Proceedings of the 1998 IEEE

International Conference on Robotics & Automation

Leuven, Belgium * May 1998

Simulator as a Virtual eality Tool

Woon-Sung Lee, Jung-Ha Kim, and Jim-Hee Cho
Kookmin University

Contact address :
Dept. of Automotive Engineering, Kookmin University
86 1- 1 Chungnung-dong, Sungbuk-g,u,
Seoul, 136-702, KOREA
Phone:+82-2-9 10-47 12, Fax:+82-2-9 10-47 1 8
E-mail: [email protected]

ABSTRACT
Driving simulators are used effectively for vehicle system development, human factor study, and other
purposes by enabling to reproduce actual driving conditions in a safe and tightly controlled environment.
This paper describes a driving simulator developed for design and evaluation of full-scale driving
simulators and for driver-vehicle interaction study. The simulator consists of a real-time vehicle
simulation system, a visual and audio system, a motion system, a control force loading system, and an
experiment console. The real-time vehicle simulation system supervises overall operation of the
simulator and also simulates a'ynamic motion of realistic vehicle models in real-time. The economical
visual system generates high fidelity driving scenes that are displayed on a screen by a projector. The
motion system generates realistic motion cue using a six degree-of-freedom Stewart platform driven
hydraulically. The control force loading system acts as an interface between a driver and the simulator.
The experiment console monitors the status of the simulator in opera,tion and also collects and manages
experimental data.

Keywords: Driving Simulator, Real-Time Vehicle Simulation, VisuaVAudio System, Motion

System, Control Force Loading System, System Integration

INTRODUCTION advent of computer technologies, Daimler-Benz in

Germany launched a high fidelity driving simulator in the
A driving simulator is a virtual reality tool that gives a early 1980s ,]!I[ which created wide interests throughout
driver on board impression that he/she drives an actual the world. Since then, many automotive makers and
vehicle by predicting vehicle motion caused by driver research institutions have developed and applied their own
input and feeding back corresponding visual, motion, simulators that meet application purposes and target
audio and proprioceptive cues to the driver. The performance [4-61. Recently, construction of the state-of-
simulator normally consists of several subsystems as the-art National Advanced Driving Simulator (NADs) [7]
follows: a real-time vehicle simulation system performing in the United States has begun and will be completed in
real-time simulation of vehicle dynamics; motion, visual 1999.
and audio systems reproducing vehicle motion, driving Many factors should be carefully considered in
environment scenes and noise sensed by a driver during developing and applying full-scale driving simulators
driving; a control force loading system acting as an effectively, such as construction costs, application areas
interface between the driver and the simulator; an operator and target performance. A prototype simulator, down-
console for monitoring system operation; and system scaled, yet consisting of all the necessary subsystems can
integration managing information and data transfer among be used effectively for evaluation of the factors mentioned
subsystems and synchronization. The driving simulators above and design of the full-scale simulators. The
have been used effectively for vehicle system development, objective of this study is thus to develop an economical
safety improvement and human factor study. and effective driving simulator as a virtual reality tool and
The driving simulators, having their roots on flight apply it to design and evaluation of full-scale simulators
simulators applied since the early 1900s, have begun to and driver-vehicle interaction study. This paper describes
appear in primitive forms in the 1970s [1,2]. With the mainly the development of the simulator.

$10.00 0 1998 lEEE

0-7803-4300-~-5/98 71
SIMULATION CQMPONENTS system characteristics. These feedback cues must be
accurate and realistic enough to ensure high fidelity. They
The essence of interactive driving simulation is the must also be received by the driver without significant
cuing feedback that a driver receives based on control delay over what would be experienced in a real vehicle,
responses, input commands, and disturbances. The realism otherwise the effective vehicle dynamics will be degraded,
of the simulation then depends on the basic fidelity of the resulting in reduced fidelity and validity.
cuing, and the relationship between the driver’s control
responses and the response of the cuing devices. SYSTEM CONFIGURATION
Figure 1 illustrates various components that control
and generate visual, motion, control loading, instrument, Figure 2 shows a basic concept and configuration of
and auditory display feedback cues to a driver in driving the driving simulator developed in this study. The
simulation. Driver control inputs to the vehicle dynamics simulator is controlled and operated in a network of four
generate vehicle positions, velocities, accelerations, and personal computers connected by Ethernet.
orientations that then provide inputs to the various The driver on board the simulator takes driving action
feedback cuing pathways. In driver-driving simulator in terms of steering, braking, and accelerating in the given
closed-loop operation, the driver receives visual and driving scenarios. The driver input is then sensed by the
motion feedback of vehicle motion related with the control force loading system and fed back to the real-time
driver’s inputs, and proprioceptive feedback of steering dynamic computer through the RS-232C port. Simulation
torques as influenced by vehicle motions and steering of vehicle dynamics is then performed in real-time to

I User interface for Svstem Confiauration I

t t

Real-Time
+ Vehicle d
______

Instrument Driver

Fig 1 Interactive Driving Simulation Components

Visual I Audio Computer

Console Computer

6 D 0 F Motion Bed
Monitoring Console

Safety System

Algorithm Communication System

Ethernet 10 Base-T

Fig. 2 Driving Simulator Functional Diagram

72
predict the resultant vehicle motion. The visual and audio a steering system and a brake system. The models should
computer, with the real-time vehicle simulation output, also have flexible expandability. The modular vehicle
generates and displays realistic graphic image and noise of modeling approach has been adopted in this study for
driving environment corresponding to driving conditions. systematic development of the vehicle models.
The motion system computer runs a drive logic to control a For chassis and suspension motion, both a 16 DOF
hydraulic system that drives a 6 degree-of-freedom lumped mass model and a 14 DOF multibody model with
Stewart platform for creating realistic motion cue. The front Macpherson and rear trailing arm suspensions have
control force loading system, in addition to sensing the been developed. Based on a simple linear relationship
driver input, generates realistic reaction forces and torques between the accelerator pedal angle and the throttle valve
in the driving mechanism for proprioceptive cue. The angle, the input to the engine model has been determined
monitoring system is also included in the simulator to according to driving action. The engine speed and torque
monitor entire system operation and perform effective fail- have been computed using a simple engine model based on
safe functions. The driver cab is composed of mainly the the reference 8. A power train model, consisting of a static
driver side of an entire passenger car, along with the actual torque converter and an automatic transmission model
instrument panels and driving mechanism. with a transmission map, has been used to compute driving
A photograph of the simulator in operation is shown in torque that will be included in tire rolling dynamics. For
Figure 3. the brake system, a quasi-static model has been used to
compute brake pressure in relation to the brake pedal angle,
without considering hydraulic system dynamics. The
steering system has also been simplified to account for a
rack and pinion type.
Several real-time integration methods such as Euler,
Heun and A d a m methods have been used for real-time
simulation. I n selecting the right integration method,
several factors must be considered including accuracy,
stability, phase, delay, and execution time. Based on
results from several numerical experiments, the Adanis
method has been selected for this purpose [9]. The zero or
low speed contiitions occurring during various maneuvers,
including starting after stop, high speed driving, braking
and complete stop, cause undefined longitudinal slip and,
as a result, difficulty in simulation. This problem has been
solved by follolwing the approach taken by Bernard [ I O ] .
A personal computer equipped with dual-Pentium pro
processors has been set LIPas a real-time computer in t h i s
study. The execution time of the lumped mass and the
multibody models on the computer is less than 2 and 6
msec per step, respectively. For more complex and
realistic vehicle models, the parallel processing technology
is being considered for efficient parallel execution of the
Fig. 3 Driving Simulator in Operation vehicle models based on parallelism identified in
REAL-TIME VEHICLE SIMULATION SYSTEM - independent kinematic chains using shared meinor)'
The real-time vehicle simulation system is a key element multiprocessors [ 1 I ] .
of the driving simulator because accurate prediction of VISUAL AND1 AUDIO SYSTEM - Since visual cue is
vehicle motion with respect to driver input is essential. most significant in controlling and maneuvering a vehiclc
Three factors that are correlated each other must be during driving, the key clement for ensuring high fidclit)
considered for performing real-time vehicle simulation: in driving simulhtion is a visual system. Processing of high
vehicle modeling, real-time simulation and a real-time resolution graphics i n the visual system is essential l'or tlii'
computer. driver to have realistic driving feel and react to driving
Vehicle models used in the driving simulator must environment precisely. The visual system is divided into
satisfy contradicting conditions of realism and real-time In two paits for computer image generation arid displa)
order to predict vehicle motion caused by various driving Figure 4 shows a functional diagram of the visual/audio
action of the driver on board the simulator, the vehicle system i n this study.
models should include, in addition to basic chassis and The visual 5,ystems in the liill-scale driving siiniilators
suspensions, subsystems such as an engine, a power train, normally utilize very expensive high-end workstatiotis

73
 and 32 degrees of horizontal and vertical field of views,
Simulation
Results Driver Generator respectively.
1 a Speaker An audio system is an important addition to fidelity
enhancement of the driving simulator. In this study, an
Screen audio system has been set up with a Sound Blaster
AWE32 sound card and speakers. A sound database has
been constructed to store and manage tire, engine and
driving environment noise. The MIDI function of the
Fig. 4 VisuaVAudio System Functional Diagram sound card has then been used to combine and play
appropriate noise in coordination with driving conditions.
with dedicated graphic cards and graphic software to MOTION SYSTEM - Reproducing ride and handling
generate high fidelity visual image. in this study, an characteristics of a vehicle that a driver feels through
economical, yet very efficient visual system computer has chassis linear accelerations and angular velocities is an
been set up with a personal coinputer with a Pentiuin pro important element for enhancing siinulator fidelity.
processor and a Glint 500TX+Delta 3D graphic However, a care should be taken not to cause conflicting
accelerator [ 121. Visual image generation software has effects among subsystems and induce simulator sickness,
been developed using OpenGL 1 . 1 , and objects have been due to high sensitivity of the motion system.
inodeled using V R M L 1 .O. A visual database has also been The kinematic structure of the motion system should be
developed for storing and inanaging model, object and considered first in developing the motion system. A widely
terrain data systematically. used 6 DOF Stewart platform, driven hydraulically, has
A system configuration file in ASCII format has been been chosen to be the motion platform in this study. Target
utilized for defining driving scenarios and measuring performance and motion envelope of the motion system
performance easily. A typical driving scene with texture have been defined with reference to MIL-STD-1558
map generated by the visual system is shown in Figure 5 . specification for flight simulators, as shown in Table 1 .
Fog and night effects can be imposed on the same scenario Based on the target performance, motion envelope and
easily. For display of generated driving scenes, a single driving scenarios, a hydraulic circuit and its elements
channel projection system has been developed having 42 including actuators, servovalves, and other components
have been designed. The lower part in Figure 3 shows the
motion system develope' in this study

Table 1. Motiot Platform Requirements

A drive logic of the motion system for generating

realistic motion cue imludes an washout algorithm,
inverse kinematic analysis, and a control algorithm, as
organized in Figure 6. The washout algorithm recovers
Fig. 5. Typical Driving Scene with Texture Map motion cue that is realizable within the motion envelope

Motion
Actuator
Kinematics Controller Platform

Algorithm

Fig. 6 Motion System Functional Block Diagram

74
from command cue of vehicle simulation output. A
washout algorithm has been developed in this study for
software limitation of platform motion based on high-pass
filtering, and tilt coordination for generating sustained
inertial acceleration required in some maneuvers such as J-
turn based on low-pass filtering. Chassis position and
orientation information from the washout algorithm has
then been converted into actuator lengths by the inverse
kinematic analysis. A PID control algorithm has been
developed to ensure movement of the motion platform in
the right position. The drive software has been developed
based on the flexible LabVIEW software [13].
Figure 7 shows a typical response of the motion system
in terms of voltages representing actuator lengths. As Fig. 8 Driver Cab
shown in the figure, the motion system response indicated
by the solid line closely follows the driving input from the SYSTEM INTEGRATION
washout algorithm represented by the dotted line.
The system integration technology, managing
........... information and data transfer among subsystems,

t
Wi& Enpsrt
___ Gwrtrolled Re€.pdn*e synchronization, and fail-safe functions, is a key factor in
./-\>.
.i n determining the fidelity and performance of the driving
simulator. Thus, the method for integrating the subsystems
should be carefully considered from the early stages of
simulator development.
The Windlows NT software has been chosen as the
operating system of the driving simulator in this study, due
to reliability and many convenient features. A software
time interrupt function has been added to the basic
Windows NT software for real-time operation.
Definition of input/output signal flows among the
5 6
fW%) subsystems is a prerequisite to effective system integration.
The signal flow diagram including types and sizes of I/O
Fig. 7 Typical Motion System Response signals has been carefully prepared at the initial
CONTROL FORCE LOADING SYSTEM - The development stage, and then the network communication
control force loading system acts as an interface between flow based on Ethernet has been defined. LabVlEW
the driving simulator and the driver, in that it senses driver software has been utilized to develop a flexible
input and feeds it back to the real-time vehicle simulation communication module as shown in Figure 9.
system, displays vehicle operating conditions on
instrument panels, and generates reaction forces and
torques in the driving mechanism for proprioceptive cue.
The system developed in this study consists of the
following components: a controller with five micro-
controllers and a communication module for managing
data transfer, computing reaction forces and torques, and
controlling encoders, motors and lamps; rotary encoders
for sensing driver input; and DC geared motors for
actuating reaction forces and torques. The driver cab with
the control force loading system implemented is shown in
Figure 8. i
Driver software has been developed using an assembler Fig 9 Network Module Using by LabVlEW
to control appropriate functions effectively. An algorithm
for computing reaction forces and torques have been The operation procedure of the simulator includes
developed based on the table look-up approach. Use of system initialization, a series of tests for checking normal
vehicle dynamics for more accurate and flexible operability ancl controllability, and simulation according to
computation is also being implemented. driving scenarios. The operation module has also been

75
developed by combining individually developed REFERENCES
subsystem modules in LabVlEW and FORTRAN format.
Transport delay is a phenomenon whereby the response 1. B. Repa and W. Wierwille, "Driver Performance in
of a dynamic element falls behind in time relative to its Controlling a Driving Simulator with Varying Vehicle
input. Without including human perception delay and Response Characteristics," SAE Paper 760779, 1976.
physical vehicle delay occurring in real driving situations,
there are three major sources of transport delay in the 2. B. Richter, "Driving Simulator Studies: The Influence
driving simulator: the vehicle simulation, visual, and of Vehicle Parameters on Safety in Critical Situations,"
motion systems. Real-time vehicle simulation is achieved SAE Paper 741 105, 1974.
as mentioned above, and thus delay associated with
vehicle dynamics computation remains below the 7
3. J. Drosdol and F. Panik, "The Daimler-Benz Driving
perception threshold of the driver. Transport delay in the Simulator A Tool for Vehicle Development," SAE
visual system includes data acquisition, image processing Paper 850334, 1985.
and display time, and is in the range of 25 and 50 msec
depending on the quality of visual images. The delay in the 4. G.P. Bertollini, et. al., "The General Motors Driving
motion system is around 50 msec, including data Simulator," SAE Paper 940176, 1994.
acquisition, drive logic computation, and motion platform
I-esponse time. Compensation of transport delay is critical 5. T. Suetomi, et. al., "The Driving Simulator with Large
i n ensuring the fidelity of the simulator, in that delay Amplitude Motion System," JSAE Paper No. 902278
degrades system performance, and further system stability. (In Japanese).
The prediction method based on numerical integration [ 141
is being applied to compensate for individual delays and 6 J.S. Freeman, et. al., "The Iowa Driving Simulator: An
guarantee synchronization. Implementation and Application Overview," SAE
The fail-safe function has also been carefully Paper 950174, 1995.
considered here. The monitoring console, as shown in
Figure 10, has been developed to monitor entire system 7 E.J. .Haug, et. al., Feasibility Study and Conceptual
operation periodically and, when emergency occurs, move Design of a National Advanced Driving Simulator:
the motion platform to the neutral position safely. Final Report, NHTSA Report DOTHS 807596, 1990.

8 D. Cho and J.K. Hedrick, "Automotive Powertrain

Modeling for Control," ASME Journal of Dynamic
Systems, Measurement and Control, Vol.111, No.4,
1989, pp. 568-576.

9 W.S. Lee and S.S. Kim, "Real-Time Vehicle Dynamic

Simulation," Proc. of AVEC '94, Japan, 1994, pp.105-
109.

10 J.E. Bernard and C.L. Clover, "Tire Modeling for Low-

Speed and High-speed Calculations," SAE Paper
950311. 1995.

11 W.S. Lee and T.O. Tak, "Parallel Processing of

Fig. I O Monitoring Console for System Safety
Controlled Multibody Mechanical System Dynamics,"
Engineering with Computers, Vol. 7, 1991, pp. 161-
CONCLUSIONS 172.

12. GLINT 500TX Professional 3D Graphic Processor,

An economical, yet effective driving simulator as a
virtual reality tool, consisting of all the necessary Internet Documents.
components required in full-scale simulators, has been
13. LabVIEW User Manual, National Instruments, 1996
developed. It will be used for design and evaluation of the
full-scale simulators and driver-vehicle interaction study.
Experiments with the developed simulator are underway 14. W.S. Lee, "Compensation of Transport Delay in
and the results will be reported separately. Driving Simulator," 1991 American Control
Conference.

76