The Use of A Variable-Stability Vehicle in Handling Research

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Downloaded from SAE International by University of Michigan, Monday, September 17, 2018

650659

The Use of a Variable-Stability


Vehicle in Handling Research

R. Thomas Bundorf
Research Laboratories, General Motors Corp.

A VARIABLE STABILITY vehicle is some system such as an The optimum in vehicle directional control charac­
airplane or passenger car which can be made to adopt vari­ teristics is that combination of handling qualities that is
ous control characteristics. As an example, the variable sta­- best used by the driver. It is probable that it is also a con­
bility passenger car can be made to "handle" like a sports figuration that is well liked by the driver. This vehicle
car or like a grossly overloaded station wagon. The control configuration is somewhere within an extremely large variety
characteristics can be varied through a very wide range, of possible combinations. The number of these configurations
which includes most present-day vehicles as well as untried is readily apparent when one considers that the directional
vehicle configurations and tried configurations that have nott control characteristics of a passenger car are affected by no
proved to be acceptable. The range includes vehicles that less than 27 design parameters, most of which interact with
are stable and vehicles that are divergently or oscillatorily each other. Any survey of vehicle handling quality obviously
unstable; hence, the name "variable stability vehicle." requires a special vehicle that can easily and quickly adopt
A variable stability vehicle usually embodies one or more3 a wide variety of directional control characteristics.
electrohydraulic servo systems. The system discussed here The central topic of this paper is the use of the GMR var­
has a servo mechanism that actuates and positions the front iable stability vehicle, built by Cornell Aeronautical Labor-
[.
wheels and another that applies torques to the steering wheel. atory, Inc., in the early stages of a research program directed
Signals to these servo actuators are electrical; consequently,, toward definition of optimum vehicle handling qualities. A
it is very easy to modify them by attenuation or by addition pilot study of driver performance in a maneuvering task is
of new signals. The operator of a variable stability vehicle outlined, in which the variable stability vehicle is used in
has only to turn some knobs in order to change his vehicle's several handling configurations. Particular attention is given
control characteristics. This is in marked contrast to the to the technical problems that must be solved in such an in­
process of interchanging vehicle parts to obtain different vestigation.
configurations.
Such a vehicle has many uses. It may be used as a test THE GMR VARIABLE STABILITY VEHICLE
bed for new design ideas for conventional vehicles or as a
tool in a search for new subsystems that could be added to The GMR variable stability vehicle in its present config­
conventional vehicles for handling improvement. However, uration is in a 1956 passenger car chassis. All original d e ­
its most important current use is in the search for optimum sign parameters of the vehicle have remained unchanged e x ­
vehicle directional control characteristics. cept for the steering system. Here, the mechanical linkage

ABSTRACT

This paper describes the use of the GMR variable stabil­ types of passenger car directional control characteristics were
ity passenger car in a brief study of driver performance in simulated, and each configuration was driven by each of six
a maneuvering task. The study was part of a pilot program different drivers through a complex course. The results of
for evaluation of test methods and equipment for future and the investigation are presented in terms of the average driver
more extensive human factors evaluations. Three distinct performance with each vehicle configuration.

227
Downloaded from SAE International by University of Michigan, Monday, September 17, 2018

228 R. T. BUNDORF

CAR MOTION SENSORS

I ' 1

STEERING TORQUE
COMPUTER
-STEERING TOROUE
LATERAL TRANSDUCER
ACCELERATION STEERING TORQUE
SERVO

,....,.
FRONT WHEEL FRONT WHEEL
STEERING TORQUE
MOTOR
POSITION COMPUTER POSITION SERVO

- FRONT WHEEL ACTUATOR

Fig. 1 - Variable stability vehicle operation schematic

between steering wheel and front wheels may be disengaged angle would appear exactly like front roll understeer and have
completely in favor of two servo mechanisms, which provide the same effect on the vehicle's transient response. More­
for control of front wheel position and steering wheel forces. over, it is apparent that the amount of artificial front roll
An operational schematic is shown in Fig. 1. understeer may be controlled by attenuating this electric sig­
Rather than directly turning the front wheels, any steering nal.
input by the driver turns a potentiometer located on the steer­ Not all feedback signals bear such a direct relation to a
ing column, thereby sending an electric signal to the front physical parameter. Yaw acceleration feedback,for e x a m ­
wheel position servo. This electrohydraulic servo then posi­ ple, produces an effect quite analogous to a change in yaw
tions the front wheels in proportion to the magnitude of the moment of inertia, but some analysis is required to substanti­
electric voltage signal, and the vehicle turning motion is ate this, and the relationship is not exact.
initiated in this manner. The separation of the steering wheel from the front wheels
Also included in the vehicle are many motion transducers. also eliminates the normal tire aligning torque feedback to
Examples are a yaw acceleration transducer and a lateral the driver's-hands. The steering wheel torque servo replaces
accelerometer, each of which generates electric signals in this with various kinds of torque feedback. In general, m o ­
proportion to the detected motion. Any combination of these tion transducers supply signals to the torque servo, but other
signals and others may be fed to the front wheel position servo important sources are pickups that sense steering wheel pos­
as additions to the original steering input signal. Thus, the ition and velocity.
original input signal is modified by the feedback from the The selection and attenuation of the various feedback sig­
motion transducers, and the vehicle response is modified in nals is made at a control panel located on the vehicle dash.
turn. This is shown in Fig. 2. The upper row of potentiometers
The driver is, of course, aware only of his own steering meter input feedback signals to the front wheel position
input and the consequent vehicle response. To him, the o p ­ servo, and the lower row meters signals to the steering c o l ­
eration appears normal, although the vehicle may have d i ­ umn torque servo. In this way, the directional control char­
rectional control characteristics not in keeping with those acteristics of the vehicle can be changed in a few moments
of a typical passenger car. without the operator's leaving the driver's seat.
A detailed account of.the system design is not within the
scope of this paper. For this information, reference may be METHODS FOR DESCRIBING VEHICLE CONFIGURATIONS
made to the current paper by Segel.* It is possible to
understand the general nature of the system by noting that
Vehicle Description by Physical Parameters or by Normal
any added steer angle to the front wheel servo that is
Force Coefficients - It may be assumed at this point that if
applied as a result of a feedback signal proportional to roll
a vehicle or its directional control characteristics can be
described by any accurate method, the variable stability v e ­
*L. Segel, "The Variable-Stability Automobile: Concept hicle can be made to reproduce these directional control
and Sesign." SAE Transactions, Vol. 74(1966), paper qualities. This, then, introduces a primary problem in the
605658. planning of the handling study which is that of deciding what
Downloaded from SAE International by University of Michigan, Monday, September 17, 2018

VARIABLE-STABILITY VEHICLE IN HANDLING RESEARCH 229

Table 1 - Physical Design Parameters


k <r for a Passenger Car Vehicle
(Linear Directional Control Simulation)

Typical Range of
Parameter Variation

Inertia
Sprung Yaw
1650-5500 slug-ft
Unsprung Yaw
2
Sprung Roll 300-600 slug-ft

Sprung Yaw-Roll Product -50 to + 100 slug-ft'


Fig. 2 - Variable stability vehicle control panel
Mass
Sprung
80-180 slugs
vehicle configurations to test. (Other very significant prob­ Unsprung
lems not considered in detail in this report are the selection Mass eg Location
of tasks, drivers, and performance criteria.) This isbecause Sprung Fore-Aft 35-55% forward
of the vast number of vehicle configurations and the obvious weight distribution
impossibility of a complete parameter study. Sprung Height 1.7-2.0 ft
Table 1 lists the physical design parameters that affect Unsprung Fore-Aft
the linearized directional control response of the passenger Unsprung Height
car. The range of variation in these parameters in current Roll Centers
passenger cars is also given. It is apparent that the variabil­ Front Height - 1 to +3 in. above ground
ity in current vehicle design is very great and that some Rear Height +4 to +15 in. above ground
simpler method with fewer parameters is required. (This is Wheelbase 8.2-10.8ft
a principal reason for the difficulty in analyzing and defin­ Roll Steer Coefficients
ing vehicle directional control. The design engineer, the Front 10founder- to 10"^over-
analyst, and the human factors researcher all must contend steer
with the problem of many parameters and variables that often Rear 10% under- to 0% over-
interact with each other.) To attempt a handling study based steer
on design parameter variation is appealing, but is almost Roll Camber Coefficient
certainly impossible. Front 0.70-1.0 deg/deg
Another possibility is the use of the equations that d e ­ Rear -0.1 to 0.8 deg/deg
scribe vehicle steady-state and transient responses. This tech­ Roll Ra'te (total) 375-975 ft-lb/deg
nique usually reduces the number of parameters. For exam­ Roll Damping
ple, a mass-spring-damper, a three parameter system, may Tire Cornering Stiffness
be alternatively described by two parameters, its natural fre­ Front 110-230 lb/deg (per tire)
quency and its damping factor. Rear
An example applied to the automobile equations is the Tire Camber Stiffness
stability derivative notation that has been used in the air­ Front
10-40 lb/deg (per tire)
craft industry and later applied to the automobile by Milli- Rear
ken, Segel, et a l . * at the Cornell Aeronautical Laboratory. Tire Aligning Torque
This reduces the number of variables to 17, as may be seen Coefficient
in Table 2, but also results in the loss of the design para­ Front
15-60 ft-lb/deg (pertire)
meter meaning except for the inertial parameters. Instead, Rear
the stability derivatives have a meaning that is associated Steering Ratio
with the dynamic response characteristics of the vehicle. The Overall 18.0:1 to 28.0:1
coefficient N ' , for example, represents the yawing moment Vehicle Forward Velocity 0-100 mph

on the vehicle induced by the existence of one radian per

second of yaw velocity, and the stability derivativeN rep-


* W. F. Milliken, D. W. Whitcomb, L. Segal, W. Close,
5
C. L. Muzzey, A. G. Fonda, "Research in Automobile Sta­
resents the yawing moment induced by the existence of one
bility and Control and in Tire Performance," Inst.ofMech.
radian of steer angle. (Actually, the design parameter m e a n ­
Engr., London, 1965.
ing is not completely lost, since the stability derivatives may
Downloaded from SAE International by University of Michigan, Monday, September 17, 2018

230 R. T. BUNDORF

Table 2 - Directional Control Equations and Parameters in Stability


Derivative Notation

v(M + M ) u 0 + M (-z ) rp =v -(M + M ) ur + Y„ 6 + Y r


s u s s s u 0 r

+ Y^ 0 + Y 6
0 6

(I +I ) r +I p = N 6 + N I + N J + N J
zs zu xzs 8 r 0 5

I p+I r + M ( - z ) u j 3 = - M (-z ) ur + L p + L 0
xs xzs s s s s p 0

Parameters
= Total yaw inertia
zs zu
= Roll inertia of sprung mass
= Product of inertia of sprung mass
xzs
M = Sprung mass
s
M = Unsprung mass
u
(-) = Sprung mass eg height above roll axis
u = Forward velocity
6 = Sideslip angle
r = Yaw velocity
= Roll angle
= Steer angle
Y (n = 6, r,0, 6) = Lateral force per unit input of motion variable, n
n
N (n = 6, r, 0, 6) = Yawing moment per unit input of motion variable, n
n
Ln (n = p, 0) = Rolling moment per unit input of motion variable, n

be expressed as relatively simple functions of the design par­ duced by the existence of 1 radian/sec of yaw velocity, and
ameters.) the stability derivative N ' represents the yaw acceleration
This stability derivative form may be further reduced to
induced by the existence of 1 radian of steer angle. The num­
a "normal" form with a minimum number of parameters,
ber of parameters required for description of the vehicle sys­
called "normal form coefficients." In this case, the left-
tem by this method has been reduced to 15. This seems to
hand sides of the equations are greatly simplified by absorbing
be the simplest exact form for the directional control equa­
inertial parameters into the stability derivatives and by
tions; and for a simpler description, one must resort to ap­
separating the acceleration variables. This is accomplished
proximations.
by matrix techniques, and the equations of Table 3 are the
Vehicle Description by Response Characteristics -A sim­
result.
pler, although approximate, characterization may exist in
A disadvantage of this form is that the design parameter
the description of the steady-state and transient responses
meaning is even more difficult to retain. In the case of the
of each vehicle motion variable: yaw velocity, sideslip an­
stability derivative notation, expressions for the stability de­
gle, and roll angle. A steady-state response is described by
rivatives in terms of vehicle design parameters could be ob­
a single number, steady-state gain, which is a measure of
tained. However, the expressions for the normal form co­
the motion variable's steady-state amplitude for a unit input
efficients are extremely involved, with the result that one
of steer angle. For example, a typical value of steady-state
must resort to the computer to establish numerical normal
yaw velocity gain of a vehicle operating at 60mph is 0.16
form coefficients.
deg/sec/deg, meaning that if a degree of steering wheel an­
The meaning of the normal form coefficients is similar
gle is applied, 0.16 deg/sec of yaw velocity results when the
to that of the stability derivatives. The normal form coef­
vehicle has reached steady state. Values for sideslip gain
ficient N' represents the yaw acceleration of the vehicle in­
and roll gain are similarly obtainable either by computation
Downloaded from SAE International by University of Michigan, Monday, September 17, 2018

VARIABLE-STABILITY VEHICLE IN HANDLING RESEARCH 231


STEERING
Table 3 - Directional Control Equations and INPUT - DEC

Parameters in Normal Form Coefficient Notation 0.5 1.0


TIME - SEC.

;=Y'B +Y , r + Y , p +Y'0 + Y;5


6 r p 0 6

r = N * 6 + N ' r + N , 1p + N ' 0 + N * 6 YAW VELOCITY


6 r p ^ 0 5 DEGySEC. ,„

p = L ' 6 + L ' r +L'p+L'0 + L'5


6 r p 0 6

Parameters
0.5 1.0 1.5 2.0
0
0 = Sideslip angle
s
t
ANGLE - DEG.
r = Yaw velocity -0.5 Lh
p = Roll velocity t
0 = Roll angle
6 = Steer angle ROU
ANGLE DEG
Y ' (n = 6, r,p,0, or 6) = Sideslip angle time rate of " -
change per unit input of m o ­
tion variable
0.5 1.0
N ' TIME - SEC.
n (n = 6, r , p , 0 , or 6) = Yaw acceleration per unit i n ­
For'd. Velocity - 60 mph
put of motion variable
L ' (n = 6, r, p, 0, or 6) = Roll acceleration per unit in­ Fig. 3 - Typical passenger car transient responses
put of motion variable

nificance, but requires many parameters. The second is


quite rigorous, but is also complex and difficult to under­
with use of the vehicle directional control equations or by
stand. The third has least parameters, but is also least exact.
experimental measurement with an actual vehicle.
Although there may be other methods for description, it is
The transient response is not so easily described, since apparent that the passenger car, even in its simplest form,
it represents the "manner" in which the response reaches its is a very complicated machine and is not to be described
steady-state value. Fig. 3 shows a set of responses for a typ­ accurately by a few simple parameters.
ical vehicle operating at 60 mph. A steering wheel angle
of 10 deg was applied in the form of a step input, and the
USE OF THE VARIABLE STABILITY VEHICLE IN A HAN­
gain of each variable is indicated by its steady-state value
DLING STUDY
in relation to this input.
The transient response characteristics may be described Selection of Vehicle Configurations - For the brief han­
by several terms, as indicated in Fig. 3. Rise time refers to dling study outlined here, only three vehicle configurations
the time required for the response to reach 90^0 of its steady - were required. Each of the vehicles was to be different from
state value. Overshoot refers to the magnitude of the initial the others in a significant way, and it was decided to give
oscillation related to the steady-state value. Settling time particular attention to the response description method as
refers to the time required for the response to attain a value a criterion for describing these differences.
not exceeding a + 5°/o variation from steady state. Values
Yaw velocity response is the most significant of the three
for this vehicle include yaw velocity rise time of 0.20 sec,
commonly used motion variables. It is the primary influ­
yaw velocity overshoot of 31%, and yaw velocity settling
ence on the vehicle's heading, path, and lateral accelera­
time of 1.05 sec.
tion responses. The vehicle heading is the integral of the
If all three vehicle motion variable responses were to be yaw velocity response:
described by this method, a total of 12 parameters would be
required, 3 for transient response, and 1 for gain for each of
the three variables. The description would give the essen­ r dt
tial information even though it would not be so rigorous as
the normal form coefficient method. A tabulation for the
data of Fig. 3 is given in Table 4. where
Each of these three techniques (design parameters, nor­
mal form coefficients, and response description) has advan­ * = Heading angle
tages and disadvantages. The first has a simple physical sig- r = Yaw velocity
Downloaded from SAE International by University of Michigan, Monday, September 17, 2018

232 R. T. BUNDORF

Table 4 - Transient Response Parameter Description of a Typical Passenger Car


Settling
Gain** Rise T i m e , sec Overshoot , % T i m e , sec

Yaw Velocity n 1fi1 deg/sec


0.20 31 1.05
Response* ' deg
Sideslip Angle Q Q 5 2 deg
0.67 9 1.42
Response* ' deg
Roll Angle deg
0.098 0.64 9 1.45
Response* deg

* Forward velocity = 60 mph.


* * Referred to steering wheel.

Path change is the double integral of the yaw velocity


response with an added minor effect due to sideslip angle: YAW VELOCITY
CONFIGURATION A

DEG./SEC.

= j f r dt dt + j 0 dt
''O ^O "'0

where 1.0
TIME - SEC.

y = Lateral displacement CONFIGURATION B


YAW VELOCITY
6 = Sideslip angle DEG./SEC.

The lateral acceleration of a point on the vehicle in


steady state is equal to the vehicle forward velocity times
yaw velocity, and this transient response is markedly influ­
enced by yaw velocity response.
0.5 1.0
TIME - SEC.

Lateral acceleration = ufi + h p + ur


CONFIGURATION C
YAW VELOCITY
DEG./SEC.
where

u = Forward velocity
6 = Sideslip acceleration
p = Roll acceleration 0.5 1.0 1
h = Height of the point above the roll axis TIME - SEC.

For'd. Velocity - 30 mph 5 - 10 Degrees

The apparent primary role of yaw velocity in establishing Fig. 4 - Yaw velocity responses of variable stability v e h i ­
the transient response characteristics of a passenger car in­ cle test configurations
dicates that this variable should receive most attention
in the process of selecting vehicle configurations. Accord­
ingly, three distinct yaw velocity transient response charac­ figuration A is a typical yaw velocity response for an under-
teristics were selected and are shown in Fig. 4 . These r e ­ steering vehicle. It is marked by a short rise time and a
sponses have been obtained from computer simulation in definite overshoot. In the variable stability vehicle, it is
order that the transient responses may be shown as step input obtained by adding negative sideslip acceleration feedback
responses and described by the rise time, overshoot, and and negative yaw velocity feedback.
settling time criteria. (It is not possible to place a step in­ Configuration B represents a vehicle with less understeer
put of steer angle on an actual vehicle because of the slight characteristic but with a short rise t i m e . Also, the initial
lag that occurs through the steering system, whether it be yaw acceleration is slightly greater. In actual practice, this
of a conventional nature or a servo system.) The simulated would require that the vehicle have a high tire cornering
vehicle speed was 30 mph. stiffness relative to its yaw inertia, since response times for
These response characteristics were chosen because each such vehicles are longer than for vehicles with marked un­
has a distinct Ci aracteristic typical of a classification of dersteer quality. This type of characteristic is obtained in
passenger car vehicles. The trace representing vehicle con- the variable stability vehicle by adding a small amount of
Downloaded from SAE International by University of Michigan, Monday, September 17, 2018

VARIABLE-STABILITY VEHICLE IN HANDLING RESEARCH 233

Table 5 - Transient Response Parameter Description


of the Variable Stability Passenger Car Test Configurations

Rise Settling
Configuration Gain* T i m e , sec Overshoot, °Jo T i m e , sec

deg/sec
Yaw 0.15 0.26 7.5 0.64
deg
Velocity 0.15 0.31 1.8 0.38
Response* 0.15 0.47 0.0 0.56

* Forward velocity = 30 mph.


' * Referred to steering wheel.

positive sideslip acceleration feedback and appreciable neg­ Selection of a Tracking Task and a Performance Criter­
ative yaw velocity feedback. ion - The evaluation of vehicle handling qualities by the
Configuration C also represents a vehicle with only slight design engineer is usually accomplished by the collection
understeer characteristic. In this example, however, the in­ of subjective opinions based upon proving ground and open
itial yaw velocity response is not so fast, and the rise time road driving. At the proving ground, the driver can apply
is long. Again there is virtually no overshoot. The princi­ voluntary steering inputs that are large or small, fast or slow,
pal difference between this vehicle and configuration B is simulating passing maneuvers, fast lane changes, large fixed
not in the response curve shape but is instead in the response control inputs, or any kind of maneuver that permits a class­
t i m e . This effect is obtained in a variable stability vehicle ification of the vehicle's behavior. On the open road, he
by adding both positive sideslip acceleration feedback and may drive the vehicle for long periods without voluntarily
positive yaw velocity feedback. creating maneuvers but rather by accepting the events that
For each configuration, the steering ratio was adjusted normally occur (such as slippery roads, sharp curves, or cross-
by attenuation of the steering signals so that the yaw velo­ winds) and evaluating the vehicle on the basis of such real
city gains were equal. Consequently, in each case, a speci­ svents.
fic input of steer angle would develop the same radius of A thorough evaluation of handling qualities should include
turn, but the manner in which the response occurred would a similar variety of driving tasks. For controlled and repeat-
be different. A summary of the response parameters for the able tests, the tasks may be in the form of course tracking,
configuration is given in Table 5. crosswind disturbance correction, rolling road disturbance
These markedly different yaw velocity response charac­ correction, or obstacle avoidance. There are many possible
teristics were developed in a vehicle operating at 30 mph. examples, eaeh having a particular test format and perform­
However, they are very representative of the type of yaw ance measure.
velocity response characteristics that occur athigherspeeds. The example to be discussed is a course tracking task in
This illustrates an additional advantage of the variable sta­ which the driver is required to follow a narrow, winding course
bility vehicle in that it can simulate high speed response at a fixed speed without exceeding its boundaries. The course
characteristics at a lower and safer operating speed. is shown in Fig. 5. Its boundaries were marked by rubber
While this particular study made some investigation into traffic control cones, which could be struck by the vehicle
yaw velocity transient response variation, it did not include and moved without damage. Such a test is repeatable over
study of variation in yaw velocity gain, not did it include and over under a variety of vehicles and drivers. More i m ­
any variation of sideslip or roll characteristics. Consequently, portantly, the test can be designed to give a measure of the
yaw velocity gain was maintained at a constant level and driver's performance as well as his subjective opinion.
sideslip responses were maintained as constant as was pos­ This "serpentine course" task was first designed on paper
sible for the three configurations. Roll properties were also and then further improved by experimentation. Definition
constant, since the vehicle does not permit variation of roll of the task was very simple: The driver was to drive through
rate and roll inertia. the course, knocking over as few traffic cones as possible.
An extensive program in handling research would require His performance was measured by the number of cones that
a great many such configurations with tests made at higher were moved. Therefore, it was desired that the course should
operating speeds as well as at 30 mph. The study outlined have a proper degree of difficulty, and this was predeter­
here was one of several conducted in a pilot program for mined on the basis of the kinds of vehicles that would be
evaluation of testing methods, equipment, and perform­ tested.
ance criteria, for more extensive handling research pro­ The shape of the course was determined analytically as
grams, and is presented only as an illustration of the use the summation of six sine wave traces of different frequen­
of a variable stability automobile in such a program. cies. The amplitude of each of the six sine wave traces was
Downloaded from SAE International by University of Michigan, Monday, September 17, 2018

234 R. T. BUNDORF

COURSE I N SCALE

OO
PYLON
LATERAL 0 •■ > *J± /. 3«s, _ _ ._ / \ ■
r =
DISPLACEMENT „o'100_• N </-200 o ■"• -TOO " o •400'' *
SCALE - FT. 2 +

6.7 Ft. —
(not to scale)
Fig. 5 - Random road course layout

INPUT OUTPUT
Double
Frequency Amplitude
(cps) (feet)
0.30 2.000

0.43 0.971

0.59 0.512

0.80 0.280

1.05 0.163
1.30 0.106
1.55 0.075
Fig. 6 - Random road course computation schematic

chosen in inverse proportion to the square of its frequency,


dling engineers or professional drivers, although each was
so that the lateral accelerations consequent to the vehiclean experienced and capable driver. Each was directed to
following the course would not be dominated by the higher drive through the course four successive times, each run con­
frequencies. The frequencies were selected through and sisting of a round trip, which included both directions of travel.
slightly in excess of the bandwidth (frequency response range)
Each driver drove all three of the vehicle configurations in
of a typical passenger car. They are shown in Fig. 6. A sec­
this manner. The order of driving the configurations was
tion of the trace representing the summation was selected randomized throughout the sample of drivers so that some
and reproduced on a road surface with wave length corre­ drivers drove configuration A first, some configuration B first,
sponding to a vehicle forward velocity of 30mph. and so on. Each driver was scored as to the number of cones
The total length of the course for this speed was 461 ft.
moved on each of the four runs for each car configuration.
(At 60 mph, it would be 922 ft.) Some preliminary opera­ The results of the pilot study are presented in Fig. 7. All
tion indicated that the course difficulty could also be con­
data are represented in a single plot of performance versus
trolled by changing its width, which was finally adjusted to
the number of the driving run. Performance is indicated on
6.7ft. This permitted a perfect score with only a very good the ordinate as the average number of cones moved during
vehicle configuration. Normally, a driver would knock over any one run. The first datum point for configuration A rep­
three to six traffic cones in each ran through the course. resents the average number of cones moved by the drivers
The Test Procedure and Results - In practice, the seem­ of configuration A on their first round trip through the course.
ingly random nature of the course was an advantage because The second datum point represents the average performance
it did not permit the drivers to l'eam its patterns easily, es­
on their second trip with configuration A, and so on through
pecially since it was driven in both directions. The traffic
all four runs. Clearly, the drivers are able to adapt to the
cones clearly delineated the course and provided the drivers
vehicles and improve their performance rapidly. Nonethe­
with a simply defined but difficult series of objects around
less, slight differences still existed even after the third trial,
which to maneuver. It was discovered that car speed was very
and it is evident that the test will discriminate between ve­
important because lower speed meant in effect that the fre­hicle configurations. The best criterion, however, may be
quency content of the course had also been lowered. In addi­
one of learning time and early performance, rather than per­
tion, vehicle bandwidth becomes higher at lower operating formance after a learning period.
speeds and the vehicle handling quality improves. Since the Although the study was not intended to provide any con­
drivers tended to slow the vehicle when in difficulty, speed
clusive evaluations of vehicle handling quality, it does per­
was controlled by an automatic throttle control system. mit some tentative conclusions. First, from our personal ob­
A small group of drivers consisting of five men and one servations, the A and C configurations were significantly
woman was selected from personnel working in various areas different in handling characteristics, far greater than a sub­
of General Motors Research Laboratories. None were han­ jective "just noticeable difference." Yet, the driver per-
Downloaded from SAE International by University of Michigan, Monday, September 17, 2018

VARIABLE-STABILITY VEHICLE IN HANDLING RESEARCH 235


Q Configuration A

O Configuration 8

A Configuration C
AVERAGE DRIVER
PERFORMANCE 10
(number of cones moved
per round trip)

RUN NUMBER Fig. 7 - Driver performance in random road tracking task

formance was very nearly the same after a short learning per­ Finally, and in conclusion, the variable stability vehicle
iod. This is excellent evidence of the great adaptability of is shown to be able to produce a variety of vehicle handling
the human operator, who is able to change his control func­ configurations with no delay or apparent physical change in
tions very quickly in order to optimize the driver-vehicle the vehicle, thereby constituting an ideal tool for handling
control system. research studies. At GMR, this vehicle and future variable
Second, the results indicate that configuration C was less stability vehicles are expected to be principal tools in basic
desirable than configuration B, although each had a similar handling research activities.
response shape. The chief difference is in the rise time, in­
dicating that the difference between 0.47 sec and 0.31 sec
ACKNOWLEDGMENT
is significant.
Third, the lack of a real difference in performance data
of configurations A and B indicates that response shape may The author wishes to acknowledge the very significant
not be an extremely important factor if the rise times are contributions by Dr. Paul L. Olson, senior research psycholo­
relatively low and similar or if the areas under the curves gist of the General Motors Research Laboratories, to the han­
(which are indicators of heading response) are relatively equal. dling study presented in this paper.

You might also like