Sie sind auf Seite 1von 25

Design of an Electronic

Stability Program for


vehicle simulation software

B.J.S. van Putten

DCT 2008.138

Master Traineeship

Supervisors
Dipl. Ing. M. Baderschneider, TESIS DYNAware GmbH
Dipl. Ing. Dipl. Phys. A. Pinnel, TESIS DYNAware GmbH
Dr. Ir. I.J.M. Besselink, Technische Universiteit Eindhoven

Eindhoven University of Technology


Department of mechanical engineering
Automotive Engineering Science

Eindhoven, November 2008


Abstract
Whereas these days in most vehicles an Electronic Stability Program (ESP) is standard or
available as an option, in highly realistic vehicle simulators and vehicle simulation packages
ESP is not always included as such. The investigation is aimed at obtaining an efficient and
effective ESP for this application, which is then built in a Matlab/Simulink environment to suit
veDYNA vehicle simulation software. An analytical analysis, supported by experimental
results, shows the abilities of an ESP, using sensitivities to determine at which wheel an
intervention takes place and a combination of the direct influence of braking as well as its
secondary effect, to control vehicle behavior under critical circumstances.

1
Table of Contents

1. Introduction ..........................................................................................................................3

2. ESP General and Starting Point ...........................................................................................4


2.1 Critical behavior ..........................................................................................................4
2.2 ESP Controller Essen Duisburg ..................................................................................5
2.3 Desired values ............................................................................................................5
2.4 Controller ....................................................................................................................6
2.5 Operation Algorithm ....................................................................................................7

3. Design of the Electronic Stability Program ...........................................................................8


3.1 Controller ....................................................................................................................8
3.2 Desired values ............................................................................................................8
3.3 Brake Pressure Allocation ...........................................................................................9

4. Experiments .......................................................................................................................14
4.1 Test cases ................................................................................................................14
4.2 Test maneuver ..........................................................................................................15
4.3 Test 1: Vehicle without ESP ......................................................................................15
4.4 Test 2: Vehicle with ESP control torque ....................................................................17
4.5 Test 3: Vehicle with improved ESP algorithm vs. ESP Duisburg ...............................18
4.6 Test 4: Redesigned ESP vs. ESP Duisburg ..............................................................21

5. Conclusion and Recommendations ....................................................................................23


5.1 Conclusion ................................................................................................................23
5.2 Recommendations ....................................................................................................23

List of Symbols.......................................................................................................................24

2
1. Introduction
Weather conditions as well as critical situations on the road can cause a driver to be unable to
control behavior of his or her vehicle. In this case, the Electronic Stability Program (ESP)
helps to stabilize vehicle movement. Driver assistance systems like ESP which improve a
cars active safety are on the market for about two decades and have almost become
standard equipment. Thorough testing and development have made these systems into
lifesavers that can assist the driver in most hazardous situations. Contrary to the actual car, in
the digital testing and simulation environment, ESP is not always included as such. However,
because ESP has become a very important and well recognized safety system, it is useful to
include this system as a parameterizable stand-alone unit in vehicle simulation software. This
feature improves opportunities to simulate the entire vehicle behavior in the development and
testing phase. In this phase it is not possible to implement the actual ESP, because its
algorithm is not freely available. The investigation presented in this report is aimed at
obtaining a stand-alone ESP for the virtual-realistic car simulator Ftronik. This car simulator
runs on veDYNA vehicle simulation software and a highly realistic BMW 5-series dataset. The
goal for the investigation is formulated as follows.

Design, build and test an effective and efficient Electronic Stability Program for the veDYNA
vehicle simulation package

veDYNA
veDYNA vehicle simulation software is widely used in industry and is respected for its
precision and the fact that it can be implemented with ease in Hardware-in-the-Loop (HiL) and
Software-in-the-Loop (SiL) test benches. In contrast to other software, the veDYNA package
is built in a Matlab/Simulink environment.
The successor of veDYNA3 is DYNA4. This simulation framework covers the entire
range of vehicle simulation. In addition to the vehicle dynamics simulation package, the
framework includes engine and drive train simulation as well as a traffic simulator. The
simulation results are shown in both a graphical animation and detailed figures that show
plots of variables which are free to choose.
During the traineeship the scope of the investigation has changed from application
just for the Ftronik car simulator to general application as an add-on in veDYNA software.

Contact information TESIS DYNAware GmbH


TESIS DYNAware GmbH
Baierbrunnerstrae 15
81379 Mnchen, Germany

3
2. ESP
E Gen
neral and
d Startin
ng Pointt
The Electronic
E Sttability Progra
am (ESP) iss originally developed
d byy the German car part
supplier Bosch and d car manufaacturer Merce edes-Benz in the early 198 80s. Its purp
pose, as its
namee implies, is to
o increase vehhicle stability under critical circumstancees.

2.1 Critical
C be
ehavior
Critica
al dynamical vehicle beha
avior can be divided into two
t situationss, excessive understeer
and excessive
e ove
ersteer. Underrsteer and oversteer generrally are car specific
s prope
erties.

Underrsteer
Whenn, to maintain
n a certain curvature
c with constant radius
r during increasing speed,
s the
gle has to be increased, the vehicle is said to be understeered.
steering wheel ang

steer
Overs
Whenn, to maintain a constant ra easing speed, the steering
adius at incre g wheel angle
e has to be
decre
eased, the vehhicle is called oversteered..

Not only
o this bas
sic chassis la ayout of the vehicle dete ermines vehiicle behaviorr under all
condittions. In certain cases, vehicle beh havior can be
b provoked d, i.e. by inffluence of
accele eration or deceleration in a corner or load transfer. Excessive understeer
u ho
owever is a
stablee situation whhere grip on thhe front axle is reduced an
nd the car slides curve outward over
its fro
ont tyres. On the
t contrary, excessive ovversteer is an unstable situ uation where grip on the
rear axle
a is lower and
a the car tu urns curve inwward and tend
ds to spin aro
ound its axis. These two
situations are show wn in figure 2..1 and figure 2.2 below.

Figure 2.1 Excessive un


ndersteer Figure 2.2 Excessive overssteer

This controller
c usu
ually controls 2 motion va ariables, vehicle side slip angle and vehicle yaw
velociity. In a real vehicle espe s hard to determine and has to be
ecially the firrst variable is
estimaated. In this study, vehiccle side slip angle is a variable directly available e from the
veDYNA vehicle model.
m
2.2 ESP Controller Essen Duisburg
As starting point for the design of an ESP controller, at hand there is a design of University
Essen Duisburg. This design can be divided into 3 parts;

calculation of desired values


controller
operation algorithm

These three parts are subsequently shortly explained in the following sections.

2.3 Desired values


The original controller includes 2 Single Track Models (STM), one to determine the desired
values and one to provide a state space model of the vehicle. This means every parameter
that has to be defined by the user in these models, occurs twice. The models are not exactly
equal; the cornering stiffnesses of both models can be adjusted to achieve certain behavior.
The first STM is used to calculate the desired values for vehicle side slip angle and
vehicle yaw velocity. Under steady state conditions this model, whose parameters are
determined by experiments, delivers fairly acceptable values which can be used as an input
variable for the controller.
Following equations are used to determine desired values for yaw velocity and side slip angle.

Desired yaw rate

L v
& des =
iL l v2
1+ 2
vch

With
L
iL wheel toe angle due to steering wheel angle

Maximum desired yaw rate

s g
& des max = 0.8
v (2.1)

Desired vehicle side slip angle

lR ml F v 2
des =
C 2 l

With
lR distance CoG to rear axle
lF distance CoG to front axle
l wheel base
v
= inverse corner radius
&
C 2 cornering stiffness rear axle

5
Empirical relation to determine maximum vehicle side slip angle

des max = tan 1 (0.02 s g ) (2.2)

With
s friction coefficient tyres to road
g gravitational acceleration

These relations for desired values in side slip angle and yaw rate lead to the following
formulation of the error vector, which serves as input for the controller.

act des
e=
& act & des

2.4 Controller
A Linear Quadratic (LQ) Optimal Controller is based on the following state space description
of vehicle behavior, which is deduced from the single track model with state variables and
& .

C 1 + C 2 C 2lR C 1lF
1+ 0
mv mv 2
C 2lR C 1lF C l + C 2lR2
2
1 F
Jz J zv

C 1 + C 2 C 2 l R C 1l F
mv
1+
mv 2
A= 2
C 2 l R C 1l F C l + C 2 l R
2
1 F
Jz J zv

0
E = 1


J z

e& = Ae + EM
M = Ke (2.3)

6
The LQ Controller has to be chosen such that it minimizes


(
J = e T Qe + RM 2 dt)
0

With
1 2
0
lim
Q=
1
2

0
lim&
2
1
R=
lim M
Where the limits are weighting factors, free to choose to influence controller output. Essen
Duisburg proposed to use user defined maximum values for respectively side slip angle, yaw
rate and output control torque for these weighting factors. Limit factors lim and lim& in
this case are chosen equal to the maximum vehicle side slip angle (2.2) and vehicle yaw
velocity (2.1), respectively.

The controller calculates the control torque M by solving the Riccati equation every time
step, due to non linearity caused by changing vehicle speed and steering wheel angle. This
linearizes the system every time step and thereby allows for the linear observer.

The actual control torque which is required to stabilize the vehicle is determined by (2.3). This
control torque is subsequently used in an operation algorithm to prescribe brake pressures to
the ABS brake system.

2.5 Operation Algorithm


The operation algorithm is based on basic theory stating oversteer requires an ESP
intervention on the outer front wheel, whereas in case of understeer the inner rear wheel is
braked. Inner and outer wheels of course are determined by the direction of the curve. At first
two boundary conditions are checked, vehicle speed should be above 5 m/s and control
torque has to be larger than some threshold. Then a distinction is made between several
cases. The sign of the steering wheel angle is used to recognize left and right curves,
whereas the error in yaw velocity distinguishes between understeer and oversteer. In each of
these cases, brake pressures are calculated by incrementally increasing or decreasing brake
pressure for the chosen wheel as long as a control torque exists.

7
3. Design of the Electronic Stability Program
Electronic Stability Program
Brake
State Space Description Pressures
LQR Controller Wheel FR

Operation Algorithm
Wheel FL
-
Desired Side Slip Angle K1 Control
Wheel RR
V Desired Yaw Velocity K2 Torque
- Brake Pressure Allocation
Wheel RL
& act

act

Figure 3.1 Schematic representation of the ESP

A schematic representation of the Electronic Stability Program is show in figure 3.1 above. It
shows the two controlled variables, the controller which provides controller gains K1 and K2
and the operation algorithm and brake pressure allocation. In this study, the focus is on
improving this last sub block.

3.1 Controller
As the controller of the design at hand already is influenced by both yaw rate and side slip
angle and optimizes the control torque to comply to both variables, this controller is used in
the new ESP design. Correct operation of the controller is supported by simulation results,
presented in chapter 5.4.

3.2 Desired values


Instead of using the output of a Single Track Model, the desired values for yaw velocity and
side slip angle are measured in a single measurement and stored in a 2D look-up-table. To
be sure to take variables into account that influence these 2 reference values for the controller,
measurements are carried out. Results of these measurements are that both yaw velocity and
side slip angle are mainly influenced by vehicle velocity and steering wheel angle. Therefore,
these 2 variables as well as the 2 output variables are measured in a semi-steady-state
cornering experiment in which speed varies from 30 to 180 [kmh] and steering wheel angles
vary from 0 to 720 [deg]. Afterwards, the results are mirrored to cover all possible situations.
The advantage of the direct use of measurements is not only that the actual data of
the vehicle is used, but also parameterization of the controller becomes more clear. The use
of actual vehicle data improves parameter correspondence to the actual vehicle behavior,
especially at high lateral acceleration. One of the 2 Single Track Models that determine ESP
behavior is now obsolete.

8
3.3 Brake Pressure Allocation
Whereas the operation algorithm of the ESP Duisburg is based on the rule of thumb
described in chapter 2.5, here a second look is taken at the way that this ESP distinguishes
between cases and determines which intervention is done. The basic theory behind this
analysis is the moment equilibrium around the z-axis. The longitudinal, braking and driving,
and lateral forces each create a torque around the z-axis, leading to the equilibrium around
the centre of gravity in (3.1).

M z = Fpx p x + Fqx q x Frx rx + Fsx s x + Fpy p y + Fqy q y Fry ry Fsy s y (3.1)

where the lever arms are defined as shown in figure 3.2.

Figure 3.2 Definition of lever arms

Lever arms for the rear axle are constant, lever arms for the front axle however are variable
and depend on the wheel toe angles. The constant lever arms are defined as

TR
rx =
2
ry = l R
TR
sx =
2
sy = l R

Where TR is the track width rear and l R is the distance from the centre of gravity to the rear
axle in x-direction.

9
Figure 3.3 shows the variable lever arms for the front axle.

Figure 3.3 Variable lever arms front axle due to toe angle

Depending on the sign of the steering wheel angle, the lever arms are calculated as follows.

Left corner

L + R
=
2
TF2 2
aa = + lF
4
TF

= tan 2
1

lF

TF = TF cos( )
px = aa sin( )
py = aa cos( )
qx = TF px
qy = lF sin( )

Right corner Figure 3.4 Variables used in calculation of lever arms front axle

px = TF qx
py = lF sin( + )
qx = aa sin( + )
qy = aa cos( + )

where figure 3.4 shows the respective variables.

10
Brake forces are defined in negative x-direction and are always positive. As shown in (3.2),
brake forces have a direct influence on the moment equilibrium around the z-axis and
therefore can generate a control torque. The control torque created directly by the brake
forces is called primary control torque.

M PC = FBP p x FBQ q x + FBR rx FBS s x (3.2)


with FBi with i=p,q,r,s the respective brake forces for each wheel.

In addition to the primary control torque it is possible to create a second torque around the z-
axis. This is called the secondary control torque and does occur when the side force of the
current wheel is reduced by applying a brake force. This effect however can only be
generated when the tyre currently close to the friction limit. This is shown in figure 3.5 as the
red outline of the friction circle of a tyre.

Figure 3.5 Tyre friction circle

When the tyre is on the edge of slip-slide during pure cornering, the side force consumes all
transferrable force at that moment in time. This is shown in figure 3.6 below. If a brake force is
applied, the available total force is divided between a longitudinal brake force and the side
force. This causes the available side force to reduce, which is shown in figure 3.7.

Figure 3.6 Pure cornering Figure 3.7 Combined cornering and braking

The reduction in side force is a function of the brake force and maximum tyre force, which
comes from the veDYNA vehicle model, and introduces an additional torque around the z-axis.

11
F y = F max (F 2
max F b2 )
M SC = F py p y Fqy q y + Fry ry + Fsy s y (3.3)

Combining (3.2) and (3.3) gives the total available control torque which can be achieved by
braking.

M CT = FBP p x FBQ q x + FBR rx FBS s x F py p y Fqy q y + Fry ry + Fsy s y (3.4)

As Fy is only a function of the semi-constant Fmax and the applied brake force FB , control
torque M CT is only a function of variable FB .

To make a correct decision in which wheel is braked, for each wheel a sensitivity is
calculated. The sensitivity shows how much brake force is required for that wheel to produce
the control torque calculated by the controller. As an example, the calculation of the sensitivity
for wheel P is shown below.

(
M CT = FBP p x Fmax (F 2
max FBP
2
))
py (3.6)

(3.6) can be solved for FBP by solving quadratic equation (3.7)

p 2 2
x 1 F 2 2 p x Fmax P + 2 M CT p x
FBP + M CT

+ 2 M CT Fmax P = 0 (3.7)
p BP py p y2 p py
y y

This required brake force FBP now shows how much brake force is required for only
wheel P to produce the entire control torque. As this is done for every wheel, the 4 required
brake forces can be sorted and the wheel that requires least brake force is chosen to carry
out the ESP intervention. This guarantees that the intervention is done on a wheel which is
able to produce the control torque and additionally this is the wheel that requires the least
brake force, so that comfort is optimized. In certain cases it is possible that not the entire
control torque can be produced by a single wheel ESP intervention. This means there is a
remaining control torque after applying maximum allowable brake force on the first selected
wheel. Then the wheel requiring second least brake force is additionally braked.
The way a wheel is braked is not always the same. As shown before, there is a
division between primary and secondary control torque. Not in every case the secondary
control torque is useful in producing the required control torque. As shown in (3.4), the
secondary effect has an opposite sign to the primary effect for wheel P and S. This is
dependent on the direction of the curve and the sign of the control torque, so whether
understeer or oversteer has to be corrected. In practice this means that a check is necessary
to determine whether it is useful to exploit the entire potential of primary and secondary
control torque for each wheel. If the secondary control torque produced is useful, the wheel is
braked into the sliding region. If not, a safety threshold is used to make sure the wheel is kept
entirely in the slipping region. This means ESP has to work together with the Antilock Braking
System (ABS) to make sure these operations are possible.

12
In addition to the ESP intervention by applying a brake force, to effectively control
vehicle behavior it is necessary to exclude any interference by the engine driving the front,
rear or all wheels. This is done by cutting off engine torque during an ESP intervention in 3
phases.

Phase 1
Phase 1 starts at the moment an ESP intervention is initiated. At this moment the engine
torque is reduced to 0 by either reducing or increasing the throttle angle. It has to be noted
that engine torque is negative, engine braking, when the throttle valve is completely closed,
so a small angle has to be kept in order to compensate for internal losses in the engine such
as friction and changes in entropy/enthalpy.

Phase 2
Phase 2 is used to make sure the engine torque is cut off for at least 200 ms or as long as the
ESP system is active.

Phase 3
During phase 3 the maximum engine torque is raised back to its original value over 500 ms.
This is done linearly and combined with the 2 previously described phases ensures smooth
operation of the ESP system in the complete vehicle model.

The ESP system is implemented and tested in Matlab/Simulink.

13
4. Experiments

4.1 Test cases


As testing is a crucial part of the assignment, this chapter includes tests with intermediate
designs that are not explained in detail in this report. For convenience, a short description of
the versions that are tested is presented below.

1. Vehicle without ESP


A vehicle model that is not equipped with ESP serves as a reference to both assess
the maneuver in its ability to show critical vehicle behavior and demonstrate the
functionality of an ESP.

2. Vehicle with control torque


In this case, the control torque calculated by the controller is applied directly to the
vehicle. In the veDYNA vehicle model, this is possible by applying the control torque
as an external input around the z-axis. This model shows whether the control torque
in itself is able to stabilize vehicle behavior or not.

3. Vehicle with ESP with improved algorithm vs. ESP Duisburg


Some errors are detected while thoroughly assessing the algorithm that determines
ESP interventions in the starting point ESP. The changes made concern the way left
curves are distinguished from right curves and the distinction between understeer and
oversteer. In the starting point ESP, left and right curves are detected by evaluating
the sign of the steering wheel angle. As during a critical maneuver the driver reacts by
counteracting movement by steering in the opposite direction, the sign of the steering
wheel angle can be opposite to the actual curve that is followed. In the improved
algorithm, this distinction is made by evaluating the sign of the lateral acceleration.
In the improved algorithm, understeer and oversteer are no longer detected by
looking at the error in yaw velocity, but by evaluation of the sign of the control torque.
As the control torque incorporates both yaw velocity and side slip angle, this variable
is better suited.

4. Vehicle with new ESP design vs. ESP Duisburg


This test shows the performance of the design presented in chapter 3 compared to
the ESP designed by University Essen Duisburg.

14
4.2 Test maneuver
To verify the operation of the ESP system a double lane change maneuver is commonly used
and gives insight in vehicle behavior under circumstances that drives the cars chassis to its
limits. The principle of this test is shown in figure 4.1

Figure 4.1 ISO 3888-1 double lane change

To show the differences between the several designs, a test is done at 100 km/h. The driver
used during this maneuver is adjusted such that driver behavior does not interfere with the
expected ESP results. If, for example, the driver is able to counteract oversteering by a
steering wheel correction, it is impossible to identify ESP influence on vehicle behavior.
Therefore the steering wheel rate is limited to 270 deg/s. In the assessment of results, focus
is on vehicle stability, required control torque and applied brake pressure.

Table 4.1 Boundary conditions maneuver


Variable Value
Vehicle speed 100 [kmh]
Max. steering wheel angle rate 270 [deg/s]

4.3 Test 1: Vehicle without ESP


As a reference, the test is first performed with a vehicle that is not equipped with an ESP
system. The vehicle, a BMW 530i E60, has a basic chassis design which under steady state
conditions tends to understeer. However during this maneuver the vehicle is supposed to
become unstable.

15
Figure
e 4.2 and 4.3
3 below show w both side sllip angle and vehicle traje
ectory during the double
lane change
c maneeuver. It is clear that beha
avior is not as
a desired an nd the vehicle
e becomes
unstable as expectted.

Figure 4.2 Side slip ang


gle [deg] vs. simu
ulation time

Figure 4.3 Vehicle traje


ectory
4.4 Test
T 2: Ve
ehicle with
h ESP con
ntrol torqu
ue
In ord
der to test thee working prin nciple of the ESP system, a test is don ne where the calculated
controol torque is ap t control the vehicle. In th
pplied directlyy into the chasssis in order to his way it is
proveen whether the e control torque actually iss able to stabilize the vehiccle or not. Figures below,
showiing side slip angle, steerring wheel angle a and veehicle trajecto
ory, clearly provide
p the
requirred confirmatiion that the caalculated conntrol torque is able to stabilize vehicle beehavior.

Figure 4.4 Side slip ang


gle [deg] vs. simu
ulation time for E
ESP control torqu
ue

Figure 4.5 Steering whe


eel angle [deg] vs.
v simulation tim
me for ESP contro
ol torque
Figure 4.6 Vehicle traje
ectory for ESP co
ontrol torque

4.5 Test
T 3: Ve
ehicle with
h improve
ed ESP alg
gorithm vs
s. ESP Du
uisburg
As deescribed in ch hapter 4.1, ESP
E operation n algorithm is
s improved inn order to ob
btain better
vehiclle stabilizatio
on. Figures below
b show differences between
b the starting point ESP of
Unive
ersity Duisburg g and the imp
proved algoritthm for key va
ariables.

Figure 4.7 Side slip ang


gle [deg] vs. simu
ulation time for E
ESP Duisburg and improved algorithm
Figure 4.8 Steering whe
eel angle [deg] vs.
v simulation tim
me for ESP Duisb
burg and improve
ed algorithm

Figure 4.9 Vehicle traje


ectory for ESP Du
uisburg and imprroved algorithm
Figure 4.10 Brake pressure front left [Pa
a] for ESP Duisb
burg and improve
ed algorithm

Figurees above sho ow improvemments in all diisciplines, thee vehicle morre effectively follows its
track, while requiring less stee
ering wheel in nput and all in
i all also lesss brake presssure. It is
shown n clearly by the
t combination of vehicle e trajectory an
nd brake presssure that thee vehicle is
stabiliized about 2 seconds
s earliier with the im
mproved algorrithm compare ed to the starrting point.
4.6 Test
T 4: Re
edesigned
d ESP vs. ESP Duisburg
After during the reference test iti is shown th
hat vehicle be
ehavior becom mes unstable during the
double lane change maneuver, this test is pe erformed withh both the ESP system pro ovided from
Univeersity Duisburrg and the ESSP system described in th his report. First, vehicle trrajectory is
compared.

Figure 4.11 Vehicle trajjectory compared


d between ESP D
Duisburg and ES
SP 2

Figuree 4.11 shows that ESP 2 iss able to conttrol vehicle tra
ajectory more
e precisely tha
an the ESP
from University Duuisburg. Also o it is shown that although an ESP syystem is activve, still an
oversteer reaction is expected between x=300 and x=350 0 m. Assessmment of the vehicle side
ngle gives mo
slip an ore clarity abo
out this obserrvation in figure 4.12
Figure 4.12 Vehicle side slip angle [deg
g] vs. simulation ttime

arger than 5 degrees indiccate potential oversteer.


Absolute values off vehicle side slip angles la
In thiss case, the animation inde
eed shows an n oversteeredd vehicle. It iss clear that th
he changes
made e to the ESP system
s provid
de a more reliable vehicle stabilization.
s

Furtheermore, the applied


a brake pressures arre evaluated. Figure 4.13 shows brake e pressures
for the
e left front wh
heel, which is representativve for the ove
erall required b
brake force and
a thereby
experrience of comfort.

Figure 4.13 Wheel brakke pressure frontt left vs. simulatio


on time

Maximmum brake pressures are two times ass small as in the initial cooncept, also im mplying no
ABS intervention is required. Combined
C with the lowerr brake force, this largelyy increases
g comfort durring ESP interrvention.
driving
5. Conclusion and Recommendations
5.1 Conclusion
In this report, an Electronic Stability Program is presented that is implemented as a stand-
alone subsystem in the veDYNA vehicle simulation package. The ESP has following main
attributes.

Vehicle control using a Linear Quadratic Optimal controller which is calculated every
time step

Two variables are controlled, vehicle yaw velocity and vehicle side slip angle by
applying brake pressures to the individual wheels

Desired values for the two controlled variables are determined based on semi-steady
state measurements on the actual vehicle

Applied brake forces are calculated exactly, based on the control torque produced by
the controller, variable lever arms of brake forces and current slip-slide condition of
the tyres

By assessing the sensitivity of each wheel to create the required control torque, a
decision is made on where the ESP intervention takes place

In application of brake pressures both primary and secondary effects of brake forces
are taken into account

The controller is implemented in a Matlab/Simulink environment

In experiments and simulation it is shown that the ESP is effective and efficient in controlling
vehicle behavior under critical circumstances. A comparison is made between the redesign
and the starting point design delivered by University Essen Duisburg. This comparison shows
that at a double lane change test maneuver at 100 [kmh], the redesign of the ESP is able to
entirely keep vehicle behavior within stable boundaries, whereas the ESP of University Essen
Duisburg still shows unstable vehicle movement. In addition to better stabilization, the
required peak brake forces are only half as large.

5.2 Recommendations
Below 30 [kmh] the ESP is not active. Under normal conditions, this is not considered a
problem. At specific conditions such as a road with a low friction coefficient however, this can
still cause undesired vehicle behavior. It is recommended to execute more experiments in
order to better investigate vehicle yaw velocity and side slip angle below 30 [kmh].

Dynamic tyre behavior and vehicle chassis dynamics are not taken into account in the
calculation of desired values. In certain cases, this causes the ESP to become active during
swift alterations in direction. Possibly a more detailed vehicle model as observer or look-up-
tables that take dynamics into account, are able to solve this issue.

Controller parameterization is unclear because the parameters ruling controller behavior have
no physical meaning. Iteratively an optimum for the division between yaw velocity and side
slip angle influence can be found. Possibly these controller parameters have to be chosen
variable to vehicle speed and/or steering wheel angle.

Other tests can be performed to examine behavior of the ESP system under more
circumstances. Although the double lane change maneuver identifies critical vehicle behavior
in most cases, it does not cover every situation that can be encountered. This additional
testing could include the J-turn maneuver, fishhook maneuver and step steer response, as
well as testing on road surfaces with low friction coefficient.

23
List of Symbols

vehicle side slip angle


steering wheel angle
s friction coefficient
inverse corner radius
& vehicle yaw velocity

C effective cornering stiffness


FB brake force
Fmax maximum tyre force
g gravitational constant
i steer ratio
Jz inertia around z-axis
l wheel base
lF distance front axle to centre of gravity
lR distance rear axle to centre of gravity
lim weighting factor, maximum allowable vehicle side slip angle
lim& weighting factor, maximum allowable vehicle yaw velocity
lim M z weighting factor, maximum allowable control torque around z-axis
M CT overall control torque
M PC primary control torque
M SC secondary control torque
Mz moment around z-axis
p wheel index front left
q wheel index front right
r wheel index rear left
s wheel index rear right
TF track width front
TR track width rear
v vehicle velocity
vch vehicle characteristic velocity

24

Das könnte Ihnen auch gefallen