Dynamic Model and Control

of Vehicles

By: Hsueh-Yuan Lu

Supervised by: Dr. Meilan Liu

A Thesis submitted to the Faculty of Graduate Studies

in partial fulfillment o f the requirements for the

Master o f Science in Engineering

Control Engineering

Lakehead University
September, 2006

ii

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Library and Bibliotheque et
Archives Canada Archives Canada

Published Heritage Direction du

Branch Patrimoine de I'edition

395 W ellington Street 395, rue W ellington

Ottawa ON K1A 0N4 Ottawa ON K1A 0N4
Canada Canada

Your file Votre reference

ISBN: 978-0-494-21534-0
Our file Notre reference
ISBN: 978-0-494-21534-0

NOTICE: AVIS:
The author has granted a non­ L'auteur a accorde une licence non exclusive
exclusive license allowing Library permettant a la Bibliotheque et Archives
and Archives Canada to reproduce, Canada de reproduire, publier, archiver,
publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public
communicate to the public by par telecommunication ou par I'lnternet, preter,
telecommunication or on the Internet, distribuer et vendre des theses partout dans
loan, distribute and sell theses le monde, a des fins commerciales ou autres,
worldwide, for commercial or non­ sur support microforme, papier, electronique
commercial purposes, in microform, et/ou autres formats.
paper, electronic and/or any other
formats.

The author retains copyright L'auteur conserve la propriete du droit d'auteur

ownership and moral rights in et des droits moraux qui protege cette these.
this thesis. Neither the thesis Ni la these ni des extraits substantiels de
nor substantial extracts from it celle-ci ne doivent etre imprimes ou autrement
may be printed or otherwise reproduits sans son autorisation.
reproduced without the author's
permission.

In compliance with the Canadian Conformement a la loi canadienne

Privacy Act some supporting sur la protection de la vie privee,
forms may have been removed quelques formulaires secondaires
from this thesis. ont ete enleves de cette these.

While these forms may be included Bien que ces formulaires

in the document page count, aient inclus dans la pagination,
their removal does not represent il n'y aura aucun contenu manquant.
any loss of content from the
thesis.
i*i

Canada
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 ABSTRACT

In this thesis the author develops a 14 degrees-of-freedom (DOF) full-car model. The

model draws from and improves upon features and setups of certain existing vehicle

dynamics models. The proposed model provides a means to simulate vehicle ride and

handling behaviors. An accurate prediction o f such behaviors will lead to the proper

control and design of vehicles.

The vehicle’s kinematics and dynamics are developed to reflect the interactions

between the rigid mass elements of the model such as the vehicle body and the wheels.

The mathematical model includes the nonlinear characteristics of the tires, the three

dimensional motions of the sprung and unsprung masses, the inertial coupling between

the sprung and unsprung masses, and the restraints and forces imposed by the suspension

components. The frictional forces developed at the road-tire contacts are modeled by the

single point contact version of the Lund-Grenoble (LuGre) dynamic friction model. An

extension of the LuGre friction model is presented to take into account the coupling

between the rotational and translational motions of the wheels.

Three different numerical study cases are selected to verify the model’s capability in

representing various vehicle dynamic situations with respect to the model’s accuracy and

to the m odel’s range of applicability.

iii

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 The issue of active suspension is subsequently discussed. A non-switching sliding

mode controller is incorporated into the proposed vehicle model and a substantial

reduction in the spectral intensity of a vibration mode of the vehicle body is achieved.

Simulation results suggest that the rigorous modeling and mathematical

development yields a model that captures satisfactory ride comfort and vehicle

performance.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 ACKNOWLEDGEMENTS

I would like to thank the following people for their help and support during the past two

years of study at Lakehead University.

First and foremost, my deepest gratitude goes to my advisor, Dr. Meilan Liu, for

guiding my work with valuable insights and detailed comments, for working closely with

me and for providing support and encouragement during the course o f this research

program.

I am very grateful to my co-advisor, Dr. Xiaoping Liu, for his valuable suggestions

and help on my research topic.

I am also very grateful to Dr. Abdelhamid Tayebi for the valuable lectures that

introduced me to the modern control concepts.

Special thanks go to the examiners, Dr. Alexander Sedov (internal) and Dr. Yuping

He (external), for devoting their effort and time to review and make suggestions to my

work and helping me to finalize the details.

My sincere thanks go to all my fellow graduate students, for creating a pleasant

working environment. Their friendship and support has made my Lakehead experience

very enjoyable. Thanks also go to the staff members of the M aster’s Program in Control

Engineering who have helped me in many ways.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Financial support, received by my advisor from the Natural Sciences and

Engineering Research Council of Canada (NSERC) that supported this work is gratefully

acknowledged.

Finally, I would like to thank my family members and my fiancee, W eilin W ang, for

loving me and supporting my work all these years.

vi

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 TABLE OF CONTENTS

CERTIFICATE OF EXAMINATION
ABSTRACT iii
ACKNOWLEDGEMENTS v
TABLE OF CONTENTS vii
LIST OF TABLES ix
LIST OF FIGURES x
LIST OF APPENDIX xv

CHAPTER 1 INTRODUCTION

1.1 Vehicle Dynamics - A Historical Perspective 1

1.2 Objectives 3
1.3 Organization o f the thesis 3
1.4 Relevant Terminologies 5

CHAPTER 2 OVERVIEW OF VEHICLE DYNAMICS MODELS

2.1 Existing Vehicle Dynamics Models 7

2.2 The Proposed Vehicle Dynamics Model 14

CHAPTER 3 THE 14-DOF FULL-CAR DYNAMIC MODEL

3.1 Coordinate Systems 15

3.2 The 14-DOF Full-Car Model 19
3.3 Kinematics of Vehicle Body and Wheels 22
3.4 Dynamics of Vehicle Body and Wheels 26
3.5 Simplification of the 14-DOF Full-Car Model 41
3.6 Concluding Remarks 47

CHAPTER 4 ROAD-TIRE FRICTION

4.1 Literature Review 49

4.2 The LuGre Dynamic Friction Model 52
4.3 Steady-State Characteristics 56

vii

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 4.4 Validation of Steady-State Behavior with the Magic
61
Formula
4.5 Incorporating the LuGre Friction Model 63
4.6 Concluding Remarks 68

CHAPTER 5 NUMERICAL STUDIES OF VEHICLE DYNAMICS

5.1 A Planar Model of a Three-W heeled Vehicle 71

5.2 A 7-DOF Full-Car Model 79
5.3 A Full-Car Model for M aneuver Simulation 86
5.4 Conclusions 114

CHAPTER 6 SLIDING MODE CONTROL

6.1 Background 118

6.2 Vehicle Model (A Pitch-Bounce Half-Car) 121
6.3 Sliding M ode Controller Design 123
6.4 Simulations and Discussions 126
6.5 Conclusions 139

CHAPTER 7 CONCLUSIONS AND RECOMMENDATIONS

7.1 Conclusions 140

7.2 Recommendations 144

APPENDIX A DERIVATION OF VELOCITIES AND

147
ACCELERATIONS

REFERENCES 149

viii

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 LIST OF TABLES

Table 4.1 Magic Formula Parameters 63

Table 4.2 LuGre Model Parameters 63

Table 5.1 Parameters for LuGre Friction Model 73

Table 5.2 Parameters for the TWV 74

Table 5.3 Parameters s and R 80

Table 5.4 Model Parameters 83

Table 5.5 Model Parameters 91

Table 6.1 System Parameters 122

ix

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 LIST OF FIGURES

Figure 1.1 Vehicle Motions 5

Figure 2.1 Quarter-Car Models [2.3] 8

Figure 2.2 A 2-DOF Quarter-Car Model [2.4] 9

Figure 2.3 Half-car model involving pitch and roll motion 10

Figure 2.4 Planar Model of a Three-W heeled Vehicle [2.6] 11

Figure 2.5 Full-Car Model of A Three-W heeled Vehicle [2.7] 12

Figure 3.1 Vehicle body frame of reference 16

Figure 3.2 Wheel center and wheel base coordinate frames of reference 18

Figure 3.3 The four frames of reference used 19

Figure 3.4 The full-car model 19

Figure 3.5 Axle roll center bounce and axle rotation as DOFs 21

Figure 3.6 Schematic of the 14-DOF full-car dynamics model (spins not
21
shown)
Figure 3.7a Side view of the vehicle body showing —x n sin 7 24

Figure 3.7b Front view of the vehicle body showing yn sin 0 24

Figure 3.7c Locations of wheel centers in the vehicle body frame of

25
reference
Figure 3.8 FBD of vehicle body (showing forces and torques from
28
wheels 2 and 3 only)

Figure 3.9a FBD of the vehicle body as viewed along the x-axis 31

Figure 3.9b FBD of the vehicle body as viewed along the y-axis 31

Figure 3.9c FBD of the vehicle body as viewed along the top 32

Figure 3.10 Vehicle body as viewed along the y-axis showing H, z and zn 32

Figure 3.11 Forces exerted by road surface onto the wheel bases 33

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Figure 3.12 FBD of a wheel 34

Figure 3.13 Free body diagram o f a wheel (showing only forces in the
36
z-direction)

Figure 3.14a [A] Matrix as inputted into MAPLE 40

Figure 3.14b [A] '1 Matrix as determined by M APLE 40

Figure 3.15 A quarter-car model [3.13] 43

Figure 3.16 A half-car model showing pitch and bounce motions [3.14] 44

Figure 3.17 A half-car model showing roll and bounce motions [3.15] 45

Figure 3.18 A 7-DOF full-car Model [3.12] 47

Figure 4.1 The average internal deflected state rj in the direction of

53
motion

Figure 4.2 The effect on g (Vr ) due to changes in friction parameters 54

Figure 4.3a Slip for braking 59

Figure 4.3b Slip for acceleration 59

Figure 4.4 Steady-state friction forces Fx(s) and Fy(fi) 63

Figure 4.5 a Bristle deformation before rotation 64

Figure 4.5b Bristle deformation after rotation 64

Figure 4.6 Total bristle deflection with respect to the global frame 66

Figure 4.7 Vehicle steering geometry definition [4.10] 69

Figure 5.1 The planar TW V model 71

Figure 5.2 Half-sine wave road profile 74

Figure 5.3 Normal reaction at front/rear wheel 75

Figure 5.4a Vertical acceleration of the vehicle body - Time history 76

Figure 5.4b Vertical acceleration of the vehicle body - Frequency
76
spectrum

xi

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Figure 5.5 Comparisons of the present model and Ref. [5.1] 78

Figure 5.6 Frictional forces 78

Figure 5.7 Dry friction behavior 80

Figure 5.8 Ramp-like bump profile 82

Figure 5.9 Time-delay at 10 m/s 82

Figure 5.10a Vertical displacement o f vehicle body 84

Figure 5.10b Pitch angle of vehicle body 84

Figure 5.10c Vertical acceleration o f vehicle body 85

Figure 5.11 Damper friction forces 85

Figure 5.12 Steering geometry definition (showing right steering) [5.3] 87

Figure 5.13 Driver steering input 89

Figure 5.14 Longitudinal/Lateral friction forces 92

Figure 5.15 Vehicle body speeds for pavement road with no torque input 93

Figure 5.16 W heels’ rotational speeds with no torque input 95

Figure 5.17 Longitudinal/Lateral relative velocity of each wheel with no

96
torque input

Figure 5.18 Longitudinal/Lateral bristle deflection at each wheel 97

Figure 5.19 Longitudinal/Lateral dynamic friction coefficients 98

Figure 5.20 Normal forces during turn maneuver 99

Figure 5.21 Vehicle position with no torque input 100

Figure 5.22a Comparison of vehicle body speeds 102

Figure 5.22b Comparison of vehicle position 103

Figure 5.23 Comparison of longitudinal/lateral frictional forces 103-104

Figure 5.24 Comparison of longitudinal/lateral relative velocities 104-105

xii

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Figure 5.25a Vehicle body speeds, no torque input 106

Figure 5.25b Vehicle position, no torque input 107

Figure 5.26a Vehicle body speeds, 50 N.m input torque 107

Figure 5.26b Vehicle position, 50 N.m input torque 108

Figure 5.27a Normal forces for turn maneuver with no torque input 108

Figure 5.27b Normal forces for turn maneuver with 50 N.m torque input 109

Figure 5.28 Longitudinal/Lateral friction forces, no torque input 109-110

Figure 5.29 Longitudinal/Lateral friction forces, 50 N.m torque input 110-111

Figure 5.30 Comparison o f vehicle positions, no torque input 112

Figure 5.31 Comparison o f vehicle positions, slippery road 113

Figure 5.32 Vehicle spins out of control, 100 N.m torque to rear wheels 113

Figure 5.33 Lateral frictional forces, 100 N.m torque to rear wheels 114

Figure 6.1a Schematic of a passive suspension system 120

Figure 6.1b Schematic of an active suspension system 120

Figure 6,2 A single ramp-step bump 126

Figure 6.3a Vehicle body bounce 128

Figure 6.3b Vehicle body pitch 129

Figure 6.3c Acceleration of vehicle body bounce 129

Figure 6.3d Acceleration of vehicle body pitch 130

Figure 6.4 Control input forces at front/rear suspension 130-131

Figure 6.5 Frequency spectrum of body bounce/pitch accelerations 131-132

Figure 6 .6a Vehicle bounce acceleration - Change in sprung mass 133

Figure 6 .6b Vehicle bounce acceleration - Change in spring coefficients 134

Figure 6 .6c Vehicle bounce acceleration - Change in damping coefficients 134

xiii

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Figure 6.7 Control input forces atfront/rear suspension 136

Figure 6 .8a Vehicle body bounce 137

Figure 6 .8b Vehicle body pitch 137

Figure 6.9 Frequency spectrum of body bounce/pitch accelerations 138

xiv

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 LIST OF APPENDIX

APPENDIX A DERIVATION OF VELOCITIES AND

ACCELERATIONS

xv

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 CHAPTER 1

INTRODUCTION

1.1 Vehicle Dynamics - A Historical Perspective.

The study of vehicle dynamics started probably in the early 1930s [1.1]. The tools used in

those days were mainly experimental observations. Ride comfort was first considered an

important issue of vehicle performance during this period. This period also saw

theoretical development that led to the practical design of suspension systems.

From the 1930s to 1950s, the importance of achieving a satisfactory compromise

between ride comfort and vehicle handling performance was recognized. The importance

o f the main force-generating element, the tire, had also been recognized by experimental

measurement of the force and moment properties. Accordingly, the design of the

suspension system was advanced. It is interesting to note that the development of

independent suspension was introduced during this period [ 1.1].

For the next three to four decades, the maturing theories and technologies expanded

significantly. More accurate rig results and mathematical models were developed for

which the tire and vehicle dynamics behaviors could be studied and verified

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 experimentally. Dynamic analysis was also broadened to consider studies of vehicle

stability, handling and vibration to predict ride comfort [1.1].

The last decade or two witnessed research in vehicle dynamics moving toward and

relying on more and more the development of computer modeling and simulation

methods [1.2]. Currently, these computer codes provide a range of ride and handling

models of varying degrees of complexity, which could hardly be prepared manually. All

of the commonly required calculations for vehicle dynamics studies have been embodied

in multi-body system (MBS) dynamics codes. These codes are expressed and solved

either numerically or analytically. The latter approach has the advantage of fast

simulation run time and ease of parameter change and control system implementation.

Depending on user preference, there are many computer packages commercially

available. Examples of stand-alone packages include ADAMS/CAR [1.3] and CarSim

[1.4]. There are also “add-ons” available that serve the same purpose such as the many

Matlab “tool boxes” contained in [1.5].

Although one could logically suggest the use o f these codes for a vehicle dynamic

study, the complexity of these codes makes it difficult to add user routines or add-ons, not

to mention that one may not be able to test the different underlying assumptions inherent

in the development of such codes. In addition, due to the complexity and the size o f the

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 output data that the codes tend to generate, it is difficult to introduce and investigate

various control schemes. Therefore, in this thesis the author will propose and develop a

vehicle model that is of medium complexity, but nevertheless one that is able to capture

the physical essence of ride comfort and vehicle handling performance.

1.2 Objectives.

The author will propose a vehicle dynamics model consisting of a rigid vehicle body and

rigid wheels, having independent suspensions connecting the vehicle body and wheels,

and taking into consideration tire-road interaction. The model will be able to address ride

and handling simulations, and ride comfort control applications. The scope of the model

presented will include derivation of the vehicle dynamics model, modification of the

Lund Grenoble (LuGre) dynamic friction model [1.6], numerical case studies for the

modeling verifications by using Matlab [1.7], and control application using the

non-switching sliding mode control technique for the improvement of ride comfort.

1.3 Organization of the Thesis.

The thesis contains seven chapters.

Chapter 1 deals with the introduction.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Chapter 2 presents an overview of some existing vehicle dynamics models and their

applications.

In Chapter 3, the development of a 14-degrees-of-freedom (DOF) full-car dynamic

model is described. The model considers the vehicle body, suspensions as well as the

wheel motions in a pre-selected coordinate system. Derivations of equations o f motion

are governed by the principles of Newtonian mechanics. The coupling of some equations

is addressed. Simplification of the proposed model and comparisons to existing models

are then presented.

Chapter 4 is concerned with a crucial component of the vehicle model, the dynamic

friction, or the road-tire contact force. It focuses on the LuGre model. Extension o f the

existing one-dimensional friction model to a two-dimensional model is made and

validated. The 2-D LuGre model is subsequently introduced into the full-car model o f

Chapter 3.

Chapter 5 is where the developments of Chapters 3 and 4 come together and are

applied. Three numerical case studies are included to demonstrate the applicability and

accuracy of the proposed model.

Chapter 6 introduces the non-switching sliding mode control technique to the

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 proposed vehicle model. A pitch-bounce car model is used to demonstrate the

effectiveness of the sliding mode control technique.

1.4 Relevant Terminologies.

A few terminologies are introduced here to facilitate the understanding of Chapter 2 in

particular. With reference to Figure 1.1, translation along the x-axis is the longitudinal or

forward motion; lateral or sideward motion is the translation along the y-axis and bounce

refers to the vertical (z-axis) translation of the main body of the vehicle. The rotations

about the x-, y- and z-axes are known as the roll, pitch, and yaw, respectively. A w heel’s

rotation about its spin (y) axis is its spin (co) and its vertical motion (z) is its bounce.

longitudinal
Z
bounce

Figure 1.1 Vehicle Motions

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 CHAPTER 2

OVERVIEW OF VEHICLE DYNAMICS MODELS

In studying vehicle dynamics, three essential components are considered for typical

ground vehicles, the main body of the vehicle (sprung mass), the suspension system, and

the wheels (unsprung masses). The vehicle’s kinematics and dynamics can be described

from the interactions among those rigid bodies. The incorporation of road-tire interaction

is also an essential part of vehicle dynamics. Therefore, “it is important to construct a

mathematical model that includes the nonlinear characteristics of the tires, the general

three dimensional motions of the sprung and unsprung masses, the required inertial

coupling between sprung and unsprung masses, and an accurate representation o f the

restraints and forces imposed by the suspension components” [2 .1].

Vehicle dynamics covers a wide range o f subject material because it is a study of

anything relating to vehicle systems. However, two major areas have been studied

extensively; ride comfort and vehicle handling performance [2.2], Even today many

researchers are seeking ways to further improve ride comfort and vehicle handling by

developing analytical tools, and by advancing control techniques to attain the desired

goal.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Simply to avoid the complexity of coupled vehicle modes, many researchers have

studied ride comfort and vehicle handling separately. Logically one could ask, “How can

this be justified when it is clear in the real world the vehicles are subjected to ride and

handling inputs simultaneously?” [2.2], The two aspects are inextricably linked to one

another. However, it may be difficult, if not impossible, to analyze or simulate all the

vehicle dynamics simultaneously.

2.1 Existing Vehicle Dynamics Models.

Many vehicle models have been developed. They were derived by considering energy

equilibrium or dynamic equilibrium. In general, these models can be classified into three

types: (i) the quarter-car model, (ii) the half-car model, and (iii) the full-car model.

2.1.1 Ouarter-Car Models

A quarter-car model consists of one wheel and associated suspension, and a body mass.

Due to its simplicity in modeling and the relative ease in obtaining analytical results, the

model is prim arily used to study vehicle ride comfort and to implement advanced control.

In this 2-DOF model, the vertical motions of body mass (sprung mass) and the associated

wheel and suspension masses (unsprung masses) are considered.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 m

a. One DOF b. Two DOF

Figure 2.1 Quarter-Car Models [2.3]

Hac and Fratini presented their work on vehicle ride comfort using the “skyhook”

damping control law [2.3], The mathematical vehicle models used were quarter-car

models with one DOF (Figure 2.1a) and two DOFs (Figure2.1b), respectively. It should

be mentioned that in this chapter, figures retain their original notations; that is, notations

used in the references. In Chapter 3 however, figures will be annotated by symbols

consistent with those used in this thesis. The model in Figure 2.1a used a single D O F to

describe the vertical response, Ziit), of the sprung mass after road signal 12 (f) is inputted.

The 2-DOF model shown in Figure 2.1b considered motions of both the sprung and

unsprung masses. In addition to the spring (ki), a linear passive damper (Cp) was also

introduced. It should be pointed out that u(t) in both models represents the “continuously

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 variable real time damping” forces that were governed by the “skyhook” damping control

law.

A similar model was used by Simon [2.4], The focus of his work was also on the

primary suspension systems. Various suspension systems, conventional and

non-conventional, were evaluated and compared for obtaining the optimal trade-off

relationships between ride and handling. A prototype of a continuously variable

semi-active system implementing the “skyhook” control algorithm was constructed, and

tested for its dynamic effect on vehicles. The work was also based on a quarter-car model

of 2 DOFs (Figure 2.2).

damper
spring

spring

Figure 2.2 A 2-DOF Quarter-Car Model [2.4]

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 2.1.2 Half-Car Models

A half-car model typically consists of either the left or right half of a vehicle, or the front

or rear half of the vehicle. The former is usually called the pitch-bounce model (Figure

2.3a), while the latter is the roll-bounce model (Figure 2.3b). There is also the so-called

roll-yaw model that includes, as its DOFs, the lateral, roll and yaw motions of the

vehicle.

. (B

Figure 2.3a Half-car model involving pitch motion

Figure 2.3b Half-car model involving roll motion

10

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Chen et al. [2.5] presented a pitch-bounce model along with their work on

constrained Hx control to active suspension systems on a vehicle. The half-car model had

4 DOFs, the bounce and pitch of the vehicle body and the bounces of the two wheels. The

suspension and tire were modeled by linear springs and viscous dampers. Gawade et al.

showed the in-plane 7-DOF mathematical model of a three-wheel vehicle to study the

influence of bump profiles on occupant injury [2.6]. The planar three-wheeled vehicle

model considered the motions of longitudinal, bounce and pitch of vehicle body, and the

motions of bounce and rotation of the wheels (Figure 2.4).

z f
Figure 2.4 Planar Model of a Three-Wheeled Vehicle [2.6]

11

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 2.1.3 Full-Car Models

Further to their in-plane model [2.6], Gawade et al. developed a three dimensional model

of the three-wheeled vehicle with suspensions and compliant tires [2.7]. It had 6 DOFs

for the vehicle body, and another 6 DOFs describing the vertical displacements and

rotation motions of the wheels (Figure 2.5).

Figure 2.5 Full-Car Model of A Three-Wheeled Vehicle [2.7]

For this seemingly simple full-car model, a good amount of detail had to be

incorporated. For example, transformation matrices were used to describe the relation

between the inertial coordinate frame and body centered coordinates. These matrices

were written in terms of three independent Euler angles (the yaw, roll, and pitch angles).

The equations of motion also took into account the steering effect. The so-called “M agic

12

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Formula” [2.8] was used to describe the tire’s behavior in the lateral direction for

different slip angles, and the Coulomb friction model was used to describe rolling

resistance, or the tire’s behavior in the longitudinal direction.

A simplified full-car model, together with the 2-D LuGre friction model, was

presented by Villella in [2.9] to study the handling responses with implementation o f the

input-output linearization control technique. The main focus of the w ork was on

incorporating the LuGre dynamic friction model into the full-car model and on simulating

the handling responses under a less aggressive lane change maneuver by the driver over a

smooth road.

Villella’s model was developed with reduced complexity in mind. The suspensions

were absent. They were replaced by four rigid joints, which resulted in a 7-DOF model.

The DOFs were, the longitudinal, lateral, and yaw motions of the vehicle body, and the

rotation motion of each wheel. As for the absence of suspensions, the equations of normal

reaction forces among the four wheels were determined analytically by using static force

balance in the z direction, static moment balances about the pitch and roll axes, and a

hypothetical suspension model with infinitely large values of the spring stiffness.

Accordingly, the normal reaction forces depend only on the geometric parameters o f the

vehicle, the steering angles, and the tire-road friction functions.

13

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 2.2 The Proposed Vehicle Dynamics Model.

This thesis proposes a 14-DOF full-car model. This model can be viewed as an improved

version on those presented in [2.7, 2.9]. Specifically, the vehicle body will have 6 DOFs

representing the six rigid body motions. Each wheel will have a bounce and a spin motion

associated with it. Suspensions will be present, so will road-tire interaction. The detailed

description and derivation of the model will be presented in the next two chapters, where

the reader may note that, for the kinematics, this thesis draws upon and expands the work

presented in [2.9]. For the dynamics, however, the free-body diagrams of Reference [2.6]

will be utilized where appropriate. This more general full-car model will then be

simplified to quarter- and half-car models, and to full-car models with fewer DOFs.

14

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 CHAPTER 3

THE 14-DOF FULL-CAR DYNAMIC MODEL

The first step in vehicle modeling is a thorough understanding of the physics - kinematics

and dynamics - of motion of the vehicle. This chapter is devoted to such a task, to

develop a 14-DOF full-car mathematical model. This model, like any other car models,

will provide a means to simulate vehicle dynamic behaviors which can further lead to the

proper design and control of vehicles.

The study of motion will inevitably involve the setup and use of frames of

references. Once the kinematics, the absolute accelerations in particular, is understood,

the Newton’s laws are applied in order to establish the required equations of motion.

These equations of motion are then simplified so as to compare with other car models.

This not only validates the proposed full car model, it also demonstrates the versatility of

the model.

3.1 Coordinate Systems.

Four frames of reference are used throughout the study. They are introduced below.

1) Global coordinates X-Y-Z

15

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 The global coordinates are used to measure the absolute position of the vehicle. It is an

earth-fixed frame of reference, and considered as an inertial frame [3.1]. Its origin is

arbitrary, but chosen as the position that the center of gravity of the vehicle body takes

when t = to- Note that the origin is fixed once chosen.

2) Vehicle body coordinate frame x-y-z

This coordinate frame is associated with

rs

unit vectorsi , j , k (Figure 3.1). It is a

JJ
x / frame that is simultaneously coincident

with the vehicle body’s center of gravity

Figure 3.1 Vehicle body frame of reference

CG, but it is not vehicle-fixed in that the

x- and y-axes do not rotate as the vehicle rolls and pitches; That is they don’t rotate about

the longitudinal and the lateral axes. Instead they are parallel to the ground all the time,

and are instantaneously aligned with the vehicle’s longitudinal and lateral axes,

respectively, while the z-axis is determined by the right hand rule. The advantages o f such

a setup include, ( 1) that only one rotation matrix is required for coordinate transformation;

and (2) that x-y-z can be regarded as the vehicle’s principal axes of mass moments of

inertia, under the assumption of small roll and small pitch. The rotation matrix required

for the transformation from x-y-z to X-Y-Z is,

16

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 'X cosy? —siny? 0
Y ■= siny> cos tp 0 (3.1)
Z 0 0 1

where <p is the yaw angle and is determined b y ^ = f

J t0
<pdt with being the yaw rate.

3) Wheel centered coordinate frame x CWin-ycw,n-Zcw,n

This coordinate frame has unit vectors Cw,»Jcw,n»4»,» attached to the wheel center CWn,

with the x CWfn-, y cw,n- and ztx ,r axes being parallel to the vehicle’s x-, y- and z-axes,

respectively (Figure 3.2). Hence this frame differs from the vehicle body frame only in

their coordinate origins. Unless stated otherwise, the subscript n denotes the wheel and

takes on values of 1 through 4.

4) Wheel base coordinate frame x bWin-ybWttl-zbw,n

The wheel base coordinate frame is attached to a w heel’s base, with the y bwM-axis being

parallel to the w heel’s rotation axis (the spin axis), the xbWifl-axis parallel to the ground,

and the z^,„-axis parallel to the z-axis of the vehicle body (Figure 3.2). The frame rotates

with the wheel as it steers, or it is wheel-fixed. That is, the wheel base coordinate frame

differs from the wheel center coordinate frame by the steering angle 6 n of the wheel. The

wheel center unit vectors tCWJl, ]cw.n, krw nand the wheel base unit vectors \„J%n, j bWi„, kbwn are

related by,

17

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 ^cw,n C O s6n %w,n ~ @njbw,n %w,n ~ ^ ^ ^ n ^ c w , n “I” STn.dn j cw rl

jcw ,n s ill On ibw,n ~b COS Qnjbtu.n Of jbw ,n — s i l l 0n tc w n -\- COS On jew ,n ( 3 .2 )

kcw,n kbw,n ^bw,n ^cw,n

cw.

bw,n

bw,n
bw.n

Figure 3.2 Wheel center and wheel base coordinate frames of reference

Figure 3.3 compares the four coordinate frames mentioned above. It illustrates how

they relate to each other. It is noted that the earth-fixed X-Y-Z, vehicle body x-y-z and

wheel base XbW,n-ybw,n-Zbw,n frames of reference follow the SAE recommended practice

[3.2]. As will be seen later, the equations that govern the general, nonlinear motion o f the

vehicle are written in terms of the vehicle body frame of reference x-y-z. This choice of

frame of reference greatly simplifies the derivation and calculation of the vehicle model.

It should finally be pointed out that wheel chamber (rotation of a wheel about the

xCWiH-axis) is not considered in this study.

18

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 3.2 The 14-DOF Full-Car

Model.

The full-car model has in total 14

DOFs. They are x, y, z, (p, 7 , ip,

z l , z 2 i z 3 i z 4> W 1 >w 2 I w 3 I w 4

(Figure 3.4, Z4 an d 07 , u 4 not

shown) where x, y, z, p, 7 , (p are the

longitudinal, lateral, bounce, roll,

pitch, and yaw motions of the

vehicle body; z 1,z 2 ,z 3 ,z 4 the

bounce motions of the left front,

Figure 3.3 The four frames of reference used

right front, left rear and right rear
yaw

wheels, and W] ,u >2 ,<^3 ,cu4 the

angular motion (spin) of the left

lateral
front, right front, left rear and right

rear wheels, respectively. Features

longitudinal ^f
Z
bounce of this 14-DOF full-car model are,

Figure 3.4 The full-car model

a) The vehicle body is treated as

19

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 a rigid body (the sprung mass as per terminology of [3.2]) with the six DOFs,

x,y,z,<j>, 7 ,<p;

b) Each wheel is modeled as a rigid body (the unsprung mass [3.2]) with bounce

motion z n and spin con;

c) The wheel bounce zn is considered independent of z, the bounce of the vehicle body.

Moreover, the bounces o f the wheels are considered independent o f each other, so are the

spins;

Such a setup obviously has independent suspensions in mind, but can easily

accommodate solid axle suspensions. In vehicle modeling, it is often the practice to have

2 DOFs for each axle [3.3], With independent suspensions, these two axle DOFs turn out

to be the bounce motions of the wheels at the ends of the axle; while for solid axle

suspensions, one may choose to use, as the axle DOFs, the bounce o f the axle’s roll

center and the axle rotation [3.4], with the latter being easily transformed to the former

(Figure 3.5).

d) It should then be noted that each wheel center is connected to the vehicle body

through a “spring-damper” combination, km and csn; In addition, the Coulomb friction can

be included (see [3.6]). However, such friction (stop) force is not considered in the

present study.

20

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 e) The tires are considered linear springs with spring constants ktn\ a 2-Step tire

stiffness model [3.6] may also be incorporated and considered in the future work; and

f) The steering of the wheels is not treated as a DOF. However, as can be seen in

Chapter 4, steering is incorporated into the model through tire-road interaction, hence

becoming “a state”.

A schematic of the full car model is shown in Figure 3.6 where zsn denotes road

profile, or terrain.

tra ck width w

axle roll center

axle rotation

* roll center bounce

Figure 3.5 Axle roll center bounce and axle rotation as DOFs

Zsl

Z*4

Figure 3.6 Schematic of the 14-DOF full-car dynamics model (spins not shown)

21

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Other 14-DOF full-car models available in the literature have same or similar setup

of DOFs. For example, the same DOFs were used in [3.3 - 3.5], It is interesting to note

that, while Reference [3.3] pointed out that the 14-DOF model “is quite suitable for

simulating vehicle response under significant (±10 degrees) roll motions”, Reference [3.5]

suggested that “the effect of the anti-roll bar is not negligible”, and included a simple way

to integrate the roll stiffness with the suspension model. On the other hand, Reference

[3.6] presented a 16-DOF full-car model. They were, six DOFs (longitudinal, lateral,

bounce, roll, pitch and yaw) for the vehicle body, three DOFs per axle (axle roll, bounce

and steer for solid axle suspensions, and left wheel bounce, right wheel bounce and axle

steer for independent suspensions), and one DOF (the spin) per wheel. It is noted that the

model in [3.6] included that of the steering system; as a result, steering angles were

governed by the equations of motion of the steering system. Since steering is not modeled

in the present study, steering angles are incorporated as inputs rather than as DOFs.

3.3 Kinematics of Vehicle Body and Wheels.

3.3.1 Velocity and Acceleration at Center of Gravity of the Vehicle Body, CG.

Vcg and A cg are conveniently expressed in terms o f unit vectors i , j , k of the vehicle body

frame x-y-z (Figure 3.1)

22

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Since the vehicle body frame rotates with angular velocity components of [0, 0,<^],

or the angular velocity vector of the x-y-z frame is f t = ipk , one can write

i — Cl x i = tpj, j = O x j = —ipi, k — SI x k — 0 (3.4)

so that A cg becomes

(3.5)

It is noted that, though Vcg and A cg are expressed in terms o f x, y, z a n d i , j , k , they are

absolute velocity and absolute acceleration measured with respect to the global frame

X-Y-Z which is inertial.

3.3.2 Velocity and Acceleration at a Wheel Center. CWn.

In determining the velocity and acceleration at a wheel center CWn, one starts with the

velocity and acceleration of the point that is at the top end of the suspension connecting

the wheel center and the vehicle body. Since such a point is a point in the vehicle body,

one has,

^ "vb,n ^ 'e g “h X (3.6)

where Vcg is the velocity of vehicle body’s CG written with respect to the x-y-z frame,

Q — (fik is the angular velocity vector o f the frame, and r vb,n is the relative position

23

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 vector of the top end of the suspension with respect to CG. It can be shown that (Figures

3.7a, b)

r vb,n = xni + V n j + (~Xn sin 7 + Vn sin (j))k

(3-7)
- xni + ynj + (-a ^ 7 + yn(t>)k

where x„ and yn are the x- and ^-coordinates of a wheel center n (see Figure 3.7c). They

can be determined by the wheelbase and track width of the vehicle. Note that the small

roll and small pitch assumption has been invoked in equation (3.7). Substitution o f Vcg, £2

and rvb,n yields,

V vb,n = ( i - <pyn)i + (y + <pxn) ] + zk (3.8)

and

A b ,n = [ x ~ a - 'ey - ^ 2xn ) i + ( y + Cpxn + (pi. - (p2y n ) j + zk (3.9)

V - _ ----------- - f T l r
---- ■•'6 J _________" ''~ 'J X
—■ 1 - xusinY
Z’

Figure 3.7a Side view of the vehicle body showing —xn sin 7

lr z

Figure 3.7b Front view of the vehicle body showing yn sin <fi

24

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Now, if it is assumed that the wheel centers will assume the x- and y-components of

Vvb,n and A vb,n, so that the relative motion between the vehicle body and the wheels will

only occur in the vertical direction (or along the ^-direction). The velocity and

acceleration of the wheel centers are then,

Vcw,n = { i - v y n ) i + ( y + <pxn ) 3 + Znk (3.10)

and

Aw,n = { x - (pyn - < p y - + {y + + y x - <f2yn ) j + znk (3.11)

Longitudinal Axis x
i
1

Lateral Axis
v

Figure 3.7c Locations of wheel centers in the vehicle body frame of reference

25

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 It should be mentioned that VCW:ll and A CWtn are absolute velocity and absolute

acceleration measured with respect to the X-Y-Z frame but expressed in terms of x, y, z

an d i , j , k . The relative displacement between a wheel center and the vehicle body

represents the compression experienced by the suspension located between the wheel and

the vehicle body. This relative displacement and its time-derivative are

A zn = z - xn s in 7 + yn sm<f) - zn = 2 - x ny + yn<f>- zn
(3.12)
A zn - z - x„7 co s7 + yn(j)cos<j>- zn = z - x ny + yn(f>- zn

It should be pointed out that the present treatment of the kinematics of the wheel

centers is different from that of [3.6] in which the wheel centers were regarded points in

the vehicle body. Such a treatment, however, leads to the same expression for A CWin.

Details of such an approach are presented in Appendix A.

3.4 Dynamics of Vehicle Body and Wheels.

Once the absolute accelerations at CG of the vehicle body and at CWn of the wheel

centers are determined, free-body diagrams (FBDs) should be sketched and N ew ton’s

second law of motion applied. In the derivations that follow, certain FBDs of [3.7] are

utilized where appropriate. However, all equations of motion are re-developed. It is noted

that all equations of motion are written with respect to the x-y-z coordinates.

26

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 3.4.1 Linear Motions of the Vehicle Body.

The reaction force exerted by a wheel center onto the lower end o f the “spring-damper”

combination is (see Figure 3.8, showing forces from wheels 2 and 3 only),

Fnltj = Px cw ,n * + pvy c w . n J1 ~ Pzz cw ,n k (3 13)

\ J , i
J /

Therefore, the equations of motion are, by virtue of N ewton’s second law of motion,

( 1) for the linear motion in the longitudinal direction

M bx = M bw + Y J PXcv (3.14)

(2) for the linear motion in the lateral direction

M by = + J 2 py,:. , (3.15)

and (3) for the linear motion in the vertical direction (or the bounce motion)

M bz = M bg - ^ P Zmn (3.16)

Note that in above equations, the summation is with respect to n, that is, This
n —1

notation is used throughout the remainder o f the thesis unless stated otherwise.

27

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Ka

>c vV,:

Figure 3.8 FBD of vehicle body

(showing forces and torques from wheels 2 and 3 only)

3.4.2 Angular Motions of the Vehicle Body.

To arrive at the equations of motion for the angular motions of the vehicle body, one

recognizes that the x-y-z frame is a rotating frame of reference. Since the frame rotates

about the z-axis only (that is, = 0,0 ^ = O,0 2 = p ), while the vehicle rotates relative

to the x-y-z axes with the angular velocity components of [ <j>,7 , 0 ], the angular velocity

components of the vehicle body are then u x = Qx + <fr — <j>, u)y = fly + 7 = 7 and

ujz = f l, + 0 = p. The equations of angular motion of the vehicle body are therefore

[3.8]

= ai7 )

w here the angular m om entum about the CG is [3.7]

28

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 where the angular momentum about the CG is [3.7]

(3.18)

In equation (3.18), Ixxg, Iyxg and Izxg are the mass moments of inertia of the vehicle body

about the x-, y-, and z-axes, respectively. Terms such as Istxg, for example, denote the

mass product of inertia of the vehicle body about the s- and f-axes, and = I yxxg,

lyz.cg = Izy.cg and I Izcg = I zxcg. Since symmetry in the x-y and y-z planes can be

reasonably assumed for most vehicles, this leads to I Tgcg = Jyxcg = I yz cg = I zycg = 0.

Equation (3.18) then simplifies to

(3.19)

It is then recognized that Ixzxg, or I7XCn, is either not available for most vehicles, or

when available, is many orders of magnitude smaller than Ixxg, I rxg and Izxg and may be

neglected. This leads to a further simplification of equation (3.19) that gives rise to

h*cg CXX'QjlJXi -f- Iy cgUJyj I ZCg'UJ~k (3.20)

Substituting equation (3.20) into (3.17) yields,

H ^ x ,c g — I x j C g ^ x ~ Iy ,c g U y Q z — 7x,cg4> ~ ^y,cgi'P

= l y ^ c g ^ y + kXiCgu>x Clz = I y , c g l + 7x ,c g <

t>'f (3.21)
^ l ^ Z,cg — Iz,cgUz—^z,cg'f

29

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 where the left hand sides (LHS) represent the sums of moments due to forces about the x-,

y-, and z-axes, respectively. The angular motion about the longitudinal x-axis (or the roll

motion) is then governed by

^■x,cg4> Iy ,c g 'Y V E [ P z (Wn Un ] COS (!) E [ Py,.w:l z v jji ]

(3.22)
= I y ,c g 'i(P ~ Y l \ P z m n lln ] — Y ^ [ P y Clwnz w,n ]

the angular motion about the lateral y-axis (or the pitch motion) by

^ .< * 7 = -Iw ftP + E E + I n ] COS 7 + E [ E „ , „ E ™ ]

(3.23)
= + E E + E [ e „ , „ Z n ] + E [ e „ . „ Z w ,n ]

and the angular motion about the vertical z-axis (or the yaw motion) by

P=
I,i,{ + E f r . . . * " ] 0 -2 4 )

The FBD of the vehicle body as viewed from the x-axis, y-axis and the top is given

in Figure 3.9. In equations (3.22) and (3.23), zw,n = zn + H - z is the z-coordinate

difference between the CG and a wheel center at time t. H is the initial z-coordinate

difference, see Figure 3.10. Obviously, z(0) = z«(0) = 0.

30

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Figure 3.9a FBD of the vehicle body as viewed along the jc-axis
(wheels included for reference only)

. 0

cw.4

P;z cw,4

Figure 3.9b FBD of the vehicle body as viewed along the y-axis
(wheels included for reference only)

31

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 -vc\v,2

cw,3

Figure 3.9c FBD of the vehicle body as viewed along the top
(wheels included for reference only)

.--------

Figure 3.10 Vehicle body as viewed along the y-axis showing H , z and z n

32

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 3.4.3 Forces on the Wheel Base and Equations of Motion of the Wheels.

•cw,u
cw„
CW.
CW.ll

'CW,11

fy,n

N'n
,

Figure 3.11 Forces exerted by road surface onto the wheel bases

Reaction force exerted by the road surface onto a wheel base is (Figure 3.11)

■ffcw.n F 'fx,n * bw,n ^'fy,n3bw,n ^ r j^ b w ,n

( ^ fa ,n c®® @n 3“ F fy n sin. On'jicw^n (3 .2 5 )

4" ( F fx ,n @n F fy tn COS 0n j j Cw,n ^ n ^ c w ,n

where equation (3.2) has been applied. With the FBD of a wheel (Figure 3.12), applying

Newton’s second law of motion, the equations of motion of the wheel center are,

( 1) for the linear motion in the longitudinal direction

[:x - (pyn }m n= [ipy + (p2x n }mn - Px ^ n - Ffx^ncos0„ + Ffy>n s in 6 n (3.26)

(2) for the linear motion in the lateral direction

[y + ipxn } m n= - [ < p x - (p2 yn } m n - PVcwn - F ^ nsin#n - F ^ nc o s 6n (3.27)

33

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 O W i,

CWJi ' C W .ll

k C W .ll

Figure 3.12 FBD of a wheel

(3) for the linear motion in the vertical motion

m nz n = PZcwn + m ng - N n (3.28)

and (4) for the spin motion of the wheel, by considering the spin axis of a wheel (spin

axis passes through the wheel center and is parallel to ytw.n-)

Ispin^n tF 1f xtn (3.29)

It is observed that the LHS of equations (3.26) and (3.27) each involves two

acceleration terms. On the other hand, equations (3.28) and (3.29) do not exhibit the

coupling between accelerations. Obviously, existence o f such coupling complicates

equation solving.

34

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 3.4.4 Forces Developed in the Spring-Damper Combinations.

All “spring-damper” combinations considered are assumed massless. Equilibrium o f each

in the vertical direction yields

PzCWin = Pzcw,n + ks n ( z - X nl + yn<t>- Zn ) + csn (z - Xny + ynj> - zn ) (3 .30)

where ksn, csn are the spring constant and damping coefficient of the spring-damper

combination, P® is the initial spring force required to support the vehicle body weight

Mbg. It is found that

_ n , _ bd _ n b c -.a _ cid -.ft ., cl c

Pr . = M u g -,— , P* 0 — M u g -,— * = M bg - — ,P 2U . = M bg ,
z r.w, 1 I W c w .2 I W c w ,3 I W c w <& I yj

4 (3.31)
E
71 = 1
p L ,n = M bg.

with 1 = a + b, and w = c + d being the wheelbase and track width, respectively. The x-

and y- components of the forces are determined as follows, from equations (3.26) and

(3.27)

= M + <pyn + <py + kp2x n ] m n - FfXiU cos 9n + Ffyjn sin 9n (3.32)

Py^n = [~y - <pxn ~ ¥>* + F 2Vn ] m n - Ffx>n sin 8 n - Ffyin cos 8 n (3.33)

3.4.5 Forces Developed in the Contact Points Between a Wheel and Road Surface.

If A^is the tire’s stiffness and zm the road profile, respectively, equilibrium in the vertical

z-direction yields (Figure 3.13)

35

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 N n = PL .n + m n9 + h n {zn - Zsn ) (3.34)

P7^ c w ji

Figure 3.13 Free body diagram of a wheel

(showing only forces in the z-direction)

3.4.6 Bounce and Spin M otions of the Wheels.

Substituting equations (3.30) and (3.34) into (3.28) yields,

mn^n = kgnZ Q^sn 3“ k^n)zn ksn(xn'y

(3.35)
~k^snz CsnZn Csnip^nif Un'P') 3“ kinZm

and equation (3.29) is repeated here for easy reference.

(3.36)

3.4.7 Bounce Motion o f the Vehicle Body.

Substitution of equation (3.30) into (3.16) gives,

M bz = - J 2 l k™ ( z + Vn(t> - s n7 - zn ) + csn(z + yn4> - x nj - zn )\ (3.37)

36

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 It is noted that the four equations g o v e rn in g ^ (equation 3.35) andcu7t (equation 3.36), and

the equation governingz (equation 3.37) are not coupled to one another. This, however, is

not the case with the longitudinal, lateral, roll, pitch and yaw motions of the vehicle as

demonstrated by equations (3.38)-(3.42) given below.

3.4.8 Longitudinal, Lateral, Roll. Pitch and Yaw Motions of the Vehicle Body.

Defining M tot — M b + Y m n which is the total mass including that o f the vehicle and

those of the wheels. Substituting equation (3.32) into (3.14) yields

MfotX ~ Y ^ { y n m n \V
(3.38)
M-by^p 'y ] ( Ffx n cos 6 n Ffy>n sin 6 n ) -T y ' [ipy -f- <px n j Tnn

Substituting equation (3.33) into (3.15) results in,

MtotV + Y j ^Xnm n W
(3.39)
Mfoiip y ^ ( Ffx n sindn + Fjy^n cos 9n ) y ^[ipx <p> yn j Tnn

Substituting equations (3.30) and (3.33) into (3.22) gives,

y ]y y 1 ^w.n^n \ <
p d~ Ix,cg&

lyfigi^P d- y ' [ ( ^fx,n d- COS 9n ) Zw^n ]

(3.40)
~ Xn l + Vn<l> ~ Z n ) ] V n } ~ 5 Z { [ C« * ~ X ^ ~ Zn)]Vn }

+ Y { [ v i ~ V 2 yn]zw,nrnn }

37

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Substituting equations (3.30) and (3.32) into (3.23) results in,

^ 1f \x y ' [ Vn^w,n^n ] Iy,cg'~f

IxjCg&^P ""I- y / -^rt y y[ ( F'fx.n C*-® ^fy.n Sin @n ) %w,n ]

+ I ] { [ M 2 - Xn l + Vn4> - A O K } + 2 { [ c m ( i - xn l + Vn<P ~ Zn) \ x n } (3 '41)

+ S {[ W + ^ Xn ] Z' » , A }

And substituting equations (3.32) and (3.33) into (3.24) yields,

- J 2 ^ n m n ]x + Y j { Xnm n \ y + {hfig + + y n )™ n \}v

y 1[ ( ^fx,n sin Ffytn cos 9n 'j x n ] “H y ^ [( Ffx.n co s9n ^fy,n sin $n ) yn j (3.42)

- ^ { [ ^ - ^ 2l/n](®nWn )} - ^ 2 {[vni + V2^ ](l/» " * n )}

Equations (3.38) through (3.42) form a set of five simultaneous equations that can

be recast into the following matrix form,

025 y ^2
a 32 a 33 a 35 ■<j) ■ = 63 (3.43)
a 41 a 44 a 45 •j 64
a 51 a 52

where the nonzero elements are

(3.44a)

38

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 and

h - M byip + ^ 2 ['fV + ^ x n }m n - ( Ff*,n COS dn ~ Ffy,n sin ) (3 .44b)

h = ~ M bx<p - [fix ~ V 2 yn ]m n - ( Ffx,n sin 9n + F ^ n cos 9n ) (3.44c)

h = h f i g W + Z ^ { [ ^ _ ^ V n \ z w,nm n }
T ^ ' [ ( Ffx,n sin 9n T Fjy^n cos 9n ) zw>n ]
(3.44d)
- ^ { [ ksn{z ~ X nl + Vnfi “ Zn)]yn )

'y y{[ <-'sn(-^ x n i + y n 4> zn)]yn |

^4 d~ 'y 1{ f^PV 4“ Xn ^ZwnVfln |

+ 5 Z Tn ~ J 2 K Ff C 0 S 9n ~ F fy,n s i n 9 n, J z w,n
(3.44e)
+ J 2 t t k ™ (Z - Xn l + V v ti ~ z n ) } X n }

+ S { [ Cs n ( ^ - Xn i + Vn4> ~ z n ) } x n }

^5 y / [ ( Ffx,n 9n Ffy,n sin 9n ) y n ]

y 1[ ( F fx,n sin 9n + Fjy^n cos 9n ) x n ]

(3.44f)
- J 2 { l v x - V 2y n ] ( x n m n ) }

~ Y l { [ F y + F 2X n } ( y nm n ) }

The 5x5 coefficient matrix [A] of equation (3.43) is rank-sufficient. In fact, by

inputting the matrix into the symbolic computation software M APLE [3.9] (see Figure

3.14a), the matrix can be inverted analytically (see Figure 3.14b, where the five separate

rows vectors represent each row elements of the [ A f ’matrix). Or it can be inverted

numerically depending on computational effectiveness. Note that the [A] matrix is

time-varying (see, for example, elements a jj and a .4 5 ) which requires inversion at every

39

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 time step o f integration. Following inversion, the five equations become decoupled. The

linear and angular accelerations can then be determined. That is

]T
j 4> 7 v>\ = [ 4 f 1 b\ fh h h f (3.45)

’ a ll 0 0 0 al5~
0 a ll 0 0 a.25
0 a32 a33 0 a35
-a32 0 0 a44 a45
a l5 a25 0 0 a55

Figure 3.14a [A] Matrix as inputted into MAPLE

a l I a55 - a25 ‘ a 15 a25
, 0,0,
a l l ( a l l a55 - a l 5 2 - a252) ’ a l l ( a l l a55 - a l 5 2 - a252)
a l5
a l l a55 —a l 5 2 - a 252
a !5 a25 a l l a55 - a ! 5 ‘
, 0,0 ,
a l l ( a l l a55 - al5~ - a25~) a l l ( a l l a55 - a l 5 2 - a252)
a25
a l l a55 - a 15' —a25"
a !5 ( - a l l a35 + a32 a 2 5 ) a l l a32 a55 - a l l a25 a35 - a32 a!5~
a33 a l l ( a 11 a55 - a l 5 - a25 ) a33 a l l ( a l l a55 - a l 5 2 - a252)
1 - « / / a 3 5 + a 3 2 a25
, 0 ,-
«33 ’ ’ a33 ( a l l a55 - a ! 5 2 - a 252)
a l 1 a32 a55 + a l 1 a !5 a45 - a32 a 2 52 a25 ( a l l a45 + a32 a 15)
T7>0
a44 a l l ( a l l a55 - a l 5 l - a25~) a44 a l l ( a l l a55 - a l 5 z - a25L)
1 a l l a45 + a32 a l5
a44 ’ a44 ( a l l a55 - a l 5 2 - a252)
a l5 a.25
7.0,0,
a l 1 a55 - a l 5 " - a25~ a l l a55 - aJ5~ - a25
a ll
a l l a55 —al5~ - a 2 5l
Figure 3.14b [A]'1 Matrix as determined by MAPLE

40

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Now that the entire set of 14 equations of motion, equations (3.35), (3.36), (3.37)

and (3.45) with n = are defined, they will be rewritten in the state variable format,

with the 28 state variables b e in g s , z2, z3, z 4, u u uj2, w3, cu4, z, x, y, </>, 7 , </?, zu z 2, z 3,

i 4,cj1,cu2,cj3,cj4,i,i;,y,(/>,7 , 0 . O f these state variables, the vehicle body velocities x, y are

velocities written in terms of the local (vehicle body) coordinates. They should be

transformed to the global (earth-fixed) coordinates as follows

X = x cos ip — y sincp
(3.46)
Y = isiny? + ycosp

such that the 28 state variables are now z 1,z 2, z 3 ,z 4 ,uj1,Lo2 , u 3 ,LU4 ,z,X,Y,(j),j,p>, z 4,z 2 ,z 3,

z 4 , u u l j 2 ,u j3 , l ) 4 , z, X , Y,<j>, 7 , ( p . They can then be solved for by numerically integrating

the 28 first-order ordinary differential equations (ODEs) by a ODE solver such as the 4-th

order Runge-Kutta method. It should be noted that in equation (3.46), (p is the yaw angle

which is a state variable and is the integral of yaw rate over the time interval to to t,

t
pdt .
(
I ,0

3.5 Simplification of the 14-DOF Full-Car Model.

In this section, the versatility of the previously presented full-car model will be

demonstrated. The full-car model will be simplified to various car models available in the

literature, ranging from quarter-car models and half-car models to other full-car models.

41

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Most of the models have been surveyed or reviewed in Chapter 2. For easy reference,

they will be recited and renumbered. It should be noted that, in Figure 3.15 through 3.18,

the symbols have been changed, from what were used in the respective references, to

those used in the thesis. This is done with consistency in mind.

3.5.1 Simplification to Ouarter-Car Models

Quarter-car models, which include just one wheel and the associated suspension and the

vehicle body mass, are widely used for suspension analysis. A quarter-car model can be

obtained by neglecting friction and by setting x = y = cf) = 'j — ip= x = y = <p —

7 = <^ = ii = j/ = ^) = 7 = (/3 = 0, 7 = <p = a; = y = (/) = 7 = (/3 = 0,

M w, ksn ks , csn cs, kf-rj kf-, zsn zs, zn z w, zn zw, Tn 0 . Equations

(3.35) through (3.42) are then reduced to

Myj'zw ks (z zw) + cs (z zw ) -f- kf (zs zw) (3.47a)

M hz = - 4 [ks ( z - zw ) + cs (z - zw)} (3 .47b)

It is further found that equations (3.36) and (3.38) through (3.42) are identically equal to

zero. Since the sprung mass is M s = Mb / 4, un-sprung mass is the wheel mass M w, one

then has

42

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 M wZw (z zw) 4- cs ( z zw)
(3.48)
~\~kt (zs zw)

t M sz = ~ [ k s (z - zw ) + cs(z - zw)] (3.49)

Equations (3.48) and (3.49) are identical to

Nw
equation (2.1) of [3.10] and equation (8) of [3.11]

respectively, and equations (2 . 1) and (2.2) of

i— T M m m m m m m m m m
[3.12]. A typical quarter-car model [3.13] is

Figure 3.15 A quarter-car

model [3.13] illustrated in Figure 3.15.

3.5.2 Simplification to Pitch-Bounce Half-Car Models

Half-car models typically include the so-called pitch-bounce models, the roll-bounce

models and the roll-yaw models [3.10]. The pitch-bounce models represent the left- or

right-half of the vehicle, or two axles of the vehicle. The roll-bounce models consist of

the front- or rear-half of the vehicle, or a single axle. Both the pitch-bounce and the

roll-bounce models include the bounce motions of the vehicle and wheels. To the contrast,

the roll-yaw models have no bounce motions [3.10], including instead only the lateral

motions as the DOFs of the models. The more general roll-yaw models are not discussed

here because torsional deflections in the suspensions have also not been considered; and

43

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 steering angles have also not been included as DOFs of the system. A typical pitch-

bounce model is shown in Figure 3.16.

zA
........-....... ..X
Mb. ly.o? ! (-Q <x

r m1
'"'bid Qk up k,:\
up
Zi
r L

X
. ku ^ ki.

L t. L
Figure 3.16 A half-car model showing pitch and bounce motions [3.14]

Setting x ^ y = cf) = ifi = x = y = ^ = ip = x ^ y = 4> = ip = 0 ,T n = uin = u n

F fx,n = Ffy n — 0, and considering only the vertical force components from the

suspensions, equations of motion (3.35) through (3.42) are reduced to,

Tn.n Zn k sn ( Z Zn %n~i) "b dsn ( ^ X, A &tn ( z sn ) (3.50)

XlfjZ } ' [ ksn ( Z Zn X n ^f ) Csn{z Zn Xn y ) j (3.51)

ly .c g 'J 'y 1 { [ ^sn •A iT ) ~b ^ s n ( % %n •^n'T) ] } (3.52)

Equations (3.36), (3.38), (3.39), (3.40) and (3.42) are identically equal to zero. Note

that Equations (3.50), (3.51) and (3.52) are identical to equations (la), (lb), (lc) and (Id)

of [3.14],

44

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 3.5.3 Simplification to Roll-Bounce Half-Car Models

Mathematically, the roll-bounce models are not very different from the pitch-bounce

models. The assumptions axex = y = y = if = x = y — 'j = ip = x — y — /

<p = 0 , Ffx>n = Ffy .n = Tn = u n — u n = 0. The equations of motion for this case

reduce to

ksn ( z Zn yn(f>) + Csn ^Z Zn + 2/ri*^) T ( z sn zn ) (3.53)

M bZ = ~ J 2 [ k sn ( z - z n + Vn<P) + Cm(z ~ Zn + yn<j>)] (3.54)

^ x tcg& ^ ] i [ k sn (z z n "T V nft) T ^sn{z zn T J/n0)]yn} (3.55)

_ Mb, Xx \ CG:

V.

H
k , CD Kt:
cts CD

Zi
A
• Zi

K s. H I Csi K s2 j”] - I Cs2

Zsi ^ I A ZS2

/ / / / 7/ / / / / / /

Figure 3.17 A half-car model showing roll and bounce motions [3.15]

A roll-bounce half-car m odel is show n in Figure 3.17. Physically pitch-bounce and

roll-bounce models are however different and serve different purposes. Though both can

be used to analyze suspensions, the roll-bounce models enable the investigation of

45

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 rollover of vehicles; while the pitch-bounce models are appropriate for ride comfort

analysis.

3.5.4 Simplification to a 7-DOF Full-Car Models.

The 7-DOF full-car model is often studied ([3.12, 3.16, 3.18], and with the seat bounce

excluded in [3.17, 3.19]). These DOFs are shown in Figure 3.18, including the bounces of

the vehicle body and the wheels, respectively, and roll and pitch. To simplify the 14-DOF

model to the 7-DOF one, one needs to setx = y = ip = x — y — <p = x = y = Cp = 0,

Tn — con = u>n = FfXtn — Ffy<n = 0 . As a result, the equations of motion become,

mnzn = ksn(z - z n - x„7 + yn4>) + csn[z - zn - xnj + yn<j>) + ktn (zsn - zn ) (3.56)

M b'z = - J 2 [ k™ O - z n + Vn<t> - X n l ) + Csn(z - Zn + - Xnj) ] ( 3 .5 7 )

'y / 1 [ ksn {z zn %n'y + 2/n*^) "F can( i Zn ~ Xn'j -+- 2/n0)j?/n | (3.58)

Iy,cg^f y y{[ ^STi {z z n ~~ "F yn4>) "F Csn (z Zn Xn7 + yn<f>)^Xn J- (3.59)

Equation (3.36), (3.38), (3.39) and (3.42) are identically equal to zero. Equation

(3.56), (3.57), (3.58) and (3.59) hence form the equations of motion of the model.

46

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Figure 3.18 A 7-DOF full-car Model [3.12]

3.6 Concluding Remarks.

In this chapter, a 14-DOF full-car dynamic model has been developed. The resulting

equations of motion are equations (3.35) through (3.42). The last five equations, (3.38)

through (3.42), are coupled and have to be solved simultaneously.

The full car model can be simplified to (1) a quarter-car model defined by equations

(3.48) and (3.49); (2) a pitch-bounce half-car model defined by equations (3.50) - (3.52);

(3) a roll-bounce half-car model defined by equations (3.53) - (3.55); and (4) a 7-DOF

full-car model defined by equations (3.56) - (3.59). These simplified models will be

47

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 tested with results presented in Chapters 5 and 6 . Before proceeding, however, the

modeling of friction forces has to be dealt with.

48

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 CHAPTER 4

ROAD-TIRE FRICTION

So far all necessary components of the dynamic model of a vehicle have been considered

with the exception of the friction force. As was concluded in Section 3.5, vehicle motion

is primarily determined by the interaction forces between the tires and the road, or the

road contact forces. Therefore, one of the crucial elements of vehicle modeling is to

properly model the road-tire friction force.

This chapter deals with this crucial element of road-tire friction. A brief literature

review will first be given, followed by the details of the 1-D and 2-D LuGre dynamic

friction models. For purpose of comparison other friction models such as the “M agic

Formula” and the 3-D brush model are also looked at. The chapter concludes with

formulation needed to incorporate the 2-D LuGre friction model into the dynamic full-car

model that was presented in Chapter 3.

4.1 Literature Review.

The behavior o f road-tire friction is well known to be highly nonlinear. Many friction

models have been proposed attempting to capture the essence of the complicated friction

phenomena with reasonable complexity [4.1]. Since there is a wide range o f physical

49

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 phenomena that cause friction, from elastic and plastic deformations, to fluid mechanics,

etc., the nature of the models is quite different. They can be static or dynamic. They can

be described by differential equations, differential-algebraic equations, and so on [4.1].

Traditionally road-tire friction was modeled by a static (or steady-state) model.

However, this steady-state point of view was rarely valid since in reality the tires can

experience continuous phase change between vehicle’s acceleration and braking. This

called for the need in developing friction models that would capture the transient

behaviour of the road-tire contact forces under time-varying velocity. These dynamic

friction models are usually described by ordinary differential equations.

A friction model discussed extensively in the recent literature [4.2-4.7] came to the

author’s attention. This dynamic friction model, known as the LuGre model, was

introduced [4.8] as the result of collaboration between Lund Institute of Technology of

Sweden and Laboratoire d ’Automatique de Grenoble of France (LuGre). The LuGre

friction model has been demonstrated to be an accurate model for capturing most o f the

steady-state and transient friction behaviours that have been observed experimentally, and

to be suitable for the type of in-depth exploration of wheel torque capability [4.2 - 4.8,

4.10],

50

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 The LuGre model can be used when considering the cases of either rigid or

non-rigid road-tire contact. The rigid contact case treats the wheel as rigid; as a result, the

single point contact lump model is formulated [4.4, 4.5, 4.7]. The non-rigid contact

renders a distribution model to describe the interaction on the contact patch [4.2 - 4.4,

4.6]. The distributed model would no doubt yield friction behaviors that are closer to

reality. However, in keeping with the rigid wheel assumption adopted in Chapter 3, and in

aim of simplicity o f modeling and numerical simulation, the focus of the present study

will be on the point contact LuGre model. The distributed formulation will be

recommended for future work.

The advantage of the point contact LuGre model is that the physical parameters

entering the model can be selected by a vehicle designer to match the experimental data

and be used to describe the condition of road surface. Most importantly, the model is

appropriate for normal vehicle motion situation, such as steady-state or transient phases

between braking and acceleration; not to mention that the LuGre model has been

extensively discussed and applied due to its simplicity in model derivation, ease of model

parameter identification, and high accuracy in predicting the frictional behaviors.

51

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 4.2 The LuGre Dynamic Friction Model.

In Chapter 3 it has been shown that two components of friction force, Ffi and Fjy (see, for

example, Figures 3.11 and 3.12, and equations (3.25) and so on), are needed to determine

the motions of the vehicle. Thus, a two-dimensional friction model is required for

describing the longitudinal and lateral frictions. Such forces are needed in various vehicle

dynamics simulation studies, especially vehicle handling studies. The longitudinal and

lateral frictions are related to N, the normal force developed at the contact point between

the tire and the road surface:

FfZ = HxN , Fjy — flyN (4.l)

where /ix and /uy are the coefficients of friction in the longitudinal and lateral directions,

respectively. They are also known as the normalized tire friction. Note that the subscripts

x and y denote the x;w- and yt,w-axes (of the wheel base coordinate frame) defined in

Section 3.1, see Figure 3.2 in particular. In this chapter the subscript n ( n = 1, ..., 4) has

been dropped for simplicity. It is understood that the equations and discussions presented

in this chapter are applicable to all four tire-road contact points.

4.2.1 The One-Dimensional LuGre Dynamic Friction Model.

The LuGre Model interprets friction as the interaction of microscopic surface asperities

52

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 which act as the bristles (rubber elements of the tire) and deflect as the surfaces move

with respect to one another (Figure 4.1). The deformation of the bristles gives rise to the

friction. The frictional force is given as a function of internal deflected state rj, (see [4.2]

for example), with rj(t) satisfying the differential equation

Direction of motion

Figure 4,1 The average internal deflected state tj in the direction of motion

* 1( 1) _ K (4.2)
dt g( v r )

and the friction is related through

, ^(t) , TT N
0 0 7 7 ( f) + o'! — 77
dt
h 02Vr (4.3)

and the coefficient of friction, pi, is

, dr){t) , T[
M= cr077(t) + h tr2Vr (4.4)

In equations (4.2) through (4.4), rj(t) is the internal state that describes the deflection of

an elementary rubber element, Vr is the relative velocity between the contact surfaces, N

is the contact normal force exerted on the wheel by the ground, oo is the rubber lumped

53

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 stiffness, cry is the rubber lumped damping, 02 is the viscous relative damping, a n d g ( V r )

is a speed dependent sliding function which represents the transitions between the static

and kinetic friction coefficients as a function of relative velocity between the contact

surface.

( 1^1f
g( v r ) = gk + (gs -

where pik is a parameter representing kinetic friction, g s a parameter representing static

friction, and vs the Stribeck velocity [4.9], which refers to the low slip or low relative

velocity region, where a decrease in friction force is seen. The constant parameter <5 in

equation (4.5) is known as the Stribeck exponent [4.6]. It is a shape parameter used to

capture the steady-state friction or slip characteristic, and typically has values in the range

of 0.5 to 2.0 [4.6]. The Stribeck velocity and shape parameter affect the rate of transition

between g s and /uk. Figure 4.2 shows how viVand 8 affect g ( Vr ). Typically, a smaller v5or

a larger <5 suggests a quicker transition from fis to ftk-

US
0.2 0.2

0 18 0.1 8

0 16 0 .1 6

>
01 01
0 .1 4 0 .1 4

v s In c re a se s
012

-100 -5 0 0 50 100 -100 -50 0 50 100

Vr (m/s)

Figure 4.2 The effect on g ( Vr ) due to changes in friction parameters

54

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 4.2.2 The Two-Dimensional LuGre Dynamic Friction Model.

The extension of the LuGre friction from the longitudinal one-dimensional (1-D) model

to the longitudinal and lateral two-dimensional (2-D) model has been proposed in [4.3,

4.5 - 4.7]. A simple extension would be to assume that the bristle deflections are

directionally independent, so as to apply equation (4.2) along the longitudinal (x) and

lateral (y) directions separately. This would result in,

dr)X)y (.t) _ y _ IVrx,y \

P Q s .j/ ..

dt gx,y {VrXiy) ,y

Equation (4.6) should be looked at as two independent sub-models where gx<y (Vrx^y )

were two independent friction functions with two different sets of parameters; that is,

— f (-T*f
9x ( Vr x ) fJ^kx T (/i sx S1 and gy (Vry) fi^y -b (/i sy /i^ )e v

Physically this would mean that two bristles would deflect independently in two

directions. In reality there is only a single bristle at the contact point; therefore there

exists a single, unique, friction [4.6]. A different expression for gx y ( Vrx^y ) was given in

[4.6],

V
y rx,y
9x,y ( Vrx,y ) — 9(Vr) (4.7)
K

Equation (4.7) yields positive longitudinal and lateral components of the road-tire

sliding friction force g ( V r ) (see Figure 4.2). Substituting equation (4.7) into equation (4.6)

yields the deflection equations for the combined longitudinal and lateral motion. Unlike

55

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 in equation (4.6), the two sub-models are now coupled with the sliding friction function

and the relative speed. That is,

d^ {t) = V _ ^,y\Vr\
dt ra-» g (V r ) Vx's

where the longitudinal and lateral frictional coefficients are, from equation (4.3),

d V x ty ( t )
ftx.y ^Qx,yVx,y ( O "f“ &lx.y ^ 1“ &2x,y*rz,y

v <r0x,v \VT\ \ (4 '9)

^ 0 x }yVx,y ( ^ x,y rayt/ „ { \r \ 'tfz-.V ' &2x,y*rx,y
g{vr)

where a 0x, a 0y are the bristle stiffness constants; a lx, crly are the bristle damping coefficient

constants; a n d a r e the bristle viscous damping coefficient associated with the x

and y directions, respectively. These parameters can be identified from experimental data

if possible, or pre-assigned as constants for simulation purposes.

4.3 Steady-State Characteristics.

In most literature it is common that the steady-state friction characteristics are expressed

as a function of slip coefficients. In 2-D steady-state friction model, the lateral friction is

expressed as a function of slip angles, and the longitudinal friction as a function of slip

ratio. The steady-state analysis of friction models is widely employed in the study o f

road-tire interaction.

56

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 The steady-state characteristics of the deflected rubber element are obtained by

d'H' (£)
setting — - j ----- to zero and by solving equation (4.8) to obtain [4.10]
dt

(4.10)

Substituting equation (4.10) into (4.9), the steady-state frictional coefficients are found to

be [4.10]

(4.11)
\ y r \

This steady-state solution can be used to calibrate and identify the model parameters by

fitting this model to experimental data or to the “Magic formula” which will be

introduced in the next section.

4.3.1 The “Magic Formula” .

One of the most well-known models for static frictional coefficient is the Pacejka’s model

[4.11], also known as the “magic form ula” . This model has been shown to suitably match

experimental results and accurately describe the tire steady-state curves. It has been the

benchmark for validating the steady-state conditions for dynamic tire friction models. In

Reference [4.11] Pacejka presented the following formula for describing the friction

function.

H = D sin ( C arctan ( B M —E ( B M —arctan ( B M )))) (4.12)

57

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 where M is either the longitudinal slip ratio s, or slip angle /?. The parameters B, C, D and

E can be identified through curve-fitting with experiment data. Different sets of

parameters can then be used to generate plots of the longitudinal friction coefficient, pix(s),

as a function of slip ratio, s, and plots of the lateral friction coefficient, ////?), as a

function of slip angle, /?. Thus, to observe the relation of steady-state friction

characteristics between the LuGre and the “magic formula” and to identify the parameters,

it is convenient to express the LuGre friction model in terms o f the slip coefficients.

4.3.2 Definition of Slip.

In vehicle dynamics, slip is the relative motion between a tire and the road surface it is

moving on. This slip can be generated either by the tire's rotational speed (co) being

greater or less than the free-rolling speed (Vx). It is usually described as a ratio or

percentage slip (s), or by the tire's plane o f rotation being at an angle to its direction of

motion, which is also known as the slip angle (fi). Examining equation (4.2), one may

realize that when the vehicle travels at a constant speed with no slip and no steer at the

contact interfaces between the tires and the road, that is, where Vr = 0, the right hand side

(RHS) of equation (4.2) becomes zero, or the dynamic deflection of the bristle, r\(t), is

constant; since the bristle has no initial deflection, the dynamic deflection is then rj(t) = 0 .

58

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 From equation (4.3) it is seen that the tire friction forces described by the LuGre friction

model is zero, or Ff= 0.

The relative velocity is defined in the wheel plane x-y where O is the contact point

(Figure 4.3) with the following components

Vrx = V cos ( 0 ) + ru) = VX + ru>

(4.13)
Vnj = V s m { j 3 )

so that

vv r = Jv
\ 'n r 2 +
' V
v r-y2 (4.14)

Direction iII Direction

of Motion || of Motion
rco

yb»
Vr
Vr
Vx

Vx

Figure 4.3a Slip for braking Figure 4.3b Slip for acceleration

In equation (4 .1 3 ) /v —
= V
W ^ + ~rv r ?y js
is w he e i hub translational speed along the

I / 1Vy ]. uj is the wheel hub rotational speed; @is the

direction of travel with components; [1

slip angle or the angle between the direction of travel and the longitudinal axis of the

wheel coordinate. Note that for the wheel-center coordinate frame defined in Chapter 3, a

59

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 wheel coordinate. Note that for the wheel-center coordinate frame defined in Chapter 3, a

negative wheel angular velocity co implies a forward motion.

The longitudinal slip s, and the lateral slip q, both need to be defined under two

separate cases, braking and acceleration. For the braking case (Figure 4.3a), the slips

(identified by a subscript b) are given by

V cos {(3) + ru! Vr

S6= V cos ( [3) = X
yr (4.15)
96 = 4 f - = tan(/?)
*X

For the case of acceleration (Figure 4.3b), the slips (identified by a subscript a) are

V c o s ( P ) + rui Vr
s = = —- rcj > —vr and w ^ 0
ru rw .
Vr (4.16)
qa = — = (1 - sa )tan (13)
rw

The longitudinal slip is always positive within the interval [0, 1]. When s = 0 there is no

sliding, whereas j = 1 indicates full sliding or skidding. The lateral slip is a function of

slip angle and directionally dependent.

With the slip rates now defined, n SSx cs ) and p,SSy ( (3 ) of equation (4.11) can be

determined. For braking, the sliding function g ( Vr ) in equation (4.5) becomes

-[£A cgg±a.»f (4_17)

S(F-) = M* + (Ms - Mfc)e 1 ;

and for acceleration

S(Fr ) = r t +(«,-A*i)e 1 l’' J l j

60

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Finally, the frictional parameters are, from equation (4.11)

<?(K) g(Vr )
I •. + &2x*z = + OfyVz 96(/5) (4.19)
'Jlb(0')~ + &b~ \/Qb (0)~ +

for steady-state braking at some constant velocity, V, and

9(Vr ) g(Vr )
4(5,?) ■a’2xrijJ ^0’ 4 ( ,5 S ,„ 9a (^) (4.20)
'Jla.ipy + Sa2 + "2sra;

for steady-state driving at some constant a>. It should be pointed out that the steady-state

behavior of the LuGre dynamic road-tire friction model can only be obtained for a

specified constant velocity V or constant angular velocity co, and may be validated with

the Pacejka’s “magic formula”.

4.4 Validation of Steady-State Behavior with the Magic Formula.

To validate the steady-state behavior of the LuGre model presented above, experimental

data presented in [4.2] are used as the basis of comparison. Reference [4.2] listed two sets

of parameters, used in conjunction with the “magic formula”, for vehicle braking and

cornering. These two sets of parameters are given in Table 4.1, along with parametric

values used in the present study. It is seen that parameter B has a value of 18.0 instead of

the 0.178 listed in [4.2]. The 18.0 value is found after a number of trials-and-error in the

present study. This is necessary in order to best-fit experimental data of [4.2]. The source

of the discrepancy is not fully understood. One possible cause may be the lack of units in

61

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 [4.2]. The resulting tire static curves, plots of friction force versus slip, are shown in

Figure 4.4.

As to the LuGre model, model parameters are identified by best-fitting the

steady-state behavior, equation (4.19), to the “magic formula” plots. In Figure 4.4, the

plots are those of forces Fx(s) and Fy(S). They are obtained by multiplying equation (4.19)

by a normal force, N = 2000N, which was used in [4.2]. The LuGre model parameters are

identified after a number of trials-and-error in the present study. They are listed in Table

4.2 as well. Note that cr2x and co.y are set to zero, which implies an dry frictional contact

assumption. It is also noted that the jus has a value greater than unity. This is because /us

and fxk are simply frictional parameters used in conjunction with the LuGre friction model.

They are not to be interpreted as the coefficient of friction used with the well-known

theory of dry (Coulomb) friction. In Figure 4.4, Fx is plotted as a function of s for braking

by setting /? = 2.0° and V = 60 km/h. On the other hand, the plot of Fy versus /? is obtained

with the setting of 5 = 0.05 and V = 1 0 km/h. The very close match between the “magic

formula” and the LuGre model as seen in Figure 4.4 shows that by selecting appropriate

parameters, the LuGre model provides an excellent representation for the steady-state

behavior of the road-tire friction.

62

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Table 4.1 Magic Formula Parameters
Parameters B C D E
Fx (braking) 0.178 1.55 2193 0.432
Ref. [4.2]
Fv (cornering) 0.244 1.5 1936 -0.132
Fx (braking) 18.0 1.55 2193 0.432
Present study
Fy (cornering) 0.244 1.5 1936 -0.132

Table 4.2 LuGre Model Parameters

Parameters Mk Es Vs S &2x &2y
Present Fx (braking) 0.72 1.35 5.5 0.75 0.0 0.0
study Fy (cornering) 0.65 1.35 5.5 0.75 0.0 0.0

2500 2000
M a g ic - • - M a g ic
-— L uG re
1500
2000
1000

50 0
1500

z z

1000
-500

-1000

-1500

-2000
0.2 0 .4 0 6 -20 -15 -10
L o n g itu d in al S lip , s S lip A ngle ( d e g ). (5

Figure 4.4 Steady-state friction forces F x(s) and Fy(fi)

4.4 Incorporating the LuGre Friction Model.

The dynamic LuGre model, in the form of equation (4.9), is derived under the assumption

that the wheels travel along the longitudinal, or x-direction only, In order to incorporate

the steering of the wheels, it is then considered that a wheel is rotated from the

longitudinal direction by an angle of, say, 6 (see Figure 4.5b). Since “The changes in the

63

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 bristle deflection are measured as a result of rotation of the measurement frame rather

than as a result of changes in the actual displacement” [4.10], only the transferring from

the measurement frame x-y-z to the global coordinate frame X-Y-Z is required. It should

be noted that in Figure 4.5a the measurement frame x-y-z initially coincides with the

X-Y-Z global coordinate frame. As shown in Figure 4.5, this measurement frame is in fact

the wheel base coordinate frame defined in Chapter 3. The wheel rotation is taken as the

steering input 6{t) from the driver. Note that the deflection of a bristle, tj, always makes

an angle of X with respect to the global coordinate X. Expressing the tj vector with respect

to the wheel base coordinate frame, one has,

X
ted)
r
/

/
(----- /
/
/
/

//

/
/

Figure 4.5a Bristle deformation Figure 4.5b Bristle deformation

before rotation after rotation

V = Vx%w + Vyjbw + V zk'bw (4.21)

64

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 The bristle deflection component ^ m a y be neglected because it is assumed to be small

compared to the longitudinal and lateral deflections. Equation (4.21) becomes

V = Vx%w + rjyjbw (4.22)

Next, the time derivative of the bristle deformation expressed in the wheel base

coordinate frame can be written as

V = V xkw + Vyjbw + Vx%w + Vyjbw (4.23)

Since the wheel base coordinate frame rotates only about zbwas the wheel steers (see

Figure 3.2), one has

4m ^0zbjbw > jbw ” !j^zb%w (4.24)

where u zb is the rotational velocity of the wheel about zbw. Substituting equation (4.24)

into equation (4.23) yields

V = ( Vx ~ UzbVy ) L + ( Vy + ^zbVz ) Jbw (4.25)

where the coupling effect due to combined translational and rotational motion of the

wheel is clearly reflected by terms such as ojzbrjx an du zbr)y. Next, equation (4.8) is revised

taking into account (4.25) so that

drhW = _ ^ , Wr\
dt rx g(VT) V x + zbVv
dVy( t ) (To, |F r I ( 4 '2 6 )
dt ry g ( VT ) Vy Uzbr]x

65

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Finally, the dynamic tire frictional coefficients with respect to the wheel base coordinate

frame are,

Mr = <70zVx ( t ) + CTla. + a 2 x V rx

VOxVxd) + CTlx + a2x^11

V'x ~ "g(vr' ) r]x + UJ:sbT,y
dr) {t)
(4.27)
= * o y V y ( t ) + o-iy + a 2 y V ril

y/ l"r0</IFr I „ ) | _ T^
= O-OyVy(t) + &ly Vm . Tjy ^zb V x ^2 -y^ry
ry ' / ( F)

The final task is to extend the friction model, equation (4.27), from a single wheel to the

full-car model. Since the wheels are not modeled as separate systems from the vehicle

body, the friction model needs to be defined with respect to the vehicle body frame as

was done in Chapter 3.

bw,2

cg 1

L a te ra l A xis, y
Y
A\

/'

L ongitudinal A xis, x

->X
GLOBAL

Figure 4.6 Total bristle deflection with respect to the global frame

66

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 As illustrated in Figure 4.6, the total rate of rotation of the wheel base frame with

respect to the global frame is the sum of the yaw rate of the vehicle body, p , and the

steering ra te 6 ( t ) . For a full-car model, the n-th ( n = 1 , , 4 ) bristle dynamic equation

and tire frictional parameters are obtained from equations (4.26) and (4.27)

Vx,n W = K x ,n ~ Vx,n + (< ? + #n C< ) )

9 { V r, n )
(4.28)
Vy, n W = V ry>n - ~ {<P + On ( t ) ) r ) x<n
9 ( Vr;n )

and,

M z ;, n & Q x V x ,n ^ lx V x ;n ( / ) & 2 x^ rx ,?

vv T X J I <rox\Vr,
X
( v r' \ Vx,n + (v3 + en (t ) )^ , n + CT2xVrx,n
9 VVr,n )
(4.29)
fty ,n Oy V y,n ( O ~h (7l y 9 y , n ^ ) ^2 -y^ry ,'

<*0y \Vr,n
^OyVy.n (t) ~f“ &iy V
v ry,n J ~ r V y,n ~{<P + 9n (t ))Vx,n + <*2iyry,n
91(\ r)'^ / /

where the relative velocity of the n-th wheel, Vrjl = {VrXtn,Vrytn}, is defined as the velocity

of the contact point of wheel n with respect to the ground. Since the velocity of the

ground is zero, the total velocity of the contact point becomes simply the relative

velocity,

F r* ,n V tr a n la tio n .n 4 " ^ r o t a ti o n ,n ( 4 .3 0 )

where the translational velocity of contact point equals that of the wheel center, or VCWJl,

as defined by equation (3.11). That is

V tr a n la ti(m ,n T'cw .n (4.31)

67

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 The second term, Vmtatum,n, represents the velocity of the contact point rotating about the

ycw n -axis of the n-th wheel (see Figure 3.12). If the wheel is spinning atujn ,

Frotation.n (4.32)

where r is the wheel radius which is assumed to be the same for all four wheels.

Substituting equations (3.11) and (4.32) gives

Vr,n = [i ~ <PVn +[y + n ]J + znk + rujJbWin (4.33)

In the above equation, znk may be neglected just as the component o f the bristle

deflection was neglected earlier in equation (4.22). Furthermore, unit vectorsi , j

and%WJl, j bWt7l are related via equation (3.2). Noting that i = = ]CU!iU, one then has,

v r,n = [(f - a ) cos en + (y + <pxn) sin en + unr] ibwn

(4.34)
+ {{y + ) cos 0n - { x - <pyn) sin 9n} j bw,n

4.6 Concluding Remarks.

The dynamic LuGre friction model, when applied to the full-car model, consists of

equation (4.29) where the relative velocities are determined by equation (4.34). The x-

and y-components of the frictional force, which are needed for the full-car model

developed in Chapter 3, are

68

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 ^j’x.n n

= °0xVx,n ( t ) + (Tl x K x ,n ~ - ° ^ r ’y Vx,n + { V > + 8n ( t ) ) V y j + ^2xK s Nn

9 1, *r,n )
(4.35)
fyy,n =
&0y | ^r,71
= VQyVysi, (t) + Ciy V
v r y ,n -
°j} ^ i rj _ (<p + 0n ( t ) ) r ] “I” ryji N„
9 {V r,n j y .

where the relative velocities are again determined by equation (4.34). In determining

frictional force components, steering angles9n{t){n = 1,..., 4) and their time derivatives

9n(t) (n= 1,..., 4) are considered inputs to the system rather than state variables. The

steering angles and their respective derivatives are independent of each other when

considering independent suspensions. Specifically - to paraphrase the suggestion by

[4.10] - the rear steering angles are fixed at zero steers, the front steering angles are

constrained according to Figure 4.7, “so that no wheel slip is induced by the steering

geometry, allowing for the possibility of zero relative velocity solutions at all four

wheels.”

- %

Figure 4.7 Vehicle steering geometry definition [4.10]

69

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 CHAPTER 5

NUMERICAL STUDIES OF VEHICLE DYNAMICS

In Chapters 3 and 4, all the necessary components for a vehicle dynamic model have been

developed. Next, the model is to be put through different cases to verify the m odel’s

capability in representing various vehicle dynamic situations, the assumptions behind

different simplifications, and the m odel’s accuracy. The selected cases are from recent

publications in the area of vehicle dynamic and control, including issues of ride comfort

and handling.

The objective is to utilize the present vehicle model to reproduce the results as

presented by other researchers. Identical conditions and physical parameters are

identified and implemented where applicable. Because of assumptions made in the

referenced publications, modifications of the presently developed vehicle model are

necessary. The chapter emphasizes on comparing the models, and identifying the

differences or similarities among them, so as to demonstrate the range of applicability of

the present model.

70

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 5.1 A Planar Model of a Three-Wheeled Vehicle.

5.1.1 A H alf-C ar Model by Gawade et.al.

An in-plane seven degrees-of-freedom (DOF) mathematical model of a three-wheel

vehicle (TWV) was presented in [5.1] to study the effect of road bumps on occupant

injury. The system equations may be used to calculate the forces and positions of

interacting components while the TW V was passing over bumps o f different profiles, and

to examine the lift-off phenomenon (which is a measurement of vehicle stability) and the

ride comfort of the TWV. Wheel lift-off occurs when the normal reactions exerted by the

road onto a wheel goes to zero, causing the tire to lose contact with the ground. Ride

comfort, on the other hand, can be measured by the time history of the maxim um upward

acceleration and/or the frequency spectrum of the acceleration.

The planar TW V model

CO, V

shown in Figure 5.1 (where

symbols have been changed from

those of [5.3] for the sake of

consistency) was assumed to

Z

Figure 5.1 The planar TWV model travel over the bump with a

constant longitudinal speed and without steering. The vehicle body, front wheel and a

71

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 lumped rear wheel were considered rigid, and were connected with linear suspensions.

Though the wheels themselves were considered rigid, tire stiffness was incorporated into

the model. The seven DOFs considered were the longitudinal motion, bounce and pitch

for the vehicle body; and the bounce and spin motions for each of the wheels. Vehicle

body’s lateral motion, roll and yaw were not considered since it was a half-car planar

model. Furthermore, the no-slip assumption was made, such that cu — —x / r , and the

road-tire friction forces were [5.1],

n = 1,2 (5.1)
r' r

5.1.2 The Present Model.

Introducing some simplification into the general model presented in sections 3.4.6 - 3.48

by setting y = y — y = <j>= ^ = (j) = (p = ip = i p - Ffy,n - 0, the equations of motion for the

bounce and spin motions of the wheels are then given by

k stn( z XrTI Zn ) + Cs n ( z 2Vj7 %n) k tn {z n zsn)

(5.2)

Equation of motion for the bounce motion of the vehicle body is given by

M bZ = - J 2 i k Sn U ~ x n l ~ Z n ) + Cs n { z - X n7 - Zn ) ] (5.3)

and for the longitudinal and pitch motions of the vehicle body, one has

M t0t% ( Ffx.n ) (5.4)

72

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Iy,cgl y ' (% ,n% ) ^ ^X^ f x - n ^ w j i ) ^ •UiT z n)^n] (5 .5 )

“t” y 1[Csn(% %n'i A i^n]

In equations (5.2) through (5.5), m } is the mass of the front wheel, and m 2 is the

combined mass of the two rear wheels. Accordingly, Ispanj is the mass moment of inertia

of the front wheels, and Ispan,2 is the combined mass moment of inertia of the two rear

2
wheels. The wheel index n runs over 1 and 2, o r ^ = ] P ,andM t0( = M h + mx + m 2 . These
n=l

equations of motion are proven identical to those proposed in [5.1]. As to the longitudinal

frictional force, Ffx,n, the LuGre friction model may be implemented to better capture the

behavior of the tires when in contact with the road. The parameters o f the LuGre friction

model used for the present study are taken from [4.7] and listed in Table 5.1.

Table 5.1 Parameters for LuGre Friction Model

Parameters Values Units

H-k 0.57 -
t*s 1.41 -
vs 2.66 m/s
8 0.5 -
OOx 267.00 m-1
<?lx 1.33 s/m
$2x 0.0001 s/m

73

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 5.1.3 Results and Discussions.

The differential equations of motion for both models are then solved using M atLab®. For

comparison, one of the scenarios investigated in [5.1] has been chosen where the vehicle

travels over a road bump modeled as a half-sine wave with amplitude o f 0.1 m and

transverse distance of 2.0 m (Figure 5.2). The bump is located 0.5m ahead o f the center

of front wheel along the longitudinal axis. The vehicle is set to travel at its wheel lift-off

speed of 5.11 m/s which was determined by .1]. Other parameters are given in Table

5.2. Wheel torques Ti and T2 are zero.

Table 5.2 Parameters for the TWV

Parameter Value Unit
Mb 504 kg
1 ~--v 1
'\ mi 8.5 kg
\

/
\
\ m2 18.0 kg
\

/
\
\ ksi 32736 N/m
/ \

/ \ k S2 105414 N/m
\
\
/ kti 238260 N/m
\ .
/
kt2 500980 N/m
/ \ ■
j
\ Csl 3250 N.s/m
\
i \ Cs2 6470 N.s/m
j
\
QC----------- 1----------- 1----------- 1----------- X] 1.437 m
0 .5 1 1.5 2 2.5
Traverse distance (m)
X2 -0.563 m
h 0.530 m
Figure 5.2 Half-sine wave road profile
r 0.210 m
ly.cg 170.000 kg.m 2
Ispin, 1 0.110 kg.m 2
/spin,2 0.220 kg.m 2

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 The normal reaction forces, vertical acceleration of the vehicle body and its

frequency spectrum from both models are shown in Figures 5.3 and 5.4. It is clearly seen

that both models yield identical time and frequency responses. Wheel lift-off is seen

when the rear wheels are positioned at x = 3.65 m (Figure 5.3b) and the peak reaction is

observed at rear wheels at x = 4.66 m.

4000
- P re se n t Model
■ TWV Model
3500

3000

S. 2500

£ 2000
° 1500

1000
500

T rav erse d ista n ce (m)

Figure 5.3a Normal reaction at front wheel

10000
— P re se n t Model
9000 TWV Model

8000

7000
2
6000
©
(_»

£ 5000
2
o
z
4000

3000

2000
1000

T rav erse d istan ce (m)

Figure 5.3b Normal reaction at rear wheels

75

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 As to ride comfort, results of the time history and the frequency spectrum of the

vertical acceleration (Figure 5.4) show the same level of vibration with dominant

frequency at about 2 Hz with an amplitude of 5.25 m/s2.

r\ — P resen t Model
TWV Model

J)
i
in

'8
SCIJi
c
cs
c0
153
<D
<_>
< -10
u

-15
0.5
Time (sec)

Figure 5.4a Vertical acceleration of the vehicle body - Time history

5.5
— P resen t Model
CN - - TWV Model
ch
1 4.5
in

g 4
N
g> 3.5
o
2 3
o
2.5

oca
"5
X<
3
D
D
Q.
a
<
0.5

Frequency, (Hz)

Figure 5.4b Vertical acceleration of the vehicle body - Frequency spectrum

To make sure that the no slip condition is true, longitudinal speed of the vehicle body and

the spinning velocity of the wheels are plotted (Figure 5.5), which verifies that

76

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 the relation cu = —x / r is maintained. Frictional forces are shown in Figure 5.6. They are

constantly zero.

Recalling from Chapter 4 (see Sections 4.2.1 and 4.3.2, in particular) that when the

vehicle travels at a constant speed with no slip and no steer, the tire friction forces will be

constant and equal to zero. The wheels will experience the pure rolling motions. In the

LuGre friction model, the frictional force is given as a function of the bristle deflected

behavior and is depended on the rate of change in relative speed or slip at the contact

surface between the tire and the road. Thus, with no-slip at the contact interfaces and the

expression given by equation (5.1), one may realize that a non-zero tire friction force

appears only when the vehicle is imposed by external forces or wheel torque for

acceleration or deceleration. The constant zero friction forces represent the dynamic

equilibrium and the steady-state motion of vehicle and wheels, which is what the present

planar TW V model is expected to simulate. Note that the constant longitudinal speed, the

zero steer and the absence of wheel torques all point to the vehicle moving along a

straight path, and having zero acceleration. Since there are no applied wheel torques,

friction forces at the tire-road contact points will have to be zero for N ew ton’s second law

to be satisfied. The friction forces may be expressed as in equation (5.1) or by the LuGre

friction model, with the latter being chosen for the present model.

77

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 a. Longitudinal velocity, Ref. [5.1] b. Longitudinal velocity, present model
Angular Velocity (rads/s)

8 10 8 10

c. Spinning velocity, Ref. [5.1] d. Spinning velocity, present model

Figure 5.5 Comparisons of the present model and Ref. [5.1]

Front W heel
0 .8

0 .6

0 .4

0 .2
%
<3> 0
s
u.
- 0 .2

- 0 .4

- 0 .6

- 0 .8

0 2 4 6 8 10
T r a v e r s e d i s t a n c e (m )

Figure 5.6 Frictional forces

78

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 5.2 A 7-DOF Full-Car Model.

A 7-DOF full-car model was discussed in [5.2], The work focused on ride comfort with

the implementation of active-suspensions, and a PID controlled passenger seat. The

controls of the vehicle body’s bounce and pitch motions, and the seat’s bounce motion

were used alternatively to obtain the optimal comfort. In the suspension systems, a linear

dry friction model was introduced in studying the vertical response of the vehicle body.

In what follows, only the mathematical structure of the vehicle model is extracted, and

the present model is simplified to having the same DOFs. The aim is to investigate the

vehicle’s responses to vertical road disturbances.

5.2.1 A Full-Car Model by Rahmi.

This model was reviewed in Section 3.5.4, see Figure 3.18 in particular. The DOFs

included the bounces of the vehicle body and the wheels, respectively; and the roll and

pitch of the vehicle body. The model was considered to be stationary in the sense that the

horizontal motions, that is, the longitudinal, lateral and yaw motions, were neglected.

Wheel spinning was also discarded. The resulting state equations were nonlinear because

of the trigonometric terms involved. In addition, the model consisted of four independent

suspensions with friction on dampers. The friction was described by, where n = 1,..., 4

79

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 fn = C en (A zn ) (5.6)

N if |A i„ | < e
C&n (5.7)
^ (2p„ - s in 2pn ) + ■ cospn otherwise
IT \ L \ Z n j

P n = Sin ( £/ A % (5.8)

In the above equations, A zn is the relative velocity between the two ends of the n-th

suspension, and is determined by equation (3.12). The parameter e is a small constant

band values implemented to prevent the complete locking of the suspension

when A zn = 0. The constant R is given as the dry friction force under the condition of

low A zn . In the range where |A in | > e , the damping friction approaches R. Reference

[5.2] listed the values of e and R (see Table 5.3). They had been verified with

experimental data, see [5.2]. Figure 5.7 shows the behavior of this friction model.

Table 5.3 Parameters s and R

<i>
o
Parameters Values Unit oL
L -e
C
o
£ 0.0012 m/s Ll.

-10
R 22 N
-1 5

from [5.2]
-20
-0 .0 1 - 0 .0 0 5 0 0 .0 0 5 0 .0 1 0 .0 1 5 0 .0 2
R elative ve lo c ity o f g a m p e r end, A Z n (m /s)

Figure 5.7 Dry friction behavior

80

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 It should be pointed out that the additional DOF representing the seat bounce motion

is excluded, in order to compare with the present model. As a result, passive, instead of

active, suspension is employed. Time responses of the model are made by simulating the

vehicle traveling over the bump without steering.

5.2.2 The Present Model.

The present model is reduced to 7 DOFs since,x = x = x = y = y = y = <fi=ip = ip = ujn = ujn = 0.

In addition, the external forces components generated from road-tire interactions and

wheel torque inputs are not considered. As a result, the internal reaction forces Pxcw,n and

PyCW,n are zero. The new set of state equations becomes, for the bounce motions of the

wheels,

Ttln Z n — k sn {z + Vn xrd z n)

“f c s n ( z + yjp Xn 7 Zn ) ktn(zn “ zsn) 4- fn (5.9)

for the bounce motion of the vehicle body,

^s n ( % Un$ %n )
£ /„ (5.10)
+ c 57l( i + yn<j>- xni - zn)

and for the pitch-roll motions of the vehicle body,

I x , c gi = “ I ] {[Kni* - A , 7 + y j - Zn) ] }

{[c » ( i - A .7 + y„4>- A )]< /„ } + £ ynfn

(5.11)
Iy,CgX = X 3 {lk sn(Z - A , 7 + Vn4> ~ Zn ) \ Xn )

+ J 2 { K n ^ - A .7 + y j - A , ) K } + £ * „ / „

For easy comparison, the dry friction model of equations (5.6) through (5.8) is used.

81

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 5.2.3 Simulation Results and Discussions.

To observe the time responses of the models, both models are hypothetically given a

constant vehicle speed of 10 m/s over the bump without steering. The model parameters

are listed in Table 5.4. A ramp-like bump profile is selected with a height of 0.035 m and

a span of 0.1 m, as shown in Figure 5.8. The road disturbance is inputted to each wheel

with a time delay (see Figure 5.9) between the front and rear axles. That is, the time delay

is A t = (a + b)jV , where a + b is the wheelbase (see Figure 3.6), and V the longitudinal

speed of the vehicle.

0 .0 4
• F ro n t W h e e l;
R e a r W h e o ls
0 .0 3 5

0 03

0 .0 2 5

- T im e D e la y -
2 0.02 | 0.02
2 0 015
E
do 0 .0 1 5

001
0 .0 0 5

004 0 .0 6
0 0 .0 5 0.1 0 .1 5 0.2
B um p sp a n (m) T im e ( se c )

Figure 5.8 Ramp-like bump profile Figure 5.9 Time-delay at 10 m/s

The simulation results are shown in Figures 5.10 and 5.11. The time responses from

both models are found to be identical. The time-delay effect of axles traveling over the

bump is clearly seen in vertical displacement and acceleration, and pitch angle (figure

5.10); it is also seen in the dampers’ friction forces (Figure 5.11).

82

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Table 5.4 Model Parameters

Vehcile Body
Parameter Description (Unit) Value
Mb Vehicle point mass at CG (kg) 1100
X j, x2 x-coordinates of front axles to CG (m) 1.2
X j, X 4 x-coordinates of rear axles to CG (m) -1.4
yi, ys y-coordinates of left wheels to CG (m) -0.75
yi, y4 y-coordinates of right wheels to CG (m) 0.75
b,cg Moment of inertia about x-axes (kg m2) 550
bxg Moment of inertia about y-axes (kg m2) 1848
Wheels
m.], m 2 Mass for front Wheels (kg) 25
m 3, m 4 Mass for rear Wheels (kg) 45
Suspension/Tire Stiffness
k s i , k s2 Front suspension spring coefficient (N/m) 15000
k s 3 t k s4 Rear suspension spring coefficient (N/m) 17000
Cs All suspension damping coefficient (N.s/m) 2500
k t All tire stiffness coefficient (N/m) 250000

83

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 . x 10
2.5
Present Model
Full-Car

1.5

£®
oo
atsi
5

0.5

2 3
Time (sec)

Figure 5.10a Vertical displacement of vehicle body

x 10'

Present Model
Full-Car

0.4

a -0.2
-0.4

-0.6

Time (sec)

Figure 5.10b Pitch angle of vehicle body

84

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 8
P resen t Model
Full-Car

<

•2

-4
0 0.2 06 0.8 1
Time (sec)

Figure 5.10c Vertical acceleration of vehicle body

Left Front Right Front

40 40
------ Present Model ------ Present Model
— - Full-Car —- - Full-Car
20 20
z Z
c:
o 0 <
o
-
__ - — ------- ---------------- U
tj
LL
-20 \ r
L)
LL
-20
j
/

-40 -40
1 2 3 1 2
tim e(sec) time(sec)

Left Rear Right Rear

40 40
Present Model Present Model
Full-Car Full-Car
20 20
z z
c co
o 0 0
tj
L.
o
LL LL
-20 -20

-40 -40
1 2 1 2
tim e(sec) time(sec)

Figure 5.11 Damper friction forces

85

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 5.3 A Full-Car Model for Maneuver Simulation.

5.3.1 The Full-Car Model by Villella.

In studying ground vehicle handling, a 7-DOF full-car model with the application o f the

2-D LuGre friction model has been introduced by [5.3] of Georgia Institute of

Technology. The focus of the study was the effect of the wheels’ input torques on the

lateral-yaw response of the vehicle. The vehicle model included five lumped masses, one

lumped translational mass that was the vehicle body, and four lumped rotational masses

that were the four wheels. The translational mass was to represent the horizontal motion

of the vehicle body while the rotational masses were to represent the spinning motion of

the wheels. M ost importantly, the model had no suspensions due to the assumption that

“suspension forces are internal to a vehicle system and have no effect on the motion of

the entire system in the horizontal plane” [5.3]. As a result, the effects of pitch-roll, and

the bounce of the wheels were neglected while the vehicle was cornering. The DOFs

included the longitudinal motion, and the lateral motions and the yaw of the vehicle body;

and the spins of the wheels, for a total of seven.

In addition to the absence of a suspension system, an analytical method for solving

for normal force distribution amongst the four wheel contact points was proposed, where

the solution produced the forces necessary to maintain zero pitch and roll conditions. The

86

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 resulting algebraic equations depended only on the vehicle’s geometric parameters,

friction forces, and steering angles. As to frictional force, the 2-D LuGre model of

Reference [4.2] was implemented. In contrast to the static friction model, such as the

“magic formula”, the large transient friction forces were captured as the steer angle is

changing, where a rapid increase of contact forces may be achieved.

The inputs to the model

were composed of four

a+b
independent wheel torques, and

:'-5©
Reference axis
Cr four time-varying steering
c+d

signals sent from the driver.

Figure 5.12 Steering geometry definition
(showing right steering) [5.3]
These four steering signals were

not independent. As demonstrated by the steering geometry definition of Figure 5.12,

which shows only the geometry of a right steer, the rear steering angles were fixed at zero

steer, and the front steer angles, d\ and 62 , were constrained by having their respective

y-axes, Ri and R 2, intersect at the same point. The reason of so doing was, according to

[5.3], “so that no wheel slip is induced by the steering geometry, allowing for the

possibility o f zero relative velocity solutions at all four wheels.” In other words, in order

not to introduce “artificial slip”, the input steer angles are not independent. Instead, the

87

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 (a + b)
angles are given by 0\,d2 = tan 1 and 0-i — 6a = 0 , for the case o f left
~ ± P - + l,c + d)
tanc^

{a + b )
steering; and by 0\,02 = tan 1 and #3 = 0A = 0 , for the case of right
(a + 6) _ (c + d)
tan#!

steering, where (a + b) and (c + d) are the wheelbase length and width of the vehicle,

respectively.

5.3.2 The present model.

The present model, whose development was dealt with in Chapter 3, has 14 DOFs in total,

and considers the vehicle body and the wheels as two inter-dependent sub-systems of

rigid-bodies. The suspensions, or the “spring-damper” units, represent the connections of

the wheels to the vehicle body, and introduce the required inertial coupling, restraints and

forces between the two sub-systems. Normal forces are solved from the consideration of

dynamic equilibrium. It is interesting to note that, although the present model has 14

DOFs compared with the 7 DOFs used in [5.3], it requires less CPU time than that by the

approach of [5.3]. A typical run of [5.3] takes 38 seconds of CPU time; while the present

model requires, on average, 5 to 10 seconds less. In addition, it should be noted that in

the remainder of this chapter, the results obtained by using the approach of [5.3] will be

denoted “G-Tech” in the plots for the abbreviation of Georgia Institute of Technology.

88

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 5.3.3 Simulation Results and Discussions.

The state-space equations of motion for both models are solved using MatLab®. A few

scenarios are considered. Note that for all scenarios, the vehicle is maneuvered under the

steering input of 9\(t) given in Figure 5.13. The steering transition begins with zero to -10

degrees (-0.176 radians) and back to zero, which causes the vehicle to turn left. The other

( a 4- b )
three steering inputs are therefore, ,02 = tan-1
(a + b)
+ ( c H- d )
ta n ^

The other model parameters are

-002
-0 04

listed in Table 5.5. Note that the friction

-0 .0 6

parameters are taken from [5.3]; the

-012
-0 .1 4
vehicle mass, mass moment of inertia

-0 .1 8
0.5 2.5 3.5 and geometry parameters are taken
tim e(sec)

Figure 5.13 Driver steering input

from measurements on a 1998 Honda

Civic by the United States National Highway Traffic Safety Administration (NHTSA)

[5.4]; the spring-damping-tire stiffness constants are taken from [5.2], It should be

pointed out that, though the w heel’s mass moment of inertia is taken from [5.3], the

wheel mass and geometry are chosen according to Honda’s specification [5.4] owing to

data availability.

89

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Several simulation scenarios are examined. First, the vehicle is given steer angles for

a turn maneuver on pavement road with no wheel torque input to the wheels. Next, the

vehicle is subjected to the same steer angles and road surface conditions but input torques

are applied to all four wheels so as to simulate four-wheel drive. Finally, the model is put

to an icy road surface and simulations are performed for a four-wheel driven, a

front-wheel driven and a rear-wheel driven vehicle, respectively.

5.3.3.1 Simulation of turn maneuver without input torques to the wheels.

Initially, the model is simulated under turn maneuver on pavement road with no wheel

torque input to the wheels. The simulation is run with an initial longitudinal speed o f 15

m/s [5.3]. Initial wheel speeds are then determined via no-slip conditions = rui, leading

to the values of -75 rad/s. The negative sign is needed due to the use of wheel center

coordinates, see Section 3.1, and Figure 3.2 in particular. All other states are initially zero.

The simulated results show the extent to which vehicle dynamics is influenced by the

friction between the road and wheels.

90

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Table 5.5 Model Parameters

Parameter Description (Unit) Value

Vehicle Body [5.4]
Mb Vehicle point mass at CG (kg) 1140
X i, x2 x-coordinates of front axles to CG (m) 1.1
X3, X 4 x-coordinates of rear axles to CG (m) -1.5
yi> yj y-coordinates of left wheels to CG (m) -0.7
y2,y4 y-coordinates of right wheels to CG (m) 0.7
H +r Height from ground surface to CG (m) 0.5
h.cg Moment of inertia about x-axis (kg m2) 365
ly.cg Moment of inertia about y-axis (kg m2) 1617
Iz.cg Moment of inertia about z-axis (kg m2) 1785
Wheels [5.4]
mn Mass of one wheel (kg) 25
r Wheel radius (m) 0.2
I spin M oment of inertia about ycw (kg m2) 0.1361
Suspension and Tire Stiffness [5.2]
ks Suspension spring constant (N/m) 17000
cs Suspension damper coefficient (N.s/m) 2500
k, Tire stiffness constant (N/m) 250000
LuGre Friction [5.3]
GOx Longitudinal rubber stiffness (m"1) 178
Olx Longitudinal rubber damping (s/m) 1
Olx Longitudinal viscous relative damping (s/m) 0
OQy Lateral rubber stiffness (m '1) 500
<J\y Lateral rubber damping (s/m) 2
&2y Lateral viscous relative damping (s/m) 0
s Stribeck exponent 0.5
Vs Stribeck velocity (m/s) 5.5
Coefficient of Friction [5.3]
Mk Kinetic, rubber-asphalt contact 0.8
Ms Static, rubber-asphalt contact 1.2
Mk Kinetic, rubber-ice contact 0.1
Ms Static, rubber-ice contact 0.2

91

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 L e ft F r o n t W h e e l R ig h t F r o n t W h e e l
4 4

2 2
z 0
0 ON
>< X
LL
O

■2 ■4
■4 ■6
0 2 3 4 0 1 2 3 4
tim e(sec) tim e(sec)
Left R ear W heel Right Rear W heel
6 4
4 2
2
LL

■2
■2
■4 ■4
0 1 2 3 4 0 1 2 3 4
tim e(sec) tim e(sec)

Figure 5.14a Longitudinal friction forces

Left Front Wheel Right Front W heel
3000 6000

2000 4000
z

>
1000 > 2000
CN
LL LL

-1000 -2000
tim e(sec) tim e(sec)
Left Rear W heel Right Rear W heel
1500 6000

1000 4000
z
^ 2000
> .
LL LL

-500 -2000
tim e(sec) tim e(sec)

Figure 5.14b Lateral friction forces

92

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Vehicle Longitudianl Veloctiy

14;

14.6

14.4
0.5 1.5 2 2.5 3.5
Vehicle Lateral Velocity
0.5

-0.5

1.5 2 2.5
Angular Velocity of Vehicle
0.5

-0.5

0.5 2.5 3.5

Figure 5.15 Vehicle body speeds for pavement road with no torque input

1) Vehicle body velocities in the horizontal plane are shown in Figure 5.15. From

Figure 5.14a, it is clearly seen that the speed change as shown in Figure 5.15 is

due to the emergence of friction forces. The longitudinal speed decreases slightly

after the turn begins (0.5 s versus 0.35 s, see Figure 5.15), where the presence of

longitudinal frictional forces dissipates the forward kinetic energy, resulting in a

lower speed slightly ahead of the completion of the turn (t = 2.75 s versus 3.0 s,

see Figure5.13). On the other hand, the appearance of lateral frictional forces

accelerates the vehicle in the lateral direction, and the moment created by these

friction forces about the mass center of the vehicle body causes the yaw motion of

93

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 the vehicle. In other words, the lateral speed and yaw rate increases and returns to

zero at the completion of the turn without visible time-lag or advance (Figure

5.15).

2) The motions of the wheels are also influenced by friction. See Figure 5.16 for

each w heel’s rotational speed, where, due to the use of wheel center coordinates, a

negative rotational speed represents a forward rolling motion. The wheel’s

rotational speeds on the left and right sides of the vehicle diverge as the left

wheels slow down and the right wheels speed up to traverse turns of differing

radii. One may also refer to Figure 5.12. As the vehicle turns about the center of

rotation, CR, each wheel center will rotate at some constant angular speed with

respect to CR', as a result, the linear speeds of the wheel centers satisfy, for a left

turn in particular, | Vcw2 \ > \Vcwi \ > \VcwX \ > \ Vcw3 \ , since it is seen that

V %/
R2 ^ ^ ^ . Thus, given the relation ofw^ — , the wheels

rotational speeds satisfy | cu21 > | cu4 1 > | ^ | > | cj3 | , which is clearly exhibited in

Figure 5.16.

3) As shown in Chapter 4, the bristle deflections are a result of the relative motions.

The computed relative velocities and bristle deflections are given in Figures 5.17

and 5.18. The bristles deflect longitudinally and laterally when steering begins,

94

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 reach steady-state values as the steer angle is held fixed, and return to zero

deflection when cornering is com plete. The dynamic coefficients of friction

are proportional to bristle deflection, which is verified by Figure 5.19 where

the dynam ic frictional coefficients are seen to have identical traits to bristle

deflections shown in Figure 5.18, albeit different m agnitudes.

-69

-70

-71

> -74

-76

-77

-78
0.5 2.5 3.5
tim e (se c)

Figure 5.16 W heels’ rotational speeds with no torque input

95

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 x 'lO " 3 L e ft F r o n t W h e e l x -)0 R ig h t F r o n t W h e e l
2 3

2
CO1 ‘'“I
E,
ON
2 £
0 >
0

1 1
0 1 o 3 4 0 1 2 3 4
tim e(sec) tim e(sec)
x10’ Left Rear Wheel x10 Right Rear W heel
20 20

15 15

10 10
00
a 5 5
>
0 0
•5 ■5
0 1 2 3 4 0 2 3 4
tim e(sec) tim e(sec)

Figure 5.17a Longitudinal relative velocity of each wheel with no torque input
Left Front Wheel Right Front Wheel
0.8

0.6 0.6

0.4
1 °'4
0.2 & 0.2
>
0
- 0.2 - 0.2
0 1 2 3 4
tim e(sec) tim e(sec)
Left Rear Wheel Right Rear Wheel
0.015 0.015

0.01 0.01

| 0.005
:/ J\ 7
1- 0
>
-0.005
|r
- 0.01 - 0.01
0 1 2 3
tim e(sec) tim e(sec)

Figure 5.17b Lateral relative velocity of each wheel with no torque input

96

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 : -IO"6 L e ft F r o n t W h e e l x 10 R ig h t F r o n t W h e e l
4 2

2
0
0 <N
X
PC
X
P" 1
■2 ■2

■4 •3
0 1 2 3 4 0 9 3 4
tim e(sec) tim e(sec)
x10“ Left R ear Wheel x 1 0 6 Right R ear Wheel
2

1
0.5

E 0
CO
X
SC
-0.5
•2
■3
0 1 2 3 4
tim e(sec) tim e(sec)

Figure 5.18a Longitudinal bristle deflection at each wheel

x io Left Front W heel x10 Right Front Wheel

3

2 2

E,
1
>.

0 0

1 1
0 1 2 3 4 0 1 2 3 4
tim e(sec) tim e(sec)
x10 Left Rear W heel x10 Right R ear W heel
3 3

2 2
E,
CO 1 ^r 1
>< EC

0 0

1 1
0 1 2 3 4 0 1 2 3 4
tim e(sec) tim e(sec)

Figure 5.18b Lateral bristle deflection at each wheel

97

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 x 'io L e ft F r o n t W h e e l x io R ig h t F r o n t W h e e l

0.5
1
<N
0 ’V _ ..
-0.5

2
0 1 2 3 4
tim e(sec) tim e(sec)
x 1 Q J Left Rear Wheel x10 Right Rear W heel
4

0.5
2

0
-0.5
■2

■4
0 1 2 3 4
tim e(sec) tim e(sec)

Figure 5.19a Longitudinal dynamic friction coefficients

Left Front W heel Right Front W heel

_/

-0.5 -0.5
tim e(sec) tim e(sec)
Left R ear W heel Right R ear W heel

-0.5 -0.5

tim e(sec) tim e(sec)

Figure 5.19b Lateral dynamic friction coefficients

98

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 4) Normal forces exerted on the wheels are shown in Figure 5.20. Higher normal

forces at the front wheels are the result of a center of gravity which is closer to the

front wheels. As the vehicle turns, the normal forces increase on the right side and

decrease on the left side by the same amount. This means that the springs and tires

on the left side of the vehicle are compressed more than those on the right side.

Since the road surface is assumed even, the differences in tires’ normal

compressions or forces could result in the different lateral friction forces among

the tires and further introduce a greater or a less turning moment to facilitate the

tuning of the vehicle.

Left Front Wheel Right Front W heel

4000 5000

3500 /\ 4500
\ / \ ■
\ / z I
i
“ 3000
\\ / CN 4000 1
z 1 Z
!I
2500 3500
\w ______ j \
2000 3000
0 1 2 3 4 0 1 2 3 4
tim e(sec) tim e(sec)
Left R ear W heel
3000 4000
r - — \
2500 3500

2000

1500
' \

\
/ "
“
z
3000

2500
/ \
1000 v — 2000
0 1 2 3 4 0 1 2 3 4
tim e(sec) tim e(sec)

Figure 5.20 Normal forces during turn maneuver

99

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Finally, by plotting the vehicle’s position with respect to the global frame one is able

to show the path of the vehicle, see Figure 5.21, where the dots represent successive

positions of the vehicle’s center of gravity, CG, and the arrows show successive traveling

directions of the vehicle. The result is as expected, since the vehicle moves along a

straight path, then turns left, and moves straight forward once the steering input ceases.

Figure 5.21 also presents the path of the vehicle by the approach of [5.3] which is

denoted as “G-Tech” as opposed to the “Present M odel”. The very close match seen in

Figure 5.21 will be seen again in Figures 5.25b and 5.26b, for example.

45

Present Model
35

30

G-Tech
I 25 *
Q.

-45 -40 -35 -30 -25 -20 -15 -10

Y-abs. p o sitio n (m)

Figure 5.21 Vehicle position with no torque input

100

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 5.3.3.2 Simulation of turn maneuver with input torques to the wheels.

Next, an all-wheel drive vehicle is assumed. The present model is simulated by applying

a 50 N-m input torque to each of the four wheels. Note that a positive torque will cause a

wheel to roll forward (see Figure 3.12). The vehicle will accelerate on a paved road and

under the same steering input as given in the previous scenario. The results of vehicle

velocities are shown in Figure 5.22. The vehicle is indeed accelerating since the

longitudinal speed has increased over the time period. However, the lateral velocity and

yaw rate are lower than in the case o f no applied torque. The lower lateral and yaw rates

give rise to less turning motion. Such an effect is seen from the different paths that the

vehicle will follow (see Figure 5.22). With torques applied to the wheels, the vehicle

makes a wider turn. This is the result of a lower lateral frictional force “pushing the

vehicle to turn.”

The longitudinal frictional forces increase by an almost equal amount among the

four wheels (Figure 5.23a). These forces are negative because o f the sign convention

defined in Chapter 3, see Figure 3.12. Magnitude-wise, they are close to T J r n = 250 N,

which is the tangential force at the wheel base and produced by the torque. On the other

hand, the lateral frictional forces see slight increase or reduction, compared with the

101

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 pervious no-input-torque case. Overall, lower total lateral frictional force and turning

moment lead to reduced lateral and angular accelerations of the vehicle body.

In addition, relative velocities Vrx and V,y are plotted in Figure 5.24. According to

the LuGre friction model, frictional forces are directly affected by Vr. It is seen in Figure

5.24 that Vrx and Vty of the front wheels are much higher than the no-input-torque case.

Since g(Vr) becomes closer to ut when Vr increases, see equation (4.5), it is concluded

that the increase in Vr causes a lower bristle deflection which in turn reduces the force

produced by the bristle deflection. This means that the wheel may be unable to produce

sufficient lateral frictional force to steer the vehicle as it did previously.

Vehicle Longitudinal V eloctiy

ON-m
- 50 N-m
E.
■X

1.5 2 2.5
V ehicle Lateral V e lo city
I I

0
E, X..
■>. -1 O N-m
50 N-m
-2
0 0.5 1 1.5 2 2.5 3 3.5 4
Angular V e lo city of Vehicle
0.5

-0.5 v. A O N-m
50 N-m
-1
0 0.5 1 1.5 2 2.5 3 3.5 4
tim e (sec)

Figure 5.22a Comparison of vehicle body speeds

102

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 45

50N-m Torque
Each wheel

No Torque Input

10 -

0 ----------------1----------------'----------------1----------------1--------------- i-
-50 -40 -30 -20 -10 0
Y-abs. position (m)

Figure 5.22b Comparison of vehicle position

Left Front W heel Right Front W heel
100 1UU
0 N-m 0 N-m

2
^ -100
X
LL

-200 -200

-300 -300
tim e(sec) tim e(sec)
Left R ear W heel Right Rear W heel
100 100
0 N-m 0 N-m
50 N-m — 50 N-m

-100 f -100
X
LL

-200 -200

-300 -300
tim e(sec) tim e(sec)

Figure 5.23a Comparison of longitudinal frictional forces

103

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 L e ft F r o n t W h e e l R ig h t F r o n t W h e e l
3000 6000
- 0 N-m — 0 N-m
- 50 N-m - • • 50 N-m

2000 4000
z
1000 CN 2000
>.
LL
>S
LL

-1000 -2000
tim e(sec) tim e(sec)
Left Rear W heel Right Rear W heel
1500 6000
0 N-m
50 N-rr

1000 4000
z z
500 ^ 2000
LL LL

-500 -2000

tim e(sec) tim e(sec)

Figure 5.23b Comparison of lateral frictional forces

Left Front Wheel Right Front Wheel

0.05 0.02
— - 50 N-m

-0.05 -0.02
a
>
-0.1 -0.04

-0.15 -0.06
time(sec) time(sec)
Left Rear Wheel Right Rear Wheel
0.01 0.01

CO
-0.01 -0.01
>a
- 0.02

-0.03 -0.03

time(sec) time(sec)

Figure 5.24a Comparison of longitudinal relative velocities

104

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Left F ro n t W h ee l R ight F ro n t W h ee l
1.5 1.5
- 0 N-m 0 N-m
50 N-m - 50 N-m

1 1
0.5 0.5
> >
0 0

-0.5 -0.5
0 1 2 3 4 0 1 2 3 4
time(sec) time(sec)
Left Rear Wheel Right Rear Wheel
0.015 0.015
0 N-m ■0 N-m

0.01 ■A
- 50 N-m
0.01 50 N-m

f 0.005 f 0.005
a 'A fiv /i,
-- 0
>S' 0 —

>
-0.005 J: ’ I *; ,l

-0.005

- 0.01 i
-0.01
0 1 2 3 4 0 1 2 3 4
time(sec) time(sec)

Figure 5.24b Comparison of lateral relative velocities

Therefore, in simulating the two scenarios, the present model seems to be able to

capture the physical behavior of the vehicle. However, it is interesting to find that, though

the vehicle speeds predicted by the present model and that by Villella [5.3] are

indistinguishable, the vehicle paths are visibly different, except for the case of no input

torques (Figure 5.25b). It seems that the difference increases as the input torque increases

(Figure 5.26b). The present model has predicted a greater turning radius. Further

examination of results suggests that the cause may lie in the difference in modeling.

Recalling that in [5.3], an algebraic method for solving for normal force distribution

amongst the four wheel contact points was proposed, where the solution produced the

forces necessary to maintain zero pitch and roll conditions. The method was governed by

105

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 the conditions of static force and moment balance for the vehicle and a hypothetical

suspension whose spring constant was allowed to approach infinity. The present model,

on the other hand, models the wheels’ bounce motion in addition to that of the vehicle

body and considers the dynamic equilibrium of the wheels as well as the vehicle body.

The differences in suspension modeling results in different normal forces distribution

between the two models (Figure 5.27), which in turn give rise to different lateral friction

forces (Figures 5.28 and 5.29), and lead to different turning radii and driven paths. Note

again that in the figures “G-Tech” refers to the approach of [5.3].

V e h ic le L o n g itu d in a l V e lo ctiy
1 i ..............i ............... i ......... r •r i

— P re s e n t M o d e l
‘■ ' S . — - G -Tech

i i i i i i i

0 0 .5 1 1.5 2 2.5 3 3.5 4

V e h ic le L a tera l V e lo c ity
i i 1 1 1

•v.
\
~ y'

------ P re s e n t M o d e l _
------ G -Tech

0 0 .5 1 1.5 2 2.5 3 3 .5 4
A n g u la r V e lo c ity of V e h icle

y '

— P re s e n t M o d e l
------ G -Tech
t 1 t I 1 i i

0 0 .5 1 1.5 2 2.5 3 3.5 4

tim e (sec)

Figure 5.25a Vehicle body speeds, no torque input

106

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Present Model

G-Tech
•2 25

Q .

-45 -40 -35 -30 -25 -20 -15 -10

Y-abs. p o sitio n (m)

Figure 5.25b Vehicle position, no torque input

Vehicle Longitudinal Veloctiy

18

air 17
-x 16 Present M odel
G-Tech
15
0.5 1 1.5 2 2.5 3.5 4
Vehicle Lateral Velocity i________
0.5

_
to 0

-°-5 Present M odel

G-Tech
-1
0 0.5 1 1.5 2 2.5 3 3.5 4
Angular Velocity of Vehicle
0.5

Tt* 0
x5
<o
.% -°-5 — Present M odel
- G-Tech
-1
0 0.5 1 1.5 2 2.5 3 3.5 4
tim e (sec)

Figure 5.26a Vehicle body speeds, 50 N.m input torque

107

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 45

Present Model
40

35

G-Tech
30

.2 25

a.
20

15

10

-40 -30 -20 -10 0

Y-abs. position (m)

Figure 5.26b Vehicle position, 50 N.m input torque

Left Front W heel Right Front W heel
4 000 6000

3000 P r e s e n t M odel 5000

Z G-Tech
<N P re s e n t M odel
z
2 2000 4000 G-Tech

1000 3000

tim e(sec) tim e(sec)

Left R ear W heel Right Rear W heel
3000 5000

2000 P r e s e n t M odel
Z G -Tech
no P re s e n t Mode!
Z 1000 G-Tech

0 1 2 3
time(sec) time(sec)

Figure 5.27a Normal forces for turn maneuver with no torque input

108

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 L e ft F r o n t W h e e l R ig h t F r o n t W h e e l
4 000 6000

i
3000 P re s e n t M odel 5000 : !i
G -Tech
" " n;
CN ;I —
z P r e s e n t Mode!
2 2000 4 000 !i - ■ G-Tech

V -
1000 3000
0 1 2 3 4
tim e(sec) tim e(sec)
Left R ear W heel Right Rear W heel
3000 5000

/ ''
I

2000
A — P re s e n t M odel
G-Tech
4000 - X ------------- 4
\

1000 . 1» 'T
z
3000
. ;j
------ P r e s e n t M odel
- - G-Tech
\ ■'/"

0 2000
0 1 2 3 0 1 2 3 4
tim e(sec) tim e(sec)

Figure 5.27b Normal forces for turn maneuver with 50 N.m torque input

Left Front W heel Right Front W heel

4
P re s e n t M ode
G-Tech
2
z
~ 0
X
LL

-2 J
\ P r e s e n t M odel
G-Tech
-4
0 1 2 3 1 2 3
tim e(sec) tim e(sec)
Left R ear W heel Right R ear W heel
6 4
P re s e n t M odel
4 G-Tech
2
Z 2
CO \ 0
x
£ 0 tv------------
LL

-2 •2
\ ------ P r e s e n t M odel
G -Tech
-4 -4
o 1 2 3 0 2 3 4
tim e(sec) tim e(sec)

Figure 5.28a Longitudinal friction forces, no torque input

109

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 L e ft F r o n t W h e e l R ig h t F r o n t W h e e l
3000 8000

6000
2000
z z 4000
1000
>.
LL £ 2000

P resent Model Present Model

- • G-Tech G-T«ch

-1000 -2000
0
tim e(sec) tim e(sec)
Right Rear W heel
1500 6000

1000 4 000

z
oo 500 2000
>-
LL U_
0
Present Model ' P re se n t Model
- ■ G-Tech G-Tech

-500 -2000
0 1 2 3
tim e(sec) tim e(sec)

Figure 5.28b Lateral friction forces, no torque input

Left Front Wheel Right Front W heel

0 0
P re s e n t Model P re s e n t Model
G-Tech G-Tech
-100 -100

CM

£ -200 £ -200

-300 -300
0 1 2 3 0 1 2 3
tim e(sec) tim e(sec)
Left Rear Wheel Right Rear Wheel
0 0
P re s e n t Model P re s e n t Model
G-Tech G-Tech
-100 -100

CO ^r
,x
£ -200 -200

-300 -300
0 1 2 3 1 2 3
tim e(sec) tim e(sec)

Figure 5.29a Longitudinal friction forces, 50 N.m torque input

110

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 L e ft F r o n t W h e e l R ig h t F ro n t W h e e l
3000

A—-
2000

Present Model
1000 G-Tech

LL
A 2000
0
P resent Model
G-Tech
-1000
o 1 . 2 , 3 0 1 2 3 4
tim e(sec) tim e(sec)
Left Rear Wheel Right Rear Wheel
1500 6000

1000 4000

2
~ 500 2000
5*
LL

P resent Model — P resent Model

G-Tech • - O-Tech
-500 -2000
1 2 3 4
tim e(sec) tim e(sec)

Figure 5.29b Lateral friction forces, 50 N.m torque input

Finally, the present model is applied to an icy road surface simulation. The same

steering input and parameters are used. The frictional parameters are however taken as /us

= 0.2 and [ik = 0.1. The resulting vehicle path on a slippery road surface with zero torque

input is shown in Figure 5.30. The slippery road condition has significantly reduced the

lateral and yaw motion of the vehicle body resulting in a much greater turning radius.

Ill

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Icy Pavement

§ 40

Dry Pav ement

-45 -40 -35 -30 -25 -20 -15 -10

Y -a b s . p o s itio n (m )

Figure 5.30 Comparison of vehicle positions, no torque input

If a front-wheel drive vehicle is considered on the same slippery road with a

100 N-m input torque applied to each of the front wheels so that the total input torque

remains at 200 N-m, Figure 5.31 shows that an even greater turning radius results. As

shown in Figure 5.32 when the same torques are inputted to the rear wheels, the vehicle

spins out of control, as much higher lateral friction forces are exerted onto the front

wheels (see Figure 5.33), producing a very high yaw moment and causing the vehicle to

rotate and spin out of control.

112

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 1 1 ! .... ... !.... i i i i
70 i
1 Front Wheel
• ' | Torque Input
60
\i\ i'?"
'■
No Torque \ >
~ 50 Input \ i
E,
co
:1O 40 y t -
Cl
tn
3 30
><: i -

20
-
10

0 \ i i i i i i i
-25 -20 -15 -10 -5 0 5 10 15 20 25
Y -a b s . p o s itio n (m )

Figure 5.31 Comparison of vehicle positions, slippery road

Q 30

-25 -20 -15 -10

Y -a b s . p o s itio n (m )

Figure 5.32 Vehicle spins out of control, 100 N.m torque to rear wheels

113

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Left Front W heel R ight Front W heel
1000 1UUU

r _ _______
500 500
— - - '1 ' Z
/ i Cl
I \
1 LL
i, n j

\
-500 -5nn
1 2 3 1 2
tim e (s e c ) tim e (s e c )
L e ft R e a r W h e e l R ig h t R e a r W h e e l

\v
/ \
/ \
\\ ■
//
J \\

0 1 2 3 4 0 1 2
tim e (s e c ) tim e (s e c )

Figure 5.33 Lateral frictional forces, 100 N.m torque to rear wheels

5.4 Conclusions.

In this chapter, the mathematical models developed and presented in Chapters 3 and 4

were applied and tested against three previous studies. Three cases were studied, 1) a

seven degrees-of-freedom half-car model of a three-wheeled vehicle, with friction forces

modeled by dry friction (with no-slip only) [5.1] or by the LuGre model as was the case

of the model presented here, 2) a seven degrees-of-freedom full-car model [5.2], with

passive suspension and dry friction (with no slip or with slip), and 3) another seven

degrees-of-freedom full-car model in which the suspension were absent [5.3], but were

presented in the model developed here.

114

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 1) The model proposed here was able to reproduce published work [5.1, 5.2]. See

Figure 5.3 through 5.5 for the three-wheeled vehicle with friction [5.1], and

Figures 5.10 and 5.11 for the full-car model of [5.2], Note that for these two

cases, the proposed model adopts either the LuGre friction model or the dry

friction model.

2) The author’s model was able to capture the essence of the vehicle’s dynamic

behavior, see Figures 5.25 through 5.29. Specifically, vehicle speeds, vehicle

positions (when there is no input torque) and longitudinal frictional forces were

found to be identical to or extremely close to those of [5.3]. However,

differences were observed in vehicle positions (with input torques present) and

in lateral friction forces. Note that the proposed model and the model of [5.3]

both employed the LuGre friction model. They differed in the treatment o f

suspensions, to include them in the proposed model and to neglect them in

[5.3],

3) Therefore, it is suggested that the proposed model is as accurate as the

publications referenced and compared. Discussions in sub-sections 5.3.3.1 and

5.3.3.2 also suggest that the rigorous modeling and mathematical development

115

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 presented in Chapter 3 yields a model that captures the physical essence o f the

vehicle.

4) The model presented here is versatile in that it can be simplified to quarter- and

half-car models, and that it can easily adopt other friction and damper models.

Now that the vehicle model has been developed and verified, a control algorithm

needs to be implemented.

116

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 CHAPTER 6

SLIDING MODE CONTROL

In recent decades, automobiles have continuously been improved with the

implementation of various control techniques. Such techniques serve to optimize the

functionality and safety of the vehicles. M eeting the demand for better handling and ride

comfort has been one of the most intensified research areas in vehicle control.

A great deal of attention has been given to the vehicle suspension system which in

turn influences the ride, handling and maneuverability of a vehicle. The induced

vibrations from the road surface pass through the suspension system before affecting the

body. In the meantime, through the suspension system, the vibration of the vehicle body

influences the tires' dynamic loading and consequently the handling of the vehicle. The

main functions of the suspension system are therefore to provide effective isolation from

road surface unevenness and to improve ride comfort while maintaining a desired level of

road holding (the ability of a car to grip the pavement, as measured by lateral acceleration

in terms of gravitational acceleration g) so as to provide stability and directional control

during handling maneuvers. The control design o f active suspension system has

undergone a major development. Many control laws have been employed for the linear

117

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 and nonlinear models of quarter-, half- and full-vehicle systems [6.1-6.9]. A detailed

comparison of performances of various active and passive suspension systems on quarter-,

half- and full-car models using full-state feedback control can be found in [6.10].

This chapter presents the active suspension control of a half-car (bounce-pitch)

model using the non-switch sliding mode control technique of [6.9]. As a case study, an

half-car model with 4 DOFs will be subjected to excitation from a ramp-step road profile.

The performance o f the active suspension will be evaluated, and compared with that of

the passive suspension. The effectiveness o f the controller for active suspension systems

will be demonstrated. The vehicle body’s bounce and pitch motions will also be

examined in the frequency domains. The robustness of the controller will then be tested

by varying the vehicle’s physical parameters within their possible operating range.

This chapter is a logical extension of the vehicle model developed previously in

Chapters 3 and 4, and verified in Chapter 5. After all, a mathematical model is more

useful if it can be used as a tool towards improvement of the physical system that it

represents.

6.1 Background.

Suspension is required for ride comfort and road holding. Excitation of vehicle vibration

is primarily due to road irregularities. W ith suspension, the vehicle body becomes less

118

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 sensitive to the disturbances generated by the road surface acting on the wheels. In the

early days, suspension system o f most vehicles was purely passive, schematically

represented by dampers and springs (see, for examples, Figure 6.1a). The quality of a

passive suspension depends on suspension parameters. For example, good ride comfort

requires soft springs but this yields poor road holding. An optimal passive suspension

system possesses properly tuned spring and damping coefficients, providing satisfactory

ride comfort and road holding simultaneously. In addition, the resonance frequencies

associated with the sprung (Mb) and unsprung (m) masses remain permanent when

passive suspensions are employed [6.5, 6.9].

Active suspensions, on the other hand, regulate the interaction between the vehicle

body and the wheel by an actuator (see Figure 6.1b). The actuator may be electronically

or hydraulically controlled and applies a force between the vehicle body and the wheel.

This force represents the control action. The advantage of the active suspension over the

conventional suspension is the capability to control the attitude of the vehicle, to reduce

the effects of braking and to reduce the vehicle roll during cornering maneuvers in

addition to increase the ride comfort and vehicle road handling. Thus, it becomes a much

focused research area in vehicle control.

119

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Mb Mb

Controller &
Ks Ks
Actuator
m m

Kt Kt

77777777777777 77777777777777

Figure 6.1a Schematic of a passive Figure 6.1b Schematic of an active

suspension system suspension system

Control design of the active suspension system has also witnessed a major

development. Many control laws have been applied to linear and nonlinear models of

quarter-, half- and full-vehicle systems. The most common type of controller studied has

been the linear quadratic regulators (LQR) with optimal state feedback control [6.1].

Other techniques have also been investigated, including PID controller [6.2], state and

output feedback scheduled controller [6.3], stabilizing controllers [6.4] and fuzzy logic

controllers [6.5]. In the area of robust control, techniques such as the HMoutput feedback

control [6.6], the mixed Ht/Ho, controller [6.7], the modular adaptive robust control

technique [6.8], and the sliding mode controller [6.9] have been investigated to increase

the robustness of suspensions designed for automobiles.

In particular, a non-switch sliding mode control (SMC) method with chattering-free

characteristic was discussed in [6.9], “Chattering” refers to the high frequency switching

of sliding mode controller, and the audible noise associated with it. Reference [6.11]

120

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 presented a survey of chattering problems in SMC systems, and provided several possible

methods for chattering suppression. Reference [6.12] showed the application of higher

order sliding mode control for eliminating chattering. The theory of SMC has been

developed to provide a systematic approach to the problem of maintaining stability and

consistent performance in the face of modeling uncertainties, in which the control

strategies based on SMC schemes are robust against disturbances and parameter

uncertainties [6.14]. Because of the insensitivity features, the SMC theory has been

applied to a wide class of systems, such as applications of the robot manipulators,

spacecraft, and power systems. M ost o f the early work in the area had been proposed by

Utkin [6.13].

6.2 Vehicle Model (A Pitch-Bounce Half-Car).

The equations of motion of a pitch-bounce vehicle model with active suspension are

given below in equations (6.1) through (6.3). The control forces generated by the front

and rear actuators are represented by un (where n - 1,2). Note that such a vehicle model

with passive suspension has been presented in Section 3.5.2.

nin^'ra ksn (z Zn ) T Cs n { z 'En'i Zn) ktn(,Zn — Rn) ~ Un (6.1)

Mb'z 'y ^ [ ksn ( Z %n'l Zn ) T Csn(^Z Zn ) Un ] (6.2)

121

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 ^y,cg'7 y ' [ksn(,Z -UiT Zn')Xn 4" Csn(z %ni ^"n)^n '^n^n ] (6.3)

These equations can be expressed in the state-space form,

£ = f( x ) + [ B] u (6.4)

where, x = [x1,x 2, x 3,x i , x 5, x 6, x 7, x s f = {zl ,z 2, z ^ , z l ,z 2, z ^ f . And f ( x ) is a vector

of functions. Matrix [5] is the controller coefficient matrix having the dimension of 8x2,

and w ii2]r is the control input vector. The system parameters are given in Table

6 . 1.

Table 6.1 System Parameters

Parameter Value Unit

M b 1500 kg
mi 100 kg
m2 200 kg
k.si 28000 N/m
k S2 34000 N/m
k ti 400000 N/m
k t2 400000 N/m
csi 2000 N.s/m
CS2 2000 N.s/m
Xl 1.0 m
X2 -1.2 m
ly.cg 1600 kg.m2

122

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 6.3 Sliding Mode Controller Design.

The theory of SMC can be found mostly in the nonlinear control literature, see, for

example, reference [6.14], The procedure of developing an SMC algorithm includes two

stages [6.15, Section 4.3], the first is to define apre-specified sliding su rfaced , and the

second to develop a control law that will guarantee the attractiveness of the system

trajectory to the surface. Once the system trajectory is confined to the pre-specified

surface, the so-called sliding mode occurs. While in sliding mode, the system is

insensitive to parameter variations and disturbances.

In implementing sliding mode control onto the system given by equation (6.4), the

sliding surface S is chosen to be the error of the system and expressed in terms of the error

state vector e = ( x rej: ( t ) — x ) such that,

S ( x , t ) = [G]e (6.5)

Here x rej ( t ) represents the state vector of the reference, and the constant matrix [G]

represents the slope of the sliding surface.

Under SMC, the system trajectories must stay on the sliding surface (that is S = 0)

for solutions to be stable. To design a control laws u for the close-loop system,

Lyapunov’s second theorem is employed to ensure the system is asymptotically stable. In

123

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 other words, the state trajectories of the controlled system satisfy S = 0 only if the

existence condition of the sliding surface is met. The existence condition is typically

identified as [6.15]

lim S S < 0
5 —>0----- _

To satisfy the condition above, it is required that

S = -D S (6.7)

Note that the derivative of the sliding surface is chosen according to the constant reaching

law [6.12], With D > 0, the solution of equation (6.7) will ensure the sliding surface to

converge to zero.

Typically, the control input vector u consists of a reaching phase, in which the

system moves from its initial position in the state space to the sliding surface, and a

sliding phase, in which it moves along the sliding surface to the desired origin. That is,

W= ueqv + E (6 .8)

According to the conditions given in equations (6.6) and (6.7)

SS = -S D S t < 0
(6.9)
5 = -D S

Employing the derivative of equation (6.5)

(6.10)

124

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 The SMC input u is found to be

M= - ( [ G ] [ 5 ] r 1{ [ G ] ( / - i , , / )} + ([G][j5])-1Z)5 (6.11)

where ([G KB])-1 must exist. The first term is referred to as the equivalent control u cqv,

which is formulated by setting S = 0, and dictates the motion of the state trajectory along

the sliding surface.

Ueqv = - ( [ G ] [ 5 ] ) - 1 { [ G ] ( / - i , , / )} (6.12)

Since when S ^ 0, equation (6.11) is in effect, forcing the system states move from their

initial position in the state space to the sliding surface; and whenG = 0 ,u = ue9„is true, or

the system moves along the sliding surface to the desired position. Thus, the two stages

of SMC are realized.

The uncertainties in system parameters may result in a poor knowledge in / a n d [B] ,

which in turn may cause the calculated equivalent control « to be far off from the actual

equivalent control. Thus, an estimation u eqv was suggested by [6.9] to replace u e q o . That is,

(6,13)

where i is the cutoff frequency. The purpose of using a low pass filter is to bypass the

high frequencies coming from undesirable system oscillations with finite frequency

caused by system imperfections, and to retain the characteristics of the signal. Finally, the

non-switch SMC is given by

125

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 U = + ( [ G ] [ B ] T l D S (6.14)

with ueqv given by equation (6.13).

6.4 Simulations and Discussions.

Simulations have been carried out to demonstrate the effectiveness of SMC for active

suspension systems in comparison to the passive suspension. Road disturbance has been

taken as a single ramp-step bump having a height of 0.02 m between t = 1.0 s and t =

1.20 s (see Figure 6.2). Two such road inputs are applied to the system, with a time delay

St between them. The vehicle is assumed to travel at a constant speed of 10 m/s. The

comparison between passive and active suspensions is made in both the time and

frequency domains.

0.02

0 .0 1 8

0 .0 1 6

_ 0 .0 1 4
E

0 .0 0 6

0 .0 0 4

0.002

8.95 1 1 .0 5 1.1 1 .1 5 1 2
T im e (s e c )

Figure 6.2 A single ramp-step bump

126

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 The control inputu is determined by a l lo w in g ^ = {0}for all reference states.

Parameter r is set to 0.05 ~ 0.1s since the two resonance frequencies pertaining to body

motion and wheel hop (the violent up and down motion of a wheel) are found to be less

than 10 Hz.

[G] and D are non-unique. From the practical point of view, these values are

determined from the design limitations of the suspension system and actuator, such as

maximum working space and actuator saturation. Two sets of control parameters ([G]

and D ) are used in the present study. These parameters are identified through

trial-and-error. The first set is used to demonstrate the effectiveness of SMC regardless of

the physical limitation on the actuator; whereas, the second set takes into account the

physical limitation.

6.4.1 First Set of Control Parameters.

TO 0 10 0 0 0 5 -2~|
[G] = Z) = 50
0 0 0 0 0 0.01 0 0.3

The controlled and uncontrolled (as in the case of passive suspension) vehicle body

bounce and pitch displacements and their accelerations are presented in Figure 6.3. The

vehicle body follows a smooth trajectory against the road irregularities being sensed by

the front and rear wheels as seen in Figures 6.3a and 6.3b (the doted lines). The vehicle

body’s controlled bounce and pitch reach the zero reference value much faster and with

127

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 much less oscillation. The decrease in vehicle body bounce and pitch accelerations give

rise to a more comfortable ride, see Figures 6.3c and 6.3d. The maximum value of the

control forces is around 900 N as seen in Figures 6.4a and 6.4b. This relatively high value,

together with the fact that the control forces, w; and a2, are varying with a frequency of

7 - 8 Ftz, makes it impossible to actually implement the actuator. Otherwise, the forces

are found to be changed reasonably smooth and without any rapid changes known as

chattering which can harm vehicle components.

0.015
— No control
■- Control
0.01

0.005

C
J -0.005

- 0.01

-0.015

- 0.02

tim e (sec)

Figure 6.3a Vehicle body bounce

128

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 0 .0 1
No control
0.008 - - - Control

0.006

0.004

0.002

SI
u
CL
- 0.002

-0.004

-0.006

-0.008

- 0.01

tim e (sec)

Figure 6.3b Vehicle body pitch

2
— No control
- - Control
1.5

0.5

•0.5

1.5

■2
0 1 2 3 4 5
tim e (sec)

Figure 6.3c Acceleration of vehicle body bounce

129

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 2 .5
— No control
2 - - Control

1.5

0.5

•0.5

Q . 1

1.5

-2.5
0 1 2 3 4 5
tim e (sec)

Figure 6.3d Acceleration of vehicle body pitch

800

6 00

400

2 200
3

-200

-400

0 1 2 3 4 5
tim e (sec)

Figure 6.4a Control input force at front suspension

130

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 1000

800

600

400

200
z

-200

-40 0

-60 0

-800

1000

tim e (sec)

Figure 6.4b Control input force at rear suspension

0.16
No control
- - - Control
0.14

0.12

0.1

0.08

0.06

0.04

0.02

0
0 5 10 15 20
frequency (Hz)

Figure 6.5a Frequency spectrum of body bounce acceleration

131

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 0.12
No control
Control

0.06

0.04
0.

0.02

frequency (Hz)

Figure 6.5b Frequency spectrum of body pitch acceleration

The frequency responses of the vehicle with passive suspension (i.e., no control) are

also examined. There are practically two effective resonance frequencies belonging to

body motion and wheel hop. Such frequencies are observed to be 1.1 and 7.5 Hz,

respectively, in the frequency spectra of vehicle body bounce and pitch accelerations

(Figures 6.5a and 6.5b). From both figures, it is seen that, when the controllers are active,

the resonance of pertaining to vehicle body’s bounce motion vanishes, albeit not entirely.

The spectral intensity at wheel hop frequency is significantly reduced, much more so for

the vehicle body’s pitch motion. One can therefore conclude that, by employing the SMC

strategy, only one mode is controlled. In the present study, this mode is the vehicle body

bounce mode.

132

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 The robustness character of SMC is illustrated in Figure 6.6 where physical

parameters of the model, such as the sprung mass Mb, and the spring and damping

coefficients, ksi, ks2, csi and cs2, are varied. A controller is said to be robust if it operates

effectively over all possible operating conditions. The results in Figure 6.6 clearly

demonstrate that the SMC is effective over a wide range of conditions, hence it is robust.

0.06
Mb = 1500kg
Mb = 2000kg

0.05

E 0.04
C
o

| 0.03
0
<_)
CO
s
1 0.02
JoD

0.01

frequency (Hz)

Figure 6.6a Vehicle bounce acceleration - Change in sprung mass

133

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 0.06
- Ks1 = 28000N/m , Ks2 = 32000N /m
■- Ks1 = 14000N/m , Ks2 = 16000N/m

0.05

g 0.04

0.03

<D
0.02
I '

0.01 ) /~ \

\/
10 15 20
frequency (Hz)

Figure 6.6b Vehicle bounce acceleration - Change in spring coefficients

0.06
— Cs1 = C s2 = 2000Ns/m
■ - Cs1 = C s2 = iOOONs/m

0.05

E 0.04

0.03

0.02

0.01

frequency (Hz)

Figure 6.6c Vehicle bounce acceleration - Change in damping coefficients

134

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 6.4.2 Second Set of Control Parameters.

rO 0 100 0 0 0 2 0]
[G] = , D -2
0 0 0 0 0 0 -1 4

This set of parameters is chosen to lower the jerk effect of the control inputs to a feasible

range; As a result, the controller is less effective in isolating vibrations induced from the

road surface. As with first case, this set of parameters was also determined by trial-

and-error.

As seen in Figures 6.7a and 6.7b, the control input forces now experience lower jerk

effect; they are decreased intheir peak values’ magnitudes and are slower acting in the

time domain. For such control inputs, the vehicle body takes longer time to reach the zero

reference for its bounce and pitch motions (Figures 6.8a and 6.8b). The frequency spectra

of the vehicle body’s bounce and pitch accelerations show the reduction of spectral

intensity only at the vehicle body’s bounce frequency.

Therefore, one can conclude that SMC provides an effective means for reducing the

displacement and acceleration of the vehicle body. On the other hand, the ability of using

such control technique in practical vehicles would depend on the development of

actuators that can withstand the large and fast momentum change required of the control

inputs.

135

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 200

150

100

-50

-100

tim e (sec)

Figure 6.7a Control input force at front suspension

200

150

100

/" \

-50

-100

-150

-200

-250

-300

tim e (sec)

Figure 6.7b Control input force at rear suspension

136

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 0.015
No control
Control
0.01

0.005

° -0.005

- 0.01

-0.015

tim e (sec)

Figure 6.8a Vehicle body bounce

0.01
— No control
0.008 - - Control

0.006

0.004

0.002

o
C l
- 0.002

-0.004

-0.006

-0.008

- 0.01

tim e (sec)

Figure 6.8b Vehicle body pitch

137

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 0.16
No control
Control
0.14

0.12

0.1

0.08

0.06

0.04

0.02

0
0 5 10 15 20
frequency (Hz)

Figure 6.9a Frequency spectrum of body bounce acceleration

0.12
No control
- • - Control

0.1

0.08

0.06

£ 0.04
V \

0.02

0
0 5 10 15 20
frequency (Hz)

Figure 6.9b Frequency spectrum of body pitch acceleration

138

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 6.5 Conclusions.

In this chapter, a sliding mode controller for the bounce-pitch model of a vehicle has been

designed and simulation results presented. The results clearly show improvements in ride

comfort. The controller is effective in isolating vibration between the vehicle body and

the irregularities in road surface. The controller is also capable of suppressing one mode

(the resonance at a frequency associated with the vehicle body’s bounce motion). M ost

importantly, it is proven robust. M oreover, one should realize that such control

methodology is possible only when the developed technology can overcome the

limitations on the physical components such as actuators.

139

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 CHAPTER 7

CONCLUSIONS AND RECOMMENDATIONS

The aim of this thesis was to develop a dynamic model of a ground vehicle and to apply

control techniques to the model to optimize the functionality and safety of the vehicle.

The theoretical development of the model is detailed in Chapter 3. The model presented

provides a means to simulate vehicle motions which can lead to appropriate control and

design of vehicles. Main conclusions and some recommendations for future work are

given below.

7.1 Conclusions.

Main conclusions resulting from the study presented are listed below.

7.1.1 Vehicle Dynamic Modeling.

A vehicle model with 14-DOF and with independent suspensions and wheels was

developed. Features of this 14-DOF full-car model were,

1. The vehicle was treated as an assembly of rigid bodies, the sprung and un-sprung

masses, with the following DOFs,

• The sprung mass, or the vehicle body, was allowed three translational and

three rotational motions.

140

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 • The un-sprung masses, or the wheels, were each allowed to have both

bounce and spin motions.

2. The suspensions were modeled such that the wheel bounces were considered to

be independent of the bounce of the vehicle body. Moreover, all bounces o f the

wheels were also considered independent of each other, as were the spins o f the

wheels.

3. Each wheel center was connected to the vehicle body through a “spring-damper”

combination.

4. The tires were considered linear springs with spring constants.

5. The steering of the wheels was not treated as a DOF. However, steering was

incorporated into the model through tire-road interaction, hence becoming “a

state” . The model provided wheel torque at each wheel as inputs and vehicle and

wheel velocities as outputs.

6. The fourteen equations of motion were contained in equations (3.35). Equations

(3.38)-(3.42) founded a set o f five simultaneous equations that were decoupled

to allow the solution to proceed.

7. The full-car model developed could be simplified into various car models with

different DOFs.

141

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 7.1.2 Road-Tire Friction Modeling.

The LuGre friction model was selected to describe the road-tire contact forces due to its

simplicity in model derivation, ease of model parameter identification, and high accuracy

in predicting the frictional behaviors. The dynamic friction model interprets friction as

the interaction of microscopic surface asperities. Existing two-dimensional, single-point-

contact LuGre model was examined, and modification of the model was discussed. The

modification took into account the coupling between the longitudinal and lateral traction

forces, which required consideration of the combined translational and rotational motion

of the wheel. This frictional model was then incorporated into the vehicle dynamic model

to form a complete set of equations of motions for the vehicle system.

7.1.3 Numerical Simulation.

The vehicle model was numerically integrated by using the built-in integration solvers in

Matlab. Three vehicle models were studied, (1) a 7-DOF half-car model of a

three-wheeled vehicle, in which the tire frictional force was modeled either by dry

friction (no-slip only) or by the LuGre model; (2) a 7-DOF full-car model, with passive

suspension and dry damping friction; and (3) a 7-DOF full-car model in which the

suspensions were either absent or present. Some of the findings were,

142

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 1. The model presented was able to reproduce published work for the first two

cases and could adopt either the LuGre friction model or the dry friction model.

2. For the third case, the present model was able to capture the essence o f the

vehicle’s dynamic and contact frictional behaviors under the variations of

applied wheel torque for normal or slick road conditions. However, discrepancy

between the models was found in vehicle positions and in lateral frictional forces

when wheel torques were applied. This may be attributed to how the suspensions

were modeled.

3. Simulation results suggested that the vehicle model presented was as accurate as

the models available in the literature. In addition, it had the versatility in that it

could be simplified to quarter- and half-car models, and could easily be adopted

to other friction and damper models.

4. In general, the rigorous mathematical development of this vehicle dynamics

model provided a model that captured the essence o f the vehicle’s behavior.

143

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 7.1.4 Control Application to the Vehicle Model.

The non-switching sliding mode control technique was implemented in designing the

active suspension control system of a half-car model (a bounce-pitch model). Sliding

mode control was selected based on its character of maintaining stability and consistent

performance in spite of the lack of modeling certainty in a close-loop system. In

particular, a non-switching sliding mode control method with chattering-free

characteristic was considered for this study. The implemented controller was effective in

isolating vibration between the vehicle body and the irregularities in road surface,

capable of suppressing one mode (the resonance at a frequency associated with the

vehicle body’s bounce motion) while reducing significantly the spectral intensity at the

other mode, the wheel bounce motion. M ost importantly, it was proven robust.

7.2 Recommendations.

The results in this thesis lay the groundwork for a better understanding o f automobile

dynamic behavior. There are some natural extensions to this work that can be attempted

in the future. They are outlined in the following, in no particular order.

Additional higher-order dynamic effects may be incorporated by adding or

modifying certain detail at the component level. For instance, the wheel chamber angle

144

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 and the effect of wheel inertia about the z-axis may be included. The center rolling axis

suggested by [3.3] may be introduced to the current model so that the model would be

able to handle a wider range of suspension systems.

The average lumped LuGre friction model as presented in [4.2 - 4.4, 4.6] may be

incorporated into the current model. This version o f LuGre model has been proven to

predict friction behaviors that are closer to reality.

In terms o f applying the dynamic vehicle model, more numerical simulation may be

conducted to demonstrate the capability and accuracy of the current model. It is

suggested to gain access to commercial software such as ADAM S/CAR, CARSim, etc.,

to further validate the model.

Other modeling approaches, for example, the Lagrange’s formulation, may also be

employed to verify the model itself.

The area of control allows for the greatest opportunities for expansion. Owing to the

non-unique nature of the control parameters, they may be fine-tuned by taking into

consideration the spatial limitation of the suspension and the actuator’s dynamic

limitations (available control action) for which active suspension output and control are

constrained [3.12]. Extension of the sliding mode control from a half-car to a full-car

model may be considered. Further, the horizontal, longitudinal, lateral, and yaw motions

145

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 may be controlled by the means of wheel torques. Reference [4.11] has discussed this

using a simpler vehicle model and provided good insights to such control application. It

serves as a good starting point of control by means of wheel torques.

In conclusion, this thesis has presented a number of unique developments that enable

the study of vehicle dynamic behaviors in both modeling and control design. The

continued development and refinement of the work are expected to lead to techniques

that could potentially improve the vehicle design process.

146

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 APPENDIX A

DERIVATION OF VELOCITIES AND ACCELERATIONS

1. Velocity and Acceleration at Center of Gravity of the Vehicle Body CG

Vcg and A cg are measured with respect to the fixed global coordinates X ,Y ,Z .

However, it is convenient to express the velocity and acceleration in terms of vehicle

body coordinates x, y, z such that

Vcg = xi + yj + zk , A cg = xi + yj + zk + xi + yj + zk (A. 1)

where, given f I = cpk , one has

i = f I x i = (pj, j = f i x j = —(pi, k = fl x k = 0 (A.2)

so that A cg becomes

A cg = ( x - y<p)i + (y + x p ) j + ( z ) k (A.3)

2. Velocity and Acceleration at a Wheel Center CWn.

In determining the velocity and acceleration at a wheel center CWn, the kinematics is

that of a point in a moving frame of reference, with the vehicle body being the moving

frame of reference. The wheel center is assumed to only have motion relative to the

moving frame in the vertical direction. The velocity of the wheel center CWn, measured

147

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 with respect to the fixed global frame o f reference but expressed in terms of vehicle body

coordinates x, y, z is

Vcw,n ^eg "f" ^ X Tcw,n 4” ^re/,n (A.4)

where Vrei:fl is the velocity of the wheel center relative to the moving frame, and rCWJl the

position vector of the wheel center in the moving frame. Defining H = h - r , the

absolute vertical distance from vehicle body center of gravity CG to centers of the wheels

in the initial configuration (Figure 3.10), and z w,n = zn + H - z , then

^cw,n Vnj %w<nk (A.5)

where x n and yn are the x- and y-coordinates of the wheel center with respect to the

vehicle body frame. They can be determined by the wheelbase and track width of the

vehicle. Since,

V = xi + yj + zk , Vre[ n = ( in - z)k
(A.6)
ft X rcw,n = ["W n ]* + [ fa n ] j

one obtains, for Vcw,n

y cw,n = [ x - ( p y n ]% + [y + <pXn]j + [(zn — z) + z ]k (A.7)

and for A CWi/l

Aw,n = I* ~ VVn ]? + [X ~ (fVn ] ? + [ij + ] 3 + [V + f a n }3

. (A.8)
+ [ ( 4 - z) + z ]k + [(zn - i) + z ]k

By virtue of (A.2) a n d zwn = i n - z , zw.n = zn - z , one further writes,

148

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 A w .n = { x - t p y - <p2xn - LpyJJ + [y + fix + (pxn - y ? y n ] j
(A.9)
+ [(^ n — ^0 +

Note that equation (A.9) is identical to equation (3.11), hence proving that the approaches

used in Chapter 3 and used above, albeit different points of view in treating the

kinematics, are identical. It is also interesting to note that the small roll and small pitch

assumption has not been invoked in the above derivation. For the derivation in Chapter 3,

though the second half of equation (3.7) assumes small roll and small pitch, equations

(3.8) - (3.11) are valid with and without the small roll small pitch assumption.

149

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 REFERENCES

C h a p te r 1

[1.1] D.A. Crolla, “Vehicle Dynamics - Theory into Practice”, Journal o f Automobile
Engineering, 210, 83-94, 1996.

[1.2] R.S. Sharp, “The Application of Multi-Body Computer Codes to Road Vehicle
Dynamics Modeling Problems”, Journal o f Automobile Engineering, 208, 55-61,
1994.

[1.3] http://www.mscsoftware.com/products/adams applications.cfm?Q=396&Z=397

[1.4] http://www.carsim.com/products/carsim

[1.5] http://www.mathtools.net/MATLAB/Automotive/index.html

[1.6] C. Canudas de Wit, H. Olsson, K J. Astrom and P. Lischinsky, “A New M odel for
Control of Systems with Friction”, IEEE Transaction on Automatic Control, 40(3),
1995.

[1.7] MATLAB Release 13, Vers. 6.5.0, The Math Works Inc., 2002

C h a p te r 2

[2.1] F.H. Speckhart, “A Computer Simulation for Three-Dimensional Vehicle

Dynamics” , SAE Paper 730526, 1973.

[2.2] D.A. Crolla, “Vehicle Dynamics - Theory into Practice”, Journal o f Automobile
Engineering, 210, 83-94, 1996.

[2.3] A. Hac and A.V. Fratini Jr., “Elimination of Limit Cycles due to Signal Estimation
in Semi-Active Suspensions”, SAE, ISSN 0148-7191, 1999.

[2.4] D.E. Simon, “Experimental Evaluation of Semiactive Magnetorheological

Primary Suspensions for Heavy Truck Applications”, M.Sc. Thesis, Virginia
Polytechnic Institute and State University, 1998.

[2.5] H. Chen, Z.Y. Liu and P.Y. Sun, “Application of Constrained Control to Active
Suspension Systems on Half-Car M odels”, J. Dynamic Systems, M easurement
and Control, 127, 345-354, 2005.

[2.6] T.R. Gawade, S. Mukherjee and D. Mohan, “Wheel Lift-off and Ride Com fort of
Three-W heeled Vehicle over Bum p”, IE(I) Journal, 85, 78-87, 2004.

150

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 [2.7] T.R. Gawade, S. Mukherjee and D. Mohan, “Six Degree-of-Freedom Three-
Wheeled Vehicle Model Validation”, Proceedings o f the IM E C H E Part D
Journal o f Automobile Engineering, 12(D4), 487-498, 2005.

[2.8] E. Bakker, H.B. Pacejka and L. Lidner. “A New Tire Model with an Application
in Vehicle Dynamics Studies”, SAE Paper #890087, 1989.

[2.9] M.G. Villella, “Nonlinear Modeling and Control of Automobiles with Dynamic
Wheel-Road Friction and Wheel Torque Inputs”, M aster’s thesis, School of
Electrical and Computer Engineering, Georgia Institute of Technology, 2004.

Chapter 3

[3.1] A. Bedford and W. Fowler, “Engineering Mechanics, Dynamics”, Addison-

Wesley, 286-289, 1994.

[3.2] “Vehicle Dynamics Terminology”, SAE Recommended Practice J-670e, Society of

Automotive Engineers, 1976.

[3.3] A.Y. Ungoren and H. Peng, “Evaluation of Vehicle Dynamics Control for
Rollover Prevention”, University of Michigan, 2003.

[3.4] F.H. Speckhart, “A Computer Simulation for Three-dimensional Vehicle

Dynamics”, SAE paper # 730526, 1973.

[3.5] F.J. Alonso, “Experimental Analysis of Vehicle Dynamics. Validation of A Novel

Vehicle M athematical Model Using Advanced Instrumentation”, Insia-Polytecnic
University of Madrid, 2004.

[3.6] T.D. Day, S.G. Roberts and A.R. York, “SIMON: A New Vehicle Simulation
Model for Vehicle Design and Safety Research”, SAE paper #2001-01-0503, 2001.

[3.7] T.R. Gawade, S. M ukherjee and D. Mohan, “Wheel Lift-off and Ride Comfort of
Three-wheeled Vehicle over Bump”, IE(I) Journal, 85, 78-87, 2004.

[3.8] A. Pytal and J. Kiusalaas, Engineering Mechanics: Dynamics, HarperCollins

Canada, 1994.

[3.9] Maple 9.01, Vers. 9, M aplesoft of Waterloo M aple Inc., 2003

[3.10] R.J. Dorling, “Integrated Control of Road Vehicle Dynamics”, Ph.D. Dissertation,
University of Cambridge, 1996.

[3.11] A. Hac and F.V. Fratini, Jr., “Elimination of Limit Cycle due to Signal Estimation
in Semi-active Suspensions”, SAE paper # 1999-01-0728, 1999.

151

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 [3.13] D.E. Simon, “Experimental Evaluation of Semiactive Magnetorheological
Primary Suspensions for Heavy Truck Applications”, M.Sc. Thesis, Virginia
Polytechnic Institute and State University, 1998.

[3.14] H. Chen, Z.Y. Liu and P.Y. Sun, “Application of Constrained Ha> to Active
Suspension Systems on Half-Car Models”, J. Dynamic Systems, Measurement
and Control, 127, 345-354, 2005.

[3.15] A. Giua, C. Seatzu and G. Usai, “AMixed Suspension System for a Half-car
Vehicle Model”, D ynam ics and Control, 10, 375-397, 2000.

[3.16] N. Yagiz and I. Yuksek, Sliding Mode Control of Active Suspensions for a Full
Vehicle Model, International Journal of Vehicle Design, 26, 264-276, 2001.

[3.17] R. Gviglii, “Active Control of Seat Vibrations of a Vehicle Model Using Various
Suspension Alternatives”, Turkish Journal o f Engineering & Environmental
Sciences, 27, 361-373, 2003.

[3.18] K. Hayakawa, K. Matsumoto, M. Yamashita, Y. Suzuki, K. Fujimori and H.

Kimura, “Robust H^-Output Feedback Control of Decoupled Automobile Active
Suspension Systems”, IEEE Transactions on Automatic Control, 44, 392-396,
1999.

[3.19] L. Zuo and S.A. Nayfeh, “Structured H 2 Optimization of Vehicle Suspensions

Based on Multi-Wheel Models”, Vehicle System Dynamics, 40, 351-371, 2003.

Chapter 4

[4.1] H. Olsson, K.J. Astrom, C. Canudas-de-Wit, M. Gafvert and P. Lischinsky,

“Friction Models and Friction Compensation”, Report, Department of Automatic
Control, Lund Institute of Technology, 1997.

[4.2] E. Velenis, P. Tsiotras and C. Canudas-de-Wit, “Extension of the LuGre Dynamic

Tire Friction Model to 2D Motion”, Report, School of Aerospace Engineering,
Georgia Institute of Technology, 2001.

[4.3] J. Yi, S. Suryanarayanan, A. Howell, R. Horowitz, M. Tomizuka and K. Hedrick,

“Development and Implementation of a Vehicle-Centered Fault Diagnostic and
Management System for the Extended PATH-AHS Architecture: Part II”, Report,
Department of Mechanical Engineering, University of California at Berkeley,
2002 .

[4.4] C. Canudas-de-Wit, P. Tsiotras, E. Velenis, M. Basset and G. Gissinger, “Dynamic

Friction Models for Road/Tire Longitudinal Interaction”, Vehicle System
D ynamics, 39(3), 189-226, 2003.

152

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Friction Models for Road/Tire Longitudinal Interaction”, Vehicle System
Dynamics, 39(3), 189-226, 2003.

[4.5] J. Svendenius and B. Wittenmark, “Review of Wheel Modeling and Friction

Estimation”, Report, Department of Automatic Control, Lund Institute of
Technology, 2003.

[4.6] J. Deur, J. Asgari and D. Hrovat, “A 3D Brush-type Dynamic Tire Friction

M odel”, Vehicle System Dynamics, 42(3), 133-173, 2004.

[4.7] M.G. Villella and D.G.. Taylor, “Input-Output Linearization of an Automobile

Model with 2D LuGre Friction”, Proc. American Control Conference, 429-434,
2005.

[4.8] C. Canudas de Wit, H. Olsson, K.J. Astrom and R Lischinsky, “A New Model for
Control of Systems with Friction”, IEEE Transaction on Automatic Control, 40(3),
1995.

[4.9] R. Stribeck, “Die wesentlichen Eigenschaften der Gleit- und Rollenlager”,

Zeitshrift des Vereines Seutscher Ingenieure, 46(38), 1342-1348, 1902.

[4.10] M.G. Villella, “Nonlinear Modeling and Control of Automobiles with Dynamic
W heel-Road Friction and Wheel Torque Inputs”, M aster’s thesis, School of
Electrical and Computer Engineering, Georgia Institute of Technology, 2004.

[4.11] E. Bakker, H.B. Pacejka and L. Lidner. “A New Tire Model with an Application
in Vehicle Dynamics Studies”, SAE Paper #890087, 1989.

[4.12] E. Bakker, L. Nyborg and H. Pacejka, “Tyre Modelling for Use in Vehicle
Dynamics Studies”, SAE Paper #870421, 1987.

C h ap ter 5

[5.1] T. R. Gawade, S. Mukherjee, D. Mohan, “Wheel Lift-off and Ride Comfort of

Three-W heeled Vehicle over Bump”, IE(I) Journal-MC, 85, 2004.

[5.2] G.. Rahmi, “Active Control of Seat Vibrations of a Vehicle Model Using Various
Suspension Alternatives”, Turkish J. Eng. Environmental Science, 361-373, 2003.

[5.3] M.G. Villella, “N onlinear M odeling and C ontrol of A utom obiles w ith D ynam ic
W heel-Road Friction and Wheel Torque Inputs”, M aster’s thesis, School of
Electrical and Computer Engineering, Georgia Institute of Technology, 2004.

[5.4] J.G. Heydinger, R.A. Bixel, W. R. Garrott, M. Pyne, J.G. Howe and D.A.Guenther,
“Measured vehicle inertial parameters-NHTSA’s data through November 1998”.
[ http://ww w -nrd.nhtsa.dot.gov/vrtc/ca/capubs/sael999-01-1336.pdf]

153

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Chapter 6

[6.1] K. Birgit, R. Georg, S. Oskar von, Z. Douglas E., “Active Suspension Design For
A Tractor By Optimal Control Methods”, Sonderforschungsbereich 438, Munchen,
Augsburg, Technische Universitat Munchen, 1998.

[6.2] G. Rahmi, “Active Control of Seat Vibrations of a Vehicle Model Using Various
Suspension Alternatives”, Turkish J. Eng. Environm ental Science, 361-373, 2003.

[6.3] I. E. Kose, F. Jabbari, “Scheduled controllers for linear systems with bounded
actuators,” A utom atica 39, 1377-1387, 2003.

[6.4] M. C. Smith, F-C Wang, “Controller Parameterization for Disturbance Response

Decoupling: Application to Vehicle Active Suspension Control.” IE E E
Transactions on Control System s Technology, 10(3), 393-407, 2002.

[6.5] L. Sakman, R. Guclu, N. Yagiz, “Fuzzy logic control of vehicle suspensions with
dry friction nonlinearity.” Sadhana, 30, Part 5, 649-659, 2005.

[6.6] K. Hayakawa, K. Matsumoto, M. Yamashita, Y. Suzuki, K. Fujimori, and El.

Kimura, “Robust EU Output Feedback Control of Decoupled Automobile Active
Suspension Systems”, IE E E Transactions on A utom atic Control, 44(2), 392-396,
1999.

[6.7] P. Gaspar, I. Szaszi and J. Bokor, “Mixed H 2 /EL, Control Design for Active
Suspension Structures”, Periodica Polytechnica Ser. Transp. Eng. 28(1/2), 3-16,
2000 .

[6.8] S. Chantranuwathana, and H. Peng, “Adaptive robust force control for vehicle
active suspensions,” P roceedings o f the Am erican Control Conference, 1702-
1706,1999.

[6.9] N. Yagiz and I. Yuksek, “Sliding Mode Control of Active Suspensions for a Full
Vehicle Model”, International Journal o f Vehicle D esign, 26(2/3), 264-276, 2001.

[6.10] J. D. Richard, “Integrated Control of Road Vehicle Dynamics”, PhD. Dissertation,

C am bridge University E ngineering D epartm ent, 1996.

[6.11] J. Guldner and V. I. Utkin, “The chattering problem in sliding mode systems”,
http://www.univ-perp.fr/mtns2000/articles/SU4 4.pdf

[6.12] M. Vig, “Higher Order Sliding Mode Control of Differentially Flat Systems”,
M.Sc. Thesis, Indian Institute o f Tech, 2004.

[6.13] V. I. Utkin, “Variable Structure Systems with Sliding Mode”, IE E E Transactions

on A utom atic Control, AC-22, 212-222, 1977.

154

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 [6.14] J.J.E. Slotine and W. Li, “Applied Nonlinear Control”, Prentice-Hall, Inc. 1991.

[6.15] M. Ahmed, “Sliding Mode Control for Switched Mode Power Supplies”, Ph.D.
Dissertation, Lappeenranta University of Technology, 2004.

155

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.