X
iAI
l
½m
i
#
’ r
ðjÞ
i
Á
#
’ r
ðkÞ
i
þ # x
ðjÞ
i
Á ðH
i
# x
ðkÞ
i
Þ;
b
j
¼
X
n
M
l¼1
X
iAI
l
½m
i
#
’ r
ðjÞ
i
Á
#
. r
i
þ # x
ðjÞ
i
Á ðH
i
#
’ x
i
þx
i
ÂH
i
x
i
Þ;
Q
j
¼
X
n
M
l¼1
X
iAI
l
½
#
’ r
ðjÞ
i
Á F
i
þ # x
ðjÞ
i
Á T
i
: ð3Þ
The pseudo velocities
#
’ r
ðjÞ
i
; # x
ðjÞ
i
and pseudo accelera-
tions
#
. r
i
;
#
’ x
i
are defined as follows (Hiller & Kecske-
m! ethy, 1989):
#
’ r
ðjÞ
i
¼
qr
i
j
;
#
. r
i
¼
X
f
j¼1
X
f
k¼1
q
2
r
i
j
k
’ q
j
’ q
k
;
# o
ðjÞ
i
¼
qo
i
q’ q
j
;
#
’ o
i
¼
X
f
j¼1
qJ
oi
j
’ q’ q
j
; J
oi
¼
qo
i
q’ q
: ð4Þ
3. Control of vehicle dynamics
Numerous reforms have taken place in the sector of
electronic control systems since their introduction in the
eighties. The field of active safety devices reaches from
the classic antilock braking system (ABS) to complex
vehicle dynamic controllers such as electronic stability
program (ESP). The objective of vehicle control systems
is support in situations of non-linear and coupled
operations, which are difficult to handle for the driver.
By means of interference in the acceleration, steering
and braking processes it is possible to maintain stability
and control (Fig. 4).
Simulation is an important tool for the design and
optimization of controllers. Due to its high adaptability
Fasim C++ ensures through its modular structure
the efficient realization and optimization of vehicle
models. Interfaces on the basis of TCP/IP, CORBA and
ActiveX permit the integration of complex control
models, implemented in controller development
software such as Matlab/Simulink (as shown in
Fig. 5).
Fig. 3. Kinematic topology.
T. Bertram et al. / Control Engineering Practice 11 (2003) 179–190 181
With the aid of the vehicles state variables
(a) wheel velocities,
(b) lateral acceleration and
(c) yaw angle
and the driver inputs
1. steering wheel angle,
2. throttle opening and
3. brake position
the brake booster or individual brake actuators can be
triggered under the terms of the control strategy. In
Fig. 6 the simulation results of steering interference are
shown. This open loop manoeuvre clarifies the possibi-
lities of active chassis components.
For this reason comprehensive examination of the
complete system is carried out to allow the implementa-
tion of concepts with regard to the priorities of the
components.
Validation of the simulation results can be obtained
by use of hardware-in-the-loop test benches as well as
driving trials.
4. Rollover simulation
An industrial application of Fasim C++ lies in the
field of passive vehicle safety. Featuring front airbags,
side airbags, seat belt pretensioners and load limiters,
existing restraint systems provide a high level of
protection. Additionally so now that knee and head
protecting side airbags are starting to come onto the
market. For the activation of these protective devices,
comprehensive sensor systems are required which can
react with the appropriate deployment of restraint
systems, taking into account any relevant accident
parameter. For this reason future sensor concepts must
supply information about vehicle stability, approaching
obstacles, vehicle interior conditions, accident type and
crash severity (Gr . osch et al., 1996).
Computer simulation plays an important role in the
development of a rollover detection system (Hiller &
Bardini, 1998). Vehicle dynamics simulation, for exam-
ple, provides the possibility to test in advance various
sensors and algorithms for rollover detection. Further-
more, occupant simulation can be used to establish
trigger times for rollover detection. For occupant
simulation the commercial simulation toolset MADY-
MO (Lupker, 1996) is used. Fasim C++ and Madymo
have been combined to form an application and
development environment for the rollover detection
system from Robert Bosch GmbH (Mehler, Mattes,
Henne, Lang, & Wottreng, 1998). Some special en-
hancements have been made in Fasim C++ for con-
ducting rollover simulations. Firstly the sensor,
including the rollover detection algorithm, was imple-
mented. Thus it was now possible to analyse the
triggering behaviour in any simulated manoeuvre.
Furthermore, it was necessary to enhance the modelling
of the environment. For the simulation of embankment
and ramp manoeuvres it is now possible to configure
surfaces such as those shown in Fig. 7.
The simulation of an embankment rollover is used
here to illustrate the application of these modelling Fig. 4. Effect of ESP.
Fig. 5. Control interface in Fasim C++.
T. Bertram et al. / Control Engineering Practice 11 (2003) 179–190 182
techniques to the analysis of real world vehicle
problems. As leaving the road is statistically the most
likely cause of a rollover situation, it is very important
that this manoeuvre is detected by a rollover protective
system.
Therefore, a full-scale rollover test with a middle class
car has been investigated in detail. As shown in Fig. 8,
very good correlation between the simulation and the
real experiment has been achieved. Only when the car
body hits the ground does the simulation yield incorrect
results, as the contact interactions between the exterior
of the vehicle and the environment have not been
modelled. Since this phase of the rollover is no longer of
importance for rollover detection these errors have been
neglected. When the car body hits the ground rollover
detection must have taken place long ago. The sensor
module that has been implemented in the vehicle is fed
with longitudinal and lateral acceleration and angular
velocity data during simulation by the chassis module,
and returns the trigger signal for controlling the rollover
protective devices. The instant of rollover detection is
visualized using a cone which has been added to the
animation and which appears above the vehicle bonnet
(hood) when the sensor triggers (Fig. 8). With the
validated model it is possible to perform parameter
studies in order to optimize the rollover sensing concept,
and to establish trigger times for rollover protective
devices such as seat belt pretensioners and window
airbags.
As a second example the application of the vehicle
dynamics simulation software for the behaviour when
driving over a ramp is chosen. Typical ramps in reality
are the beginning of crash barriers or amassments of
soil. Fig. 9 shows exemplary a comparison of a middle
class vehicle driving with a velocity of v ¼ 20 m/s one-
sided over a ramp with a height of h ¼ 0:7 m. For the
parameters ‘‘roll angle’’ and ‘‘roll rate’’ good approx-
imation of the real behaviour can be achieved.
5. Crosswind compensation
Automobiles react to crosswinds with reduced direc-
tional stability and a change in their yaw motion. In
particular sport utility vehicles (SUV), whose market
share rose around 30% in 1999, are affected by this
problem. With the help of Fasim C++ the dynamic
0 10 20 30 40 50 60 70 80
0
10
20
30
40
50
60
70
Interference of steering controller
with steering controller
without steering controller
X-coordinate [m]
Y
-
c
o
o
r
d
i
n
a
t
e
[
m
]
vehicle (position and orientation)
v = 35 m/s
= 0.1 rad
x
f
δ
Fig. 6. Simulation results of active chassis management.
b
X
Y
Z
µ=0.45
Street
h
B
H
Ramp
Embankment 2
Embankment 1
µ=0.8
Fig. 7. Example of surface contours used for rollover simulation.
T. Bertram et al. / Control Engineering Practice 11 (2003) 179–190 183
behaviour of the vehicles under crosswind conditions
can be analysed.
While driving, there is a component of the wind in x-
direction v
DW
due to the forward motion of the vehicle
and a component of the wind in lateral direction, the
crosswind v
CW
: The addition of these wind vectors gives
the resulting wind velocity v
r
; which has an angle of
approach t with respect to the vehicle. With a
symmetrical airflow over the vehicle (t ¼ 0) there is a
resisting force (aerodynamic drag) F
x
in the x-direction
and a lifting force (aerodynamic lift) F
z
in the z-
direction. An asymmetrical airflow over the vehicle
(ta0) results in an additional lateral force F
y
on the
vehicle, a roll moment M
x
and a yaw moment M
z
: The
forces and torques due to airflow over the vehicle can be
seen in Fig. 10.
A manoeuvre with the sudden appearance of cross-
wind on the highway is chosen. It is comparable to
driving over a bridge or past a gap in the wind walls.
The vehicle travels with a velocity of 160 km/h (open-
loop manoeuvre). After 10 m a crosswind springs up
from the right-hand side, and acts on the vehicle for
20 m with a velocity v
CW
¼ 100 km/h, which describes a
wind gust that can occur in Germany in autumn or
winter (Hucho, 1998). Fig. 11 displays this manoeuvre
graphically.
The resulting forces cause the car to alter its direction
and yaw to the left. For a typical lane width of 3.75 m
and a vehicle track width of approximately 1.55 m, only
1.1 m is left on both sides of the vehicle before reaching
the edge of the lane. This is often bordered with
kerbstones or crash barriers, or the vehicle could be
put into the path of other road users.
A controller has been systematically designed to
reduce the lateral offset of the vehicle. The control
concept uses the lateral acceleration a
y
and the yaw rate
’
c of the vehicle as inputs for a system that combines two
PI-controllers. These parameters are chosen since they
are already used in the ESP system, and are relatively
easy to access. The output value of this combined
controller, the steering angle d
L
; is the sum of the
outputs of the two controllers.
The comparison of the simulation results (with and
without crosswind compensation) can be seen in Fig. 12.
The driver’s reaction time (between 0.3 and 1.7 s, Bosch,
1993) depends on age, tiredness and driving experience,
Fig. 8. Embankment test.
Fig. 9. Validation results—Ramp manoeuvre.
T. Bertram et al. / Control Engineering Practice 11 (2003) 179–190 184
but 1 s after entering the crosswind area the vehicle
without the controller has already reached the periphery
of the lane. The improvement in the vehicles behaviour
obtained using the controller, in the form of a reduction
in the lateral offset, is obvious. With the controller
implemented in the vehicle, the driver has more time to
react and avoid a potential accident by applying a small
steering correction.
6. Dynamic headlamp levelling control
The approach ‘‘dynamic headlamp levelling system’’
deals with the control of the angle of vehicle headlamps.
To guarantee safe driving in the dark as well, the
headlamp must maximize the illuminated area in front
of the car under every given condition. Along with the
introduction of xenon headlamps that produce a much
higher light intensity than conventional halogen head-
lamps, the development of the dynamic headlamp
levelling system (HLS) has become more and more
important.
Vehicles that are not equipped with a dynamic HLS
will dazzle oncoming traffic while accelerating and will
have a smaller illuminated field, so that the driver’s field
of vision is reduced, while braking. These effects can be
seen in Fig. 13.
The dynamic HLS regulates the motion of the
headlamps relative to the vehicle motion so that the
range is maintained and oncoming traffic is not dazzled
(Fig. 14). To control the headlights it is necessary to
know the dynamic pitch angle of the vehicle. This
represents the disturbance variable. But the pitch angle
is difficult to measure. One possibility is to work with
inductive angle sensors (Thiemann, Stryschik, & Ho-
bein, 1998) that measure the vertical distance between
the car body and the wheel carrier at the front and at the
back axle, and to calculate the pitch angle over the
difference between these two distances and the wheel-
base. This method detects the pitch angle reliably, but it
Fig. 10. Forces and torques due to airflow over a vehicle.
y
x
10 m 20 m
v = 100 km /h
CW
v
1.1 m
3.75 m
1.1 m
Fig. 11. Crosswind manoeuvre.
Fig. 12. Lateral offset with v
DW
¼ 160 km/h and v
CW
¼ 100 km/h,
with and without control unit.
T. Bertram et al. / Control Engineering Practice 11 (2003) 179–190 185
neglects the angle that results from tire deformation.
Therefore, another possibility to calculate the pitch
angle on the basis of an Luenberger observer was
investigated.
The manoeuvre shown here represents acceleration
for a period of time, with subsequent braking. Fig. 15
shows the change in velocity and acceleration with
respect to time for the manoeuvre, as well as the scaled
accelerator pedal position and the scaled braking
pressure.
Fig. 16 shows the reference pitch angle curve calcu-
lated using Fasim C++ (Section 2). The pitch angle
decreases very strongly at the beginning of the accel-
eration process. The gearshifts can be seen as spikes in
the curve. The pitch angle measurement method that
uses inductive angle sensors approximates the reference
angle very well. The deviations can be traced back to the
neglected deformation of the tires. The pitch angle
calculated using the estimation method also approx-
imates the reference angle very well. The curves are in
the area of the acceleration nearly congruent.
The influence of the spring and damper character-
istics, with differing compression and rebound values,
on the dynamic behaviour of the pitch angle would be
interesting for further research. A more exact investiga-
Fig. 13. Vehicle without a dynamic HLS.
Fig. 14. Vehicle with a dynamic HLS.
Fig. 15. Characteristics of a simulated manoeuvre.
Fig. 16. Pitch angle calculated using the three different methods.
T. Bertram et al. / Control Engineering Practice 11 (2003) 179–190 186
tion of both the stationary and the dynamic effects of
incorrect pitch angle detection on the whole dynamic
HLS control loop should also be investigated.
7. Semi-autonomous driving
The road transport system (including automobiles,
buses and trucks) has not yet made significant use of
modern electronic technologies to enhance system
operations. However, the 21st century will make
increasing demands on modern traffic and logistics
technologies, especially in congested urban areas. An
intelligent transportation system (ITS) which contains a
wide range of systems that fully or partially take over
the tasks of driver by using intelligent systems built in
the vehicle, possibly in combination with control from
the transport infrastructure (Fig. 17) is considered to be
a promising development for the more efficient, reliable,
safer and environmentally friendly use of the transport
infrastructure.
Within the framework of this project, strategies and
concepts will be developed which enable the semi-
autonomous driving of a vehicle as a part of an ITS.
This semi-autonomous driving relates to the coordi-
nated, autonomous, steering, acceleration and braking
of a vehicle in order to let it stably follow a given lane.
The development of the vehicle controller will primarily
concentrate on driving in small convoys or ‘‘platoons’’.
A necessary recommendation for the merging of vehicles
into and out of the platoon is the development of a
model for the communication between the vehicle and
the platoon.
Several mathematical tools and experiments will be
established, with the main emphasis on merging into and
out of the platoon. In contrast to existing studies special
attention will be given to
*
robustness of the merging process with regard to
unexpected situations, disturbances, etc.,
*
dynamic vehicle behaviour,
*
trajectory planning for emergency situations.
The trajectory planning is based on B! ezier-splines and
combined with a simplified vehicle model. Simulations
results for an open-loop lane change manoeuvre ISO/
TR 3888 can be seen in Fig. 18.
The integration of this project into the project design
of an automated vehicle integrated control instrument
(DAVINCI), a cooperative with the Delft University of
Technology and the TNO Road-Vehicles Research
Institute, provides a vehicle model that includes a
controller for the semi-autonomous driving.
Fig. 17. ITS infrastructure and communications.
T. Bertram et al. / Control Engineering Practice 11 (2003) 179–190 187
8. Hardware-in-the-loop real-time simulation
Real-time simulation is becoming more and more
important for testing of electronic control unit (ECU)
software in complex mechatronic systems. Efficient and
reliable test and release of ECU software for such
systems cannot be achieved using expensive and time
consuming in-vehicle testing only. Parallel application
of in-vehicle tests, offline simulation and real-time
simulation is essential for adequate software verification
within required cost and time frames (Fig. 19). In our
case real-time simulation is used for testing safety
software in automotive ECUs. It continually checks
input and output signals for plausibility and consistency.
A low cost real-time computer with comparatively low
computing power handles the complete I/O. It is based
on a VME bus system with a Motorola 68040 CPU, a
dedicated real-time operating system (OS9) and applica-
tion specific I/O cards. In the real-time application the
workstation must be able to calculate one simulation
step Dt in less than or equal to real-time. In this case the
workstation is a DEC Alpha 600 5/333 workstation with
333 MHz clock speed. In order to model the dynamics of
the mechanical subsystems of the vehicle with sufficient
accuracy integration steps of o1 ms (and even as small
as 0.1 ms for the hydraulic components) are required.
A hardware-in-the-loop real-time simulation result of
a safety software test for the ESP is exemplary shown in
Fig. 20. The signals lateral acceleration, wheel speed and
baking pressure of the rear wheels are measured during
a braking in a turn manoeuvre. The real-time simulation
is carried out with the same program as is used for the
offline simulation. To ensure that the model is valid for
the complete range of operating conditions the model
has been validated. This means that extensive in-vehicle
measurements were performed and compared with
simulated data. For a rear wheel driven vehicle, the
vehicle model consists of 41 first-order differential
equations. In the offline simulation models with up to
70 first order differential equations are available.
Most innovations in the development of automotive
ECUs for modern cars are based on an iterative process.
In this environment hardware-in-the-loop simulation
offers a wide range of function tests in the laboratory
under close-to-real conditions. Furthermore, it is
possible to implement new algorithms and weight them
Fig. 18. Simulation results for lane change ISO/TR 3888.
Fig. 19. Hardware-in-the-loop concept.
T. Bertram et al. / Control Engineering Practice 11 (2003) 179–190 188
correspondingly with very short development cycles.
The function of an ECU can be examined often only in
the application environment with further controllers
under real-time conditions. Specifically with drive-by-
wire systems, which will increasingly replace the
mechanical and hydraulic systems in the car over the
next few years, it must be possible to guarantee at every
time that in the case of an error in one or more
components, safe stopping or control of the vehicle
remains possible. Consequently, the complete system
must have a fault tolerant design and have systems so
that individual controllers are able to diagnose an error
and to initiate the corresponding countermeasures. In
this case, the communication between the individual
modules occurs via a bus system, which has an error
resistant, redundant design. If a serious disturbance
occurs on the bus, the communications of all remaining
participants must be able to be immediately transferred
onto a backup system. In addition, the failed device(s)
must be able to detach itself from the bus system.
It must be checked continuously whether data
transmitted via the bus system is transmitted correctly
and is still valid during time-critical processes. For
example, in the case of the brakes in a drive-by-wire
system no delay in the resulting braking effect is allowed
to occur because a defective ECU is disturbing the
communications on the bus system. Consequently, every
command has a limited, temporal validity in addition to
the recommendation that it is actually correct. The
communication of the individual components occurs at
a tightly clocked frequency, which assigns a time
window to every element in which communicatios may
occur. One possible solution here is the time-triggered-
protocol (TTP), which evaluates the temporal behaviour
of the system components, as described, during each bus
cycle. A similar procedure is also used in the mobile
digital GSM communication systems (i.e. mobile cellu-
lar telephones). The individual transmission channels
transmit the digital information in a time-multiplexed
fashion in firmly assigned time slots.
The ECUs in motor vehicles today are primarily
independent, and only transmit diagnostic functions or
status information via a bus system to the outside world.
To exclude a possible system failure through erroneous
communications with the sensors and actuators, per-
ipherals and controllers are often integrated into one
control unit (i.e. ABS and ESP). If a communication
system has the described features, the reuse of sensors
and actuators for functions of a similar type is also
conceivable (Fig. 21). An ECU would consequently be
able to be reduced to the actual micro controller and the
corresponding communication hardware. This would
then be similar to a distributed computer system where,
in the case of the disturbance of single device, another
undertakes its function, switching from a safety
irrelevant functions such as the air conditioning to
controlling the brakes or some other important function
that has failed.
9. Conclusions
This paper gives an overview of current industry
based projects in the field of vehicle modelling and
simulation for the mechatronic design of automotive
systems. It shows the wide range of applications for
analysis and synthesis during the development process,
including vehicle systems, vehicle dynamics, occupant
safety, semi-autonomous driving and hardware-in-the-
loop and fault tolerant real-time systems.
The object-oriented design of the simulation environ-
ment Fasim C++ allows easy adaptation of different
extensions of the vehicle model. On the one hand this
leads to a remarkable variety of the vehicles that can be
simulated, while on the other hand, the extension of the
Fig. 21. Decoupled control system.
Fig. 20. Real-time simulation results.
T. Bertram et al. / Control Engineering Practice 11 (2003) 179–190 189
modular vehicle model is easy to manage. Additional
mechanical or non-mechanical components (e.g. control-
lers, sensors) can be easily appended. The vehicle model
described has been implemented for the development of a
rollover protective system, vehicle dynamics control
systems, and for hardware-in-the-loop simulation.
Acknowledgements
The work presented in this paper is supported by
Robert Bosch GmbH, Stuttgart (Germany), Hella KG
Hueck & Co., Lippstadt (Germany) and Ford Werke
AG, Cologne (Germany).
The project ‘‘Semi-autonomous driving’’ is supported
by the Ministry of Schools, Science and Research of the
state of Nordrhein-Westfalen (Germany).
References
Bosch Automotive Handbook (1993). Robert Bosch GmbH, (3rd ed.)
Stuttgart, Germany.
DesJardin, L. (1996). A day in the life of mechatronic engineers 10
years from now. SAE International Congress and Exposition.
Detroit/Michigan, USA, SAE96C038.
Dickinson, G. W. (1996). Cars and trucks: The ultimate consumer
electronics products. Automotive News World Congress, Detroit/
Michigan, USA.
Gr. osch, L., Mattes, B., & Schramm, D. (1996). Smart Restraints.
Airbag 2000, Second International Conference. Karlsruhe, Ger-
many.
Hiller, M., & Bardini, R. (1998). Vehicle and occupant dynamics
simulation—important tools in the development of sensor concepts
for the control of restraint systems. Eighth German–Japanese
Seminar on ‘‘Nonlinear Problems in Dynamical Systems—Theory
and Applications’’. Kobe, Japan.
Hiller, M., & Kecskem! ethy, A. (1989). Equations of motion of
complex multibody systems using kinematical differentials. Proc.
9th Symposium on Engineering Applications of Mechanics, (13)
113–121.
Hiller, M., Schuster, C., & Adamski, D. (1997). Fasim C++—a
versatile developing environment for vehicle dynamics simulation.
International Journal of Crashworthiness, 2(1), 91–105.
Hucho, W.-H. (1998). Aerodynamics of road vehicles: from fluid
mechanics to vehicle engineering (4th Ed). Warrendale, PA, USA:
Society of Automotive Engineers.
Lupker, H.A. (1996). MADYMO, a Versatile Tool for Vehicle Safety
Analyses. Sixth International MADYMO Users’ Meeting. Amster-
dam, Netherlands.
Mehler, G., Mattes, B., Henne, M., Lang, H.-P., & Wottreng, W.
(1998). Rollover sensing (ROSE). Second International Conference
on Advanced Microsystems for Automotive Applications. Berlin,
Germany.
Thiemann, M., Stryschik, D., & Hobein, D. (1998). New concepts for
automatic headlamp levelling based on an innovative, inductive
angle of rotation sensor. ATZ Worldwide, (11) 15–18.
T. Bertram et al. / Control Engineering Practice 11 (2003) 179–190 190
