MRA Vehicle Dynamics Simulation-Matlab ® /Simulink ®

Author(s): Edward M. Kasprzak and Douglas L. Milliken

Source: SAE Transactions, Vol. 109, Section 6: JOURNAL OF PASSENGER CARS:
MECHANICAL SYSTEMS JOURNAL (2000), pp. 2043-2050
Published by: SAE International
Stable URL: https://www.jstor.org/stable/44687037
Accessed: 26-04-2019 07:53 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

SAE International is collaborating with JSTOR to digitize, preserve and extend access to SAE
Transactions

This content downloaded from 85.115.52.202 on Fri, 26 Apr 2019 07:53:38 UTC
All use subject to https://about.jstor.org/terms
 2000-01-1624

MRA Vehicle Dynamics Simulation-Matlab®/Simulink

Edward M. Kasprzak and Douglas L. Milliken

Milliken Research Associates, Inc.

Copyright © 2000 Society of Automotive Engineers, Inc.

ABSTRACT augmentation and variable stability aircraft. An excellent

review of the development of active systems in aircraft is
Milliken Research Associates has developed a new given by one of the pioneers in Reference [1]. The
control of all modern large and high speed aircraft is
simulation tool, named Vehicle Dynamics Simulation-
Matlab/Simulink (VD-M/S). Produced for the heavily dependent on active control systems. Cornell
Aeronautical
government's Variable Dynamic Testbed Vehicle (VDTV), Laboratories (CAL/ Calspan) has been a
VD-M/S is an 18 degree-of-freedom simulation leader in the variable stability aircraft field since the early
programmed in the Matlab/Simulink environment. It 1 950's.
contains a detailed non-linear tire model, kinematic and
The CAL aircraft experience was transferred to the
compliance effects, aerodynamic loadings, etc. as do
development of the first variable stability automobile on
MRA's other simulation programs. Unique to VD-M/S is
contract from General Motors in 1956. This steer-by-wire
its development from Day One as a simulation catered to
automobile used an analog computer and electro-
the inclusion and exploration of active systems within the
hydraulic servo control of front wheel steering and
vehicle.
steering wheel torque, see Reference [2], Subsequently
(c. 1966), GM developed a variable response vehicle with
INTRODUCTION
front and rear steer-by-wire, [3].

Another more recent approach to automotive active

Aided by Simulink's graphical programming interface, VD-
M/S is readily modified to accept new segments of control is the Lotus active ride system [4 & 5] with control
program code representing active systems. VD-M/Sover
is both unsprung and sprung mass motions (1980's).
especially well-suited as a platform for testing new controlBy sensing motion variables and controlling diagonal
algorithms for any system attempting to control the wheel loads, this system was used to apply stabilizing
yaw moments to the vehicle at high lateral acceleration.
dynamics of the vehicle. It was used extensively in the
The Lotus system has been followed by a variety of low-
design of the VDTV which has a number of active control
bandwidth and semi-active ride systems which are
systems, described below. VD-M/S has been used to
generally aimed at sprung mass (body) control only;
determine control gains and algorithms which allow the
VDTV to emulate the handling characteristics of manysome of these systems are now in production.
different types of vehicles. In the VDTV a variable Anti-lock brake systems (also originally from aircraft) are
stability automobile has been developed in the lineage the
of core of a third approach to automotive active control.
variable stability airplanes. Effective use of the VDTVBy keeping the tires rolling, the inherent yaw damping of
relies on preliminary investigations with VD-M/S. the tires is preserved (when compared to a locked wheel)
The VDTV project is presented as a detailed example ofand driver steering control is retained. Recent
VD-M/S simulation, validation and use. The capabilitiesdevelopments in directional stability systems use active
of VDTV, and that of VD-M/S to simulate them, are control of individual wheel brakes to apply yawing
explored. Specific ways of extending VD-M/S are also moments to the vehicle under limit handling conditions.
given. Current developments are aimed at combining various
elements of these systems into an integrated active
A BRIEF HISTORY OF ACTIVE SYSTEMS
stability control system.

A general definition of an active vehicle control system is

THE VARIABLE DYNAMIC TESTBED VEHICLE
one in which vehicle motions are sensed and processed
(VDTV)
via control algorithms into signals used to drive actuators,
thereby modifying overall vehicle motion. Among the first
Toward the end of the Lotus Active period (c. 1990) an
active vehicle systems were aircraft stability 2043 on-road simulator was conceived with the ability to mimic

This content downloaded from 85.115.52.202 on Fri, 26 Apr 2019 07:53:38 UTC
All use subject to https://about.jstor.org/terms
the handling characteristics of a variety of vehicles. With preliminary analysis. This, along with the VDTV
an on-road simulator, the visual scene and motion cues specification of a dynamic computer simulation to be
are real (as compared to a fixed-base simulator). Many delivered with the vehicle, led to the development of a
research uses were foreseen, including validation testing Matlab/Simulink-based computer model which was then
of fixed-base simulators, testing new vehicle control tailored specifically to the active systems of the VDTV.
strategies and researching the influence of human The decision to create a new program rather than use
factors on handling qualities. Potentially such a vehicle one of MRA's existing vehicle simulations was driven by a
could be used in a similar fashion as variable stability desire to use the Matlab/Simulink environment.
aircraft have been used for 50 years - developing aircraft
The top level Simulink diagram is shown in Figure 1.
handling standards and supplying hands-on stability and
Execution of the simulation takes place through the
control training for test pilots.
Matlab window. Vehicle, tire, control system and program
At the same time, there was also a desire within the control data are supplied to the program through the
National Highway Traffic and Safety Administration workspace in the form of Matlab M-files.
(NHTSA) to keep abreast of developments in electronic Use of the Matlab/Simulink environment was attractive for
and computer controls applied to automobiles. The Jet
many reasons. ERIM had a large amount of in-house
Propulsion Laboratory (JPL) was contracted to study the
experience using Matlab/Simulink and the two MRA
concept of an on-road simulator [6] and JPL generated a
associates principally involved in the VDTV program were
request for proposals for NHTSA in early 1996. The
also Matlab users. Simulink's block diagram
project, termed the Variable Dynamic Testbed Vehicle or
programming technique facilitated the programming of
VDTV, was to be a platform for testing new technologies,
various active systems. Since Matlab and the VDTV's
a tool for validation of the National Advanced Driving
onboard DSpace hardware and software are readily
Simulator (NADS) and an advanced tool for use by
compatible, control algorithms developed within the
various researchers in driver behavior (human factors).
simulation would easily port to the vehicle and, if desired,
The VDTV contract was won by Environmental Research the vehicle model could be integrated with the vehicle
Institute of Michigan (ERIM) in mid-1996. The final systems to run in realtime as part of the control system.
control system specification incorporates generalized
Development of the MRA Vehicle Dynamics Simulation-
steer-by-wire (front and rear steering, and, steering
Matlab/Simulink (VD-M/S) commenced and resulted in
wheel feel servo), throttle-by-wire, brake-by-wire, active
an 18 degree of freedom, lumped parameter model.
anti-roll bars and active dampers. An onboard data
Breakdown of the principal degrees of freedom are as
system is capable of recording vehicle motions as well as
follows: six for the sprung mass, three for each of the
driver actions. Also included are extensive safety
front and rear unsprung masses, four wheel spin and two
systems, including mechanical redundancy. For
steer degrees of freedom. The three degrees of freedom
example, a critical problem with the steer-by-wire system
for the unsprung masses are ride, roll and lateral
causes a clutch to re-engage the steering column,
displacement (for each end of the vehicle), based on the
returning the vehicle to conventional steering.
equations presented in [7]; this formulation allows one set
Additionally, the VDTV specification included the delivery of equations to model both independent and axle
of a comprehensive dynamic computer simulation of the suspensions. The VDTV has four-wheel independent
vehicle, including complete models of the various active suspension with MacPherson struts on the front and
control systems. Initially, the simulation was seen as a three-link struts on the rear.
tool to develop desired VDTV configurations (vehicle
The core equations used in VD-M/S are available from
response characteristics) in advance of actual test work
the government through the Department of
with the car. In practice the simulation proved to have a
Transportation [7]. These equations had been developed
much broader influence on the VDTV project.
by System Technologies, Inc. and form the basis of STI's
As built, the VDTV is a 1995 Ford Taurus SHO reworked "VDANL" vehicle simulation. This simulation had been
to accept high performance tires, a safety roll cage and previously validated by NHTSA as part of the NADS
numerous other modifications. Vehicle motion sensors, project. Mathematical representation of a "passive"
computer and data acquisition equipment and electric & vehicle by these equations includes a representative
hydraulic servos have been installed. The VDTV number of vehicle details such as vehicle masses and
therefore is purely an experimental/research vehicle and inertias, parametric suspension details, some kinematic
is intended for proving ground use only. and compliance effects and a non-linear tire model. For
the VDTV work ERIM chose to use STI's tire model. This
VD-M/S SIMULATION model is based on "Calspan coefficients" with a number
of additions and is described in [7], Other tire models
Early in the VDTV program, ERIM requested Millikencan be easily substituted into VD-M/S.
Research Associates (MRA) to perform the required

2044

This content downloaded from 85.115.52.202 on Fri, 26 Apr 2019 07:53:38 UTC
All use subject to https://about.jstor.org/terms
 ©
i

L_J
Aero Forces
& Moments ^ _JV.

i - i I

- 5T1
Q
- -* Tire Motions

Driving/Braking (

Torques _^|

ĘI
Control Steering & Unsprung Mass Integrator
System Suspension Motions

World 'a i'

Coordinates | *

§
Figure 1 . To

While VALIDATION OF THE VD-M/S MODEL

these equ
passive vehicle model, MRA made additions and
revisions to the originai equations. These included Validation work was carried out against the original
adding modeling of static toe and camber, bump stops VDANL model and against vehicle tests performed by the
and rebound stops, lateral force inclination angle Vehicle Research and Test Center (VRTC) in Ohio.
compliance, aerodynamic downforce, a new steering Specific proving ground tests were run with a stock 1994
system model and other parameters. Furthermore, a Ford Taurus and on-board data was recorded.
choice of inputs to the model based on torque, speed or
VDANL model runs were then conducted which
acceleration for tractive/braking effort and steering wheel
simulated the same maneuvers. Inclusion of the VDANL
input, path following or a simple driver model for steering
model in the validation process was seen as an important
input were implemented.
benchmark for two reasons. First, the core equations of
To model the VDTV, alterations to the original passive VD-M/S are a revised version of the VDANL equations,
equations reflecting the active systems were required. so similarities between model behaviors should exist.
VDTV uses four wheel steer-by-wire, so the entire rear- Second, VDANL has an established history of use in
wheel steering system was modeled along with the government work and is a validated simulation.
effects of rear-wheel steering on the vehicle dynamics.
Initially VD-M/S was run with all active systems "off".
Algorithms for the electronically controlled shock
That is, the gain programming was set so that the active
absorbers were included, replacing the "passive" shock
steering system and active anti-roll bars acted as if they
absorber model. Control algorithms for front-wheel steer,
rear-wheel steer and for the front and rear anti-roll bars were passive elements. Data on the tires used with the
1994 Taurus was applied, and the vehicle parameters of
were inserted in the simulation. These will be explored in
the next section. the 1994 Taurus (not the VDTV) were used. By running
VD-M/S in this configuration the correlation with VDANL
As the project advanced and data from the actual car and the proving ground test, both of which are entirely
became available, smaller additions and revisions were passive, could be examined.
made to reflect the rate-limited or power-limited
performance of some of the active subsystems.
Figures 2 through 5 show sample plots from this
correlation. In Figures 2 and 3 the maneuver is a

2045

This content downloaded from 85.115.52.202 on Fri, 26 Apr 2019 07:53:38 UTC
All use subject to https://about.jstor.org/terms
steering wheel step steer from 0° to 45° initiated at 25
and 50 mph, respectively. The plot is of vehicle lateral
acceleration. In Figure 2 the initial part of the step is not
shown to give greater detail in the settling region. The
experimental line, taken from on-board data recorders,
shows more oscillation than the two simulations, perhaps
due to road roughness. Both simulations have nearly
identical rise time and overshoot. VD-M/S matches the
steady-state lateral acceleration almost exactly, while
VDANL is very close. The on-board data shows a higher
peak lateral acceleration (for 100 ms) than does either
simuMion. This has not been investigated, but could be
due to transient roll angle contamination of the
accelerometer signal.

Figures 4 and 5 show the results of a step steer

Figure 3. Validation plot showing lateral acceleration
maneuver from 0° to 33° initiated at 25 and 50 mph,
rise, overshoot and steady-state after a step
respectively. The plots presented are of vehicle yaw rate.
steer at 50 MPH.
In these plots all three traces show nearly identical yaw
acceleration early in the step steer. VDANL matches the
overall shape of the experimental curve better than VD-
M/S, although the peak and the time at which steady-
state is reached is less than the experimental data. VD-
M/S has a slightly different shape than the experimental
data, but the peak and steady-state values are more in
line with the proving ground test data.

Figure 4. Validation plots showing yaw rate rise time,

overshoot and steady-state from a step steer
at 25 MPH.

Figure 2. Validation plot showing lateral acceleration

rise, overshoot and steady-state after a step
steer at 25 MPH.

The above plots are a sampling of several dozen

maneuvers and responses that were studied and are
representative of the validation. From these tests it can
be concluded that VD-M/S is basically sound. It agrees
closely with proving ground tests and with the validated
VDANL simulation. Differences between the two
simulations can be attributed to alterations made in the
VD-M/S equations. Recall that even though the core VD-
M/S equations are based on those in VDANL, MRA made Figure 5. Validation plots showing yaw rate rise time,
adjustments to these equations based on their overshoot and steady-state from a step steer
at 50 MPH.
experience in vehicle simulation.

2046

This content downloaded from 85.115.52.202 on Fri, 26 Apr 2019 07:53:38 UTC
All use subject to https://about.jstor.org/terms
Future validation work may take advantage of some gain VD-M/S
programming that has been developed to emulate the specific
dynamics of the 1994 Taurus using the VDTV. Even control
though the VDTV is based on a Taurus, significant decision
alterations to the original vehicle created large use of r
differences in vehicle mass, inertias, and compliances. followin
VDTV also uses quite different tires. As a result, the gain to be si
programming required to emulate the dynamics of a 1 994 vehicle p
Taurus is very similar to that required to emulate any A succession of model runs were made to determine
other vehicle. Once the VDTV is fully functional, VD-M/S
what vehicle hardware would be required to meet thes
(including the active system models) will be validated
requirements. The initial work involved choosing ba
against the real vehicle.
algorithms for front and rear steer as well as front an
rear active anti-roll bars to give reasonable respons
VD-M/S IN THE VDTV DESIGN PROCESS
characteristics while also meeting the response metr
requirements. VD-M/S runs showed that these activ
Originally, VD-M/S was conceived as a tool to determine
systems, combined with electronically controlled shoc
specific VDTV configurations. In practice, the role of VD-
absorbers, were sufficient when appropriate control gai
M/S expanded to include analysis of key VDTV design
were chosen.
issues before the vehicle was built. This section
describes the use of VD-M/S as a case study in A the
large number of control gains on sensed variables were
considered
integration of vehicle simulation in the vehicle design for the active system algorithms. Control
process. schemes based on a knowledge of vehicle dynamics
were tested with VD-M/S to determine their viability. In
Preliminary VDTV analysis at MRA was conducted using the end, the active anti-roll bars were modeled as in
an MRA in-house vehicle model. Once VD-M/S was
Equations 1 and 2:
operating, it was tailored to the VDTV through the
additions of active systems and other subsystems Both
as the front and rear anti-roll bars are commanded by
control
noted in the previous section. This specific version of gains acting on lateral acceleration, yaw rate, roll
VD-M/S is named VDTV-Sim. angle, roll rate and roll acceleration. The anti-roll bar
algorithms are primarily responsible for determining
One of the primary objectives of the VDTV is to "emulate"
vehicle roll angle and roll damping, as well as influencing
the principal dynamics of the U.S. passenger car fleet.
the understeer gradient.
That is, the VDTV, through the use of on-board
computers, control algorithms and active subsystems is
able to exhibit specific response characteristics Qfc = KtFPHI^s + KTFAYays ^ ^ ^
+ Ktfpdox(1)s + KXFPc|)s + KXFRr
depending on the programmed control gains. There is a
significant variation among current and recently built Qrc = KtRPHI^s + KTRAYays . .
passenger cars in the commonly used response metrics. + KTRPDOT^s +KTrp<{>s +KTrrI"
Table 1 summarizes data on these metrics found in
where:
references [8], [9] and [10], representing the vehicle fleet
in the United States during the last decade. The VDTV
sought to emulate the generalized response • (ļ)s = Sprung mass roll angle
characteristics throughout these ranges.
• <ļ)s = Sprung mass roll rate

Table 1 . Performance variations across the US Fleet • <ļ)s = Sprung mass roll acceleration

Response US Fleet US Fleet VDTV • ays = Sprung mass lateral acceleration

Metric Minimum Maximum Emulation
• r = Vehicle yaw rate
Range
Understeer 0.5 6.5 -4 to +1 3 • Qfc Qrc - Commanded Front and Rear Roll
Gradient at 0.2 g Torque
(deg/g) • K** = Control Gains
Yaw Rate 0.32 0.69 0.2 to 0.9
Response Time Note that it is not enough for the active anti-roll bar to
(sec) simulate the stiffness of anti-roll bars in the car being
emulated. Due to inherent differences between the
Maximum Lateral 0.59 0.91 0.95 max.
VDTV and the vehicle being emulated (such as veh
Acceleration (g)
mass, inertia, tires, etc.) the active anti-roll bar is used
Roll Gradient 1.5 1 1 .3 1 to 1 2
respond to vehicle motions and emulate the dynamic
(deg/g) I 2047 the vehicle of interest. Thus, a more complicated g

This content downloaded from 85.115.52.202 on Fri, 26 Apr 2019 07:53:38 UTC
All use subject to https://about.jstor.org/terms
programming scheme is needed than one which output of the front control algorithm. The steer
responds solely to sprung mass roll angle as a passive algorithms serve principally to limit the maximum lateral
bar would. acceleration, and to force the sideslip gradient, lateral
acceleration rise time and yaw rate overshoot of the
The steering system algorithms contain additional
VDTV to match the vehicle being emulated. Similar to
sensed variables. The algorithms are presented in
the role of the active anti-roll bars, it is not enough to use
Equations 3 and 4:
a control algorithm to replicate a steering ratio. Along
with the other active systems, the steer control algorithms
Sfc = KFAYayt +KFRv(r/u) + KFpß and gains determine the VDTV's dynamic response.
+ Kf^ + Kppß + Kpji + Kj^rļ) (3)
Developing a control strategy to limit the vehicle's
+ Kp^ + Kf8sw (l - K Aa^t8sw
j maximum lateral acceleration was extensively studied
using VD-M/S. The VDTV is capable of attaining a lateral
5rc = KRAYayt + KRRV(r/u) + KRßß acceleration of 0.95g. Most vehicles in the US fleet do
+ KR(ļ,<|) + KRßß + Kj^ř + KR^(ļ) not reach this level of performance. To properly emulate
+ KR^ + KR5sw(l-KRAa^)ôSw (4) the lateral acceleration characteristics of a vehicle the

+ KRSFcSfc VDTV's lateral acceleration needed to be artificially

limited.
where:
Yet, it needed to be limited in such a way that the VDTV,
• ayt = Lateral acceleration at the total vehicle center operating below its performance limit, behaved as if it
of gravity was at the limit of cornering ability. Furthermore, the
• u = Vehicle forward velocity same steer-by-wire algorithms to be used to artificially
limit the lateral acceleration also had to properly
• ß = Vehicle sideslip angle represent other dynamic performance metrics such as
yaw rate overshoot and sideslip gradient. Repeated VD-
• ß= Vehicle sideslip rate M/S runs were conducted to determine steer control

• ř = Vehicle yaw acceleration algorithms which allowed all goals to be met. The result
is an algorithm which contains a term using the square of
• ^sw - Steer angle at steering wheel lateral acceleration multiplied by steering wheel angle.
• 8Fc, ôRc = Front and rear commanded steer angles This term dominates at high lateral accelerations and is
used to limit the lateral acceleration of the VDTV to match
and all other symbols are defined as before. The Simulink the vehicle being emulated.
block diagram for the front steer algorithm is shown in
Figure 4. The "f(u)" block contains Equation 3. As the VDTV was being constructed, several physical
limitations constrained the limits of some of the
subsystems. It was initially unknown if the stee
O" DeltaSW system would be rate limited or power limited. When
DeltaSW
steering system was determined to be rate limited
(ae9ree5) E> <ydd„> » constraint was added to the VD-M/S model and its effect

I [Psiaotļ^ <P3id0l£ Mux _^i f(uļ i on emulation capabilities determined. This rate limit was
sufficiently high to have only a small impact.
- ►
The rear steering subsystem was originally planned to
Diā>^pīĒ allow ±6° of steer angle. As produced, only ±3° of steer
Commanded DeltaWF
angle is obtainable. VD-M/S runs were conducted with
<Betadoi> * Steei^gle the range of steer angle reduced to the new value and
(radians)

DEââ5> <Psiddoļ> » the performance reevaluated. This constraint took away

some of the VDTV's ability to emulate very short yaw rate
DiĒāāi> <sPhid01> ► rise times or large sideslip angles, although the
ļ[sPhiddot>
performance range was still sufficient to include the
range of the US fleet.
*

Front Steer
Further issues including frequency response, bandwidth
Figure 6. Front Steer Algorithm, Simulink Block and system lag were also considered.

With the numerous modifications to the original Taurus

The steering system control algorithms are based onincluding the addition of high-performance tires-
lateral acceleration, yaw rate, yaw acceleration, speed,
increased vehicle weight, came concerns about
roll angle, roll rate, roll acceleration, sideslip, sideslip rate
rollover. VD-M/S results determined that the VDTV was
and steering wheel position gains. The rear wheel not likely to rollover in an "untripped" condition-one in
steering algorithm also contains a control gain on the
which a sudden change in road level or an impact with
2048

This content downloaded from 85.115.52.202 on Fri, 26 Apr 2019 07:53:38 UTC
All use subject to https://about.jstor.org/terms
another object did not occur. Thus, additional control CONCLUSION
gains and/or algorithms to prevent situations leading to
rollover were deemed unnecessary. While originally developed for use with the Variable
Dynamic Testbed Vehicle (VDTV), Milliken Research
Once the VDTV was completed, vehicle inputs to the VD-
Associates' VD-M/S is suitable for the simulation of
M/S model were updated with final vehicle
vehicle dynamic performance including the effects of a
measurements. Mass and inertia properties were based
on tests at SEA in Columbus, Ohio. Kinematic and wide variety of active control systems. This new model
incorporates a combination of features including:
compliance testing took place on the Anthony Best
Dynamics SPMM machine at the Goodyear Tire & • Lumped Parameter vehicle description (as opposed
Rubber Company, Akron Tech. Center. Tire data, to components in multi-body models) for quick model
provided by Goodyear early in the project, was retained. development. Real-time operation for hardware in
The result is a customized simulation of the VDTV which the loop simulation.
is used in conjunction with the vehicle. Vehicle tests are • Validated against instrumented vehicle test and
VDANL.
preceded by simulation runs on VD-M/S for a range of
experiments. A similar process can be followed using • Programmed in the block diagram environment of
VD-M/S to simulate a wide variety of vehicle active Simulink for ease of future development.
systems during vehicle concept, design and development • MRA's extensive experience in vehicle modeling and
stages. active control systems - dating back to the 1950's.

APPLICATIONS OF VD-M/S WITH VDTV It is available for use by both government and industry,
and is thus a good example of the spin-offs that can be
generated by government sponsored research.
Prior to conducting research with the VDTV at the
proving ground, VD-M/S will be run to establish the effect
DEDICATION
of proposed control gains on vehicle performance.
Depending on the goal of the research-human factors,
emulation of a particular type of vehicle, determiningThis paper is dedicated to the memory of Dr. Hugo
handling standards, etc.-the task is to make the VDTVRadt, Jr., who died suddenly in April 1999. At the
respond with the desired vehicle dynamics. The process beginning of his career, at Cornell Aeronautical
of gain programming is not trivial. By working with VD-M/Laboratory, Hugo was involved in the analysis of some of
the
S to develop control gains before using the vehicle itself, earliest variable stability aircraft and automobiles.
productivity of the VDTV can be maximized. This included major contributions to the development of
the first variable-stability, servo-controlled on-road
The use of Matlab/Simulink provides a self-contained
simulator in the mid 1950's. After an outstanding career
analysis, simulation and post-processing environment. inIt vehicle dynamics analysis at CAL (later Calspan),
also makes VD-M/S readily customizable-VDTV is only Hugo taught at the University at Buffalo and also became
one example where this simulation may be effectivelyan MRA Associate. In Hugo's last project at MRA he had
used. The block programming structure lends itself to
primary responsibility for the system analysis, control
intuitive model modifications, including other active system definition and algorithm development for the
systems. With motion variables readily available in theVDTV, including contributions to the development of the
model, a control algorithm or active system programmed VD-M/S model described in this paper.
in a Simulink block can be added to VD-M/S with a
minimum of difficulty. Some possible applications include Hugo wrote a number of SAE papers over the years and
simulation of anti-lock braking algorithms, yaw control, was a long-standing member of the SAE Vehicle
Dynamics Committee. He would (in all likelihood) have
traction control, ride-leveling., and a variety of intelligent
been a member of the Historical Panel at this first SAE
vehicle systems such as radar coupled to vehicle speed
Automotive Dynamics and Stability Conference.
control. Existing or proprietary blocks modeling such
systems can be "plugged-in" to the VD-M/S vehicle
ACKNOWLEDGMENTS
model, simulating the influence of each system on overall
vehicle performance.
The original VDANL equations were developed by
Specific blocks can be compiled and run in realtime for
Systems Technology, Inc. In our work on VDTV and the
hardware-in-the-loop development of new vehicle
Matlab/Simulink model presented herein, the authors
systems. The need to use a vehicle in the development
received generous cooperation from Ted Rosenthal &
cycle is thus minimized. The entire VD-M/S simulation Jeff Chrstos.
can also be compiled to run realtime in a vehicle. This
could be done to study or implement advanced control of Matlab® and Simulink® are registered trademarks of The
a vehicle through "model following" control techniques. Math Works, Inc.

2049

This content downloaded from 85.115.52.202 on Fri, 26 Apr 2019 07:53:38 UTC
All use subject to https://about.jstor.org/terms
REFERENCES 8. Kunkel, Dennis T. and Ronald L. Leffert, "Objective
Directional Response Testing". General Motors
Proving Ground report, May 8, 1987.
1 . Breuhaus, W. O. "The Variable Stability Airplane from
a Historical Perspective", American Aviation 9. Heydinger, Gary S., "Vehicle Dynamics Simulation
Historical Society Journal, Volume 36, No. 1, pages and Metric Computation for Comparison with
30-55, Spring 1991. Accident Data". Transportation Research Center,
2. Segel, L. and R. T. Bundorf, "The Variable Stability Inc., DOT HS 807 828, March 1991 .
Automobile", SAE SP-275, 1965. Also available as 10. Ford Motor Company. Internal report generated by
two papers, "Concept and Design", SAE 650658 and the Advanced Vehicle Technology group and
"The Use of a Variable-Stability Vehicle in Handling released to the VDTV project under non-disclosure
Research", SAE 650659. agreement. Includes response metric data on 27
3. McKenna, K., "A Variable Response Vehicle - vehicles from model years 1993-96.
Description and Applications", Joint Automatic 11. Kasprzak, E. M., "VDTV-Sim User's Manual", Milliken
Control Conference at U. of Texas, June 1974. Also, Research Associates, 1 998.
GMPG Engineering Publication No. 5665.
4. Wright, P. G. and D. A. Williams, "Applications of CONTACT
Active Suspension to High Performance Road
Vehicles", Proceedings of the Institute of Mechanical Edward M. Kasprzak & Douglas L. Milliken
Engineers, London, C239/84, 1984. Milliken Research Associates, Inc.
5. Milliken, W. F., "Active Suspension", SAE 880799, 245 Brompton Road
1988. Williamsville, NY 14221-5942 USA
6. Marriott, A. T., "The Variable Dynamic Testbed e-mail: [email protected]
Vehicle", SAE 950032, 1995.
7. Allen, R. W., et al, "Vehicle Dynamic Stability and
Rollover", DOT HS 807 956, June 1 992.

2050

This content downloaded from 85.115.52.202 on Fri, 26 Apr 2019 07:53:38 UTC
All use subject to https://about.jstor.org/terms