This article was downloaded by: [McGill University Library]

On: 15 October 2012, At: 00:55

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Vehicle System Dynamics:

International Journal of Vehicle
Mechanics and Mobility
Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/nvsd20

Nonlinear Dynamics of Vehicle

Traction
B.J. Olson, S.W. Shaw & G. Stépán

Version of record first published: 09 Aug 2010.

To cite this article: B.J. Olson, S.W. Shaw & G. Stépán (2003): Nonlinear Dynamics of
Vehicle Traction, Vehicle System Dynamics: International Journal of Vehicle Mechanics and
Mobility, 40:6, 377-399

To link to this article: http://dx.doi.org/10.1076/vesd.40.6.377.17905

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-

conditions

This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan, sub-
licensing, systematic supply, or distribution in any form to anyone is expressly
forbidden.

The publisher does not give any warranty express or implied or make any
representation that the contents will be complete or accurate or up to date. The
accuracy of any instructions, formulae, and drug doses should be independently
verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand, or costs or damages whatsoever or
howsoever caused arising directly or indirectly in connection with or arising out
of the use of this material.
 Vehicle System Dynamics 0042-3114/03/4006-377$16.00
2003, Vol. 40, No. 6, pp. 377–399 # Swets & Zeitlinger

Nonlinear Dynamics of Vehicle Traction

N3

PA
B.J. OLSON1,2, S.W. SHAW2 AND G. STE

SUMMARY

The purpose of this study is to understand the nonlinear dynamics of longitudinal ground vehicle traction.
Downloaded by [McGill University Library] at 00:55 15 October 2012

Specifically, single-wheel models of rubber-tired automobiles under straight-ahead braking and acceler-
ation conditions are investigated in detail. Customarily, the forward vehicle speed and the rotational rate of
the tire/wheel are taken as dynamic states. This paper motivates an alternative formulation in which wheel
slip, a dimensionless measure of the difference between the vehicle speed and the circumferential speed of
the tire relative to the wheel center, replaces the angular velocity of the tire/wheel as a dynamic state. This
formulation offers new insight into the dynamic behavior of vehicle traction. The unique features of the
modeling approach allow one to capture the full range of dynamic responses of the single-wheel traction
models in a relatively simple geometric manner. The models developed here may also be useful for
developing and implementing anti-lock brake and traction control control schemes.

1. INTRODUCTION

In studies of vehicle traction the gross vehicle dynamics and tire/wheel dynamics can
be captured by lumped mass models. Simplified models that are often considered for
longitudinal braking and acceleration include the single-wheel model [1–3], and a
two-dimensional, two-wheel model (front and rear) [1, 2], or full four-wheel models
for cornering [1, 2]. The dynamics of these systems involve interactions between the
vehicle, the tire/wheel assemblies, and the road surface. The force that ultimately
slows or accelerates the vehicle is the longitudinal friction force between the road and
tire, which can be empirically described in terms of a slip condition at the interface.
Thus, writing the equations of motion for any rubber-tire vehicle system requires a
description of the friction force generated at the tire/road interface, in addition to the
usual laws of motion.
Experimental evidence shows that the longitudinal friction force is proportional to
the normal force at the contact [1–3], with a coefficient of friction serving as the

1
Address correspondence to: B.J. Olson, Department of Mechanical Engineering, Michigan State
University, East Lansing, MI 48824, USA. E-mail: [email protected]
2
Department of Mechanical Engineering, Michigan State University, East Lansing, MI 48824, USA.
3
Department of Applied Mechanics, Budapest University of Technology and Economics, Budapest,
Hungary.
 378 B.J. OLSON ET AL.

‘constant’ of proportionality. This coefficient can be conveniently modeled in an

empirical manner that depends on the slip [4, 5], which is a dimensionless measure of
the difference between the vehicle speed and the circumferential speed of the tire
relative to the wheel center. During braking (resp. acceleration), this difference is
generated by a brake (resp. engine) torque on the wheel, which acts against (resp.
with) the inertia of the vehicle. The slip depends on the dynamics of the vehicle and
the tire/wheel, and it also influences their dynamics through the friction force. This
‘feedback’ results in a system of coupled equations of motion for the vehicle and the
tire/wheel. These equations of motion are most often formulated in terms of the
vehicle’s speed relative to ground and the absolute rotational rate of the tire/wheel.
Downloaded by [McGill University Library] at 00:55 15 October 2012

This is a very natural formulation, wherein the slip is merely an internal variable
defined in terms of the system’s dynamic states, which is used to compute the friction
force that appears in the equations of motion.
In this paper, a formulation is considered in which the slip is taken to be a dynamic
state variable, replacing the absolute rotational rate of the tire/wheel. Liu and Sun [6]
have developed the equations of motion for a quarter-car model in this way, but their
investigation focuses on control algorithms based on gain-scheduling, rather than
general dynamic behavior. The formulations herein focus not on control, but the
general dynamic characteristics of vehicle traction. Only single-wheel braking and
acceleration models are developed here; two-wheel models are considered in a
companion paper [7]. It will be shown that the equations of motion take on forms that
lend themselves to relatively simple investigation and interpretation using tools from
nonlinear dynamics. Specifically, this formulation allows the dynamics for the entire
range of vehicle speeds and slip values to be captured by a scalar function (one for
each model) that is defined in terms of the slip, the brake or engine torque, and the
friction/slip relationship. These functions describe completely the behavior of a given
vehicle during braking (resp. acceleration) and under a constant or slowly varying
brake (resp. engine) torque.
The dynamic models presented here are capable of describing both transient and
steady tractive performance. Most importantly, they clearly demonstrate how a
vehicle can undergo stable braking or acceleration and/or lockup, depending on the
brake/engine torque, the friction characteristic, and the vehicle parameters. These
models also allow one to clearly see how transitions between operating states occur as
parameters and conditions are varied. In fact, a single set of phase plane diagrams,
drawn at varying brake/engine torque levels, completely captures the entire range of
possible behaviors for a given friction characteristic.
One of the more interesting findings of this study is a stability result related to
brake lockup. The standard assumption is that the brake torque can increase until the
slip reaches a value that corresponds to the maximum coefficient of friction, beyond
which lockup occurs [1, 2]. Under steady-state braking conditions the corresponding
maximum brake torque is typically assumed to be equal to the peak moment provided
 NONLINEAR DYNAMICS OF VEHICLE TRACTION 379

by the friction force about the wheel center. In this work it is shown that this result is
actually an approximation that is only accurate when the inertia of the tire/wheel is
very small compared to that of the vehicle. Since this is generally the case, the
approximation works well, but the present results determine where instability to
lockup actually occurs, and also systematically shows how the approximation is
obtained.
It is recognized that slip is notoriously difficult to measure in practice, primarily
since it is not a simple matter to determine the vehicle speed relative to ground. (The
speedometer uses the tire circumferential speed, which does not match the vehicle
speed during slip – and this is precisely when both speeds are required to compute the
Downloaded by [McGill University Library] at 00:55 15 October 2012

slip [8]). However, since methods exist for estimating the slip [9–11], models such as
those developed here may be useful for developing and implementing anti-lock brake
systems (ABS) and traction control systems (TCS).
The paper is organized as follows. The single-wheel braking model is developed
first in a systematic manner, thus laying the necessary groundwork for the
development and analysis of a single-wheel acceleration model. The equations of
motion are presented and the need to quantify the available friction force for braking
is specified. This motivates an investigation of the tire/road interface and leads to the
introduction of force coefficient characteristics as a function of longitudinal wheel
slip. The equations of motion are hence cast into a framework that is convenient for a
nonlinear dynamic analysis. Two formulations are considered: one in which the
dynamic states are taken to be the forward vehicle speed and the angular speed of the
tire/wheel, and one in which wheel slip replaces the angular speed as a dynamic state.
The latter formulation is pursued in detail. Global features of the braking model are
discussed, including steady-slip conditions, local stability of slip dynamics,
hysteresis, and the transition to unstable braking (lockup). Finally, a single-wheel
acceleration model is similarly developed and analyzed. The paper closes with some
conclusions and directions for future work.

2. THE SINGLE-WHEEL BRAKING MODEL

A quarter-car model is developed in this section in order to illustrate the fundamental

aspects of vehicle braking. This single-wheel model is unrealistic by virtue of its
simplicity, and it clearly fails to capture some important dynamical features (e.g.,
dynamic load transfer). It nevertheless serves to facilitate an understanding of the
basic dynamic characteristics of vehicle braking. In fact, the approach taken here lays
the necessary groundwork for the formulation and analysis of two- and four-wheel
models. Two-wheel traction models are considered in a companion paper.
As depicted in Figure 1, the quarter-car model consists of a single wheel
constrained to move longitudinally in the x-direction at a speed u and with a rotational
 380 B.J. OLSON ET AL.

Fig. 1. Schematic of the single-wheel braking model and corresponding free body diagram.
Downloaded by [McGill University Library] at 00:55 15 October 2012

rate !. Denoted by R and J are its effective rolling radius and polar moment of inertia,
respectively. The effect of a braking mechanism on the vehicle wheel is captured by
the brake torque Tb, which opposes the forward motion of the system. The vertical
reaction force Z balances the static weight mg, while the longitudinal force X serves to
slow the vehicle in braking. By summing forces in the x- and z-directions and
moments about the mass center C of the vehicle/wheel, the system equations are
found to be:
mu_ ¼ X; ð1Þ

Z ¼ mg; ð2Þ

J !_ ¼ RX  Tb ; ð3Þ
where m is the mass of the vehicle-wheel combination and g is the acceleration due to
gravity. Overdots denote differentiation with respect to time.
In general, there are a number of forces acting on a vehicle that may give rise to a
deceleration. The model considered here includes only the longitudinal brake force X,
which is discussed in detail in the next section. Other sources of deceleration in
braking include driveline drag, grade, rolling resistance, and aerodynamic drag.

2.1. Tractive Properties

The primary force of interest in studies of vehicle traction is the longitudinal force X,
which acts on the vehicle through a tire/road contact patch. Experimental evidence
shows that this friction force is proportional to the normal force Z at the contact and is
a consequence of the relative difference between the vehicle speed u and the rolling
speed of the tire !R. The ‘constant’ of proportionality is responsible for the friction
coupling, and can be empirically determined by a friction characteristic in terms of
road test data and wheel slip, which is a dimensionless measure of the difference
between u and !R. Since the friction characteristic captures the typifying quantities of
 NONLINEAR DYNAMICS OF VEHICLE TRACTION 381

a particular tire/road combination – including slip stiffness at zero slip and peak brake
force values – it can be regarded as a tire model that characterizes the tire behavior on
a given road surface. The tractive properties are now discussed in terms of wheel slip,
the tire/road interface and attendant friction law, and a tire model.

2.1.1. Wheel Slip

The longitudinal friction force X is a consequence of the relative difference between
the vehicle speed u and the rolling speed of the tire, which is given by !R. Wheel slip
is defined in terms of this difference as1
Downloaded by [McGill University Library] at 00:55 15 October 2012

u  !R
s : ð4Þ
maxðu; !RÞ

It is assumed and taken as convention that u > 0 and 0  !R  u in vehicle braking.

Thus, s ¼ u!R
u is defined on the unit interval I ¼ [0,1], taking on the limiting values of
s ¼ 0 for free rolling (u ¼ !R) and s ¼ 1 for wheel lockup (!R ¼ 0). The former case
u ¼ !R implies the absence of a brake torque. The definition of slip, along with the
convention that !R  u allows for two possibilities for steady-state vehicle braking
with nonzero initial speed: (1) finite rotation of the wheel while the vehicle
decelerates; and (2) deceleration under lockup conditions. It is noted that the case of
lockup is undesirable since steerability, directional stability, and general control over
a vehicle is severely degraded in such a state [1, 2].

2.1.2. The Tire/Road Interface and Friction Law

In a rubber tire, wheel slip results in the deformation and sliding of tread elements in
the tire/road contact patch, which in turn sustains the friction force X in braking.
Indeed, it is through this important interface between the road surface and tire tread
that braking is negotiated. In general, the microscopic physical description of the said
phenomenon is complicated and involves more physics than what are needed here.
(See, for instance, [1, 2].) It suffices to capture these interactions by the simple
algebraic relationship:

X ¼ ðsÞZ; ð5Þ

which is known as the friction law or creep force equation. The longitudinal force
coefficient  : I ! I is experimentally determined in terms of road test data and is the
subject of the next section.

1
The maximum function max(u, !R) allows the use of Equation (4) to define longitudinal wheel slip for both
vehicle braking and acceleration. In braking u < !R, while u > !R for vehicle acceleration.
 382 B.J. OLSON ET AL.

2.1.3. Friction Characteristic and Tire Model

The friction coupling between a rubber tire and road surface depends on a number of
physical parameters involving tire construction, inflation and wear, the tire/road
interface, and vehicle speed and loading [1–3]. Since a general theory that can
accurately predict the longitudinal brake force in terms of wheel slip has yet to be
developed, friction coupling is necessarily determined experimentally. Various
methods exist to relate the brake and normal forces X and Z in terms of a friction
characteristic (s). See, for example, [9, 10]. The resulting data can then be
represented by a formula.
Figure 2 shows graphical representations of some typical longitudinal friction
Downloaded by [McGill University Library] at 00:55 15 October 2012

characteristics. The initial rate at which (s) increases with increasing slip is
dependent on the properties of the tire. For wet and dry asphalt the characteristics
increase until a peak value p ¼ (sp) is attained. This typically occurs between 10
and 20% slip, yielding maximum braking forces of 25–50% and 70–90% of the
vertical load for wet and dry asphalt, respectively. The friction characteristics then
exhibit a gradual decrease to s ¼ 1 (wheel lockup). For gravel and packed snow, the
behavior of the friction coefficient characteristics are qualitatively different. Peak
values occur at wheel lockup (here, s ¼ sp ¼ 1) and are the consequence of plowing
conditions on the deformable surfaces. Ice characteristics (not shown) are similar to
those for wet and dry asphalt, differing mostly in the resulting peak values [1–3].
An analytical treatment of these friction characteristics is possible by employing
the widely used Pacejka tire model [4, 5]. It is described by the so-called Magic
Formula which is given by:
yðxÞ ¼ D sin ðC arctanðBx  EðBx  arctanðBxÞÞÞÞ; ð6Þ
where the parameters B, C, D, and E are the stiffness, shape, peak, and curvature
factors. See [4] for typical values of these coefficients. Horizontal and vertical shifts

Fig. 2. Typical longitudinal friction characteristics: (a) dry asphalt; (b) wet asphalt; (c) gravel; and (d)
packed snow.
 NONLINEAR DYNAMICS OF VEHICLE TRACTION 383
Downloaded by [McGill University Library] at 00:55 15 October 2012

Fig. 3. Friction characteristics employed in numerical simulations.

of a characteristic are attained by the transformations:

YðÞ ¼ yðxÞ þ Sv ; x ¼  þ Sh ;
where Sv is the vertical shift and Sh is the horizontal shift. The function Y() can
represent all steady-state tire characteristics – including the brake force X, side force,
and self-aligning torque – in a physically meaningful and straightforward way. The
variable  denotes either slip angle (the angle subtended from the direction of wheel
travel to the direction of wheel heading) or longitudinal wheel slip s. In light of
Equations (2) and (5), note that the Magic Formula can be scaled to represent (s)
directly on I.
For the purpose of more efficient numerical simulations, a simple friction
characteristic was devised for the present study and is given by:
ðsÞ ¼ c1 ð1  ec2 s Þ  c3 s; s 2 I: ð7Þ
This function is unimodal with a peak value of p ¼ 0.972 at sp ¼ 0.316 and hence
behaves similarly to wet and dry asphalt characteristics. Equation (7), which is shown
in Figure 3 for c1 ¼ 1.18, c2 ¼ 10.0, and c3 ¼ 0.5, was employed for all calculations
and numerical simulations involving (s) for the vehicle braking model. The
characteristic shown for s 2 [1, 0] is employed for vehicle acceleration in Section 3.

2.2. Equations of Motion

During wheel slip, the single-wheel model possesses two dynamic states and hence
requires a set of two coordinates to describe its motion. By inspection of Equations
(1)–(5), it is clear that two of three possible variables of interest could be chosen as the
independent variables, namely, u, !, or s. The equations of motion for the single-wheel
braking model are developed first in terms of u and ! as dynamic states, and the
qualitative dynamics are captured in the (u, !R) phase plane. Though such a description
 384 B.J. OLSON ET AL.

is physically illuminating, it is subsequently shown in Section 2.2 that a formulation of

the equations of motion in terms of u and s as dynamics states lends itself to a relatively
simple interpretation. Specifically, it will be shown that the latter formulation allows
the dynamics for the entire range of vehicle speeds and slip values to be captured by a
single function that is defined in terms of wheel slip and the brake torque.

2.2.1. u and ! as Dynamic States

The common formulation of the equations of motion makes use of the speed u of the
vehicle relative to ground and the absolute rotational rate ! of the tire/wheel as
dynamic states. Then the system dynamic equations are:
Downloaded by [McGill University Library] at 00:55 15 October 2012


u_ ¼ ðu; !Þg
; 0  !R  u; ð8Þ
!_ R ¼ gHðu; !Þ
where wheel slip is merely an internal variable. The restriction 0  !R  u ensures
that s 2 I, according to the convention of wheel slip in braking. The function:
Hðu; !Þ ¼ ðu; !Þ  b ð9Þ
2
is dimensionless, where  ¼ mRJ is the dimensionless ratio of vehicle to wheel inertia
R
(typically 
1), and b ¼ Jg Tb is the dimensionless brake torque.

2.2.2. The ðu; !RÞ Phase Plane

Figure 4a shows trajectories in the (u, !R) state space for  ¼ 15 and b ¼ 12. The
rolling speed !R of the tire is defined along the ordinate while the vehicle speed u is

Fig. 4. State space descriptions for the single-wheel braking model for  ¼ 15 and b ¼ 12: (a) in u and
!R; (b) in u and s.
 NONLINEAR DYNAMICS OF VEHICLE TRACTION 385

defined along the abscissa. Wheel slip is implicitly defined in terms of these states by:
!R ¼ ð1  sÞu; ð10Þ
which follows from the definition given by Equation (4). Equation (10) shows that
radial lines originating from (u, !R) ¼ (0,0) are lines of constant slip for which there is
a linear relationship between u and !R. Since s is defined on the unit interval for
vehicle braking, the dynamics need only be considered in the region:
F ¼ fðu; !RÞju  0; 0  !R  ug: ð11Þ
Thus, trajectories are bounded by the line !R ¼ u, which corresponds to s ¼ 0 (free
Downloaded by [McGill University Library] at 00:55 15 October 2012

rolling), and the line !R ¼ 0, or the u-axis, which corresponds to s ¼ 1 (wheel lockup).
For a particular brake torque, certain constant-slip radial lines are invariant under the
dynamics. The corresponding constant slip values shall be denoted by s . Any such set
that satisfies these conditions and the equations of motion define invariant linear
manifolds in the (u, !R) phase plane, which are denoted by:
W b ¼ fðu; !RÞj!R ¼ ð1  s Þu; s 2 Ig: ð12Þ

Under certain conditions (to be determined subsequently), the u axis is also an

invariant manifold (where s ¼ 1) and is denoted by:
W Lb ¼ fðu; !RÞju  0; !R ¼ 0g; ð13Þ
where the superscript L indicates lockup. There are, for example, three such invariant
manifolds when b ¼ 12 (see Fig. 4a) with one corresponding to lockup conditions,
that is, W Lb . All trajectories started in F rapidly approach either W b or W Lb as u
decreases (since u_ < 0) and evolve essentially along the invariant manifold toward
zero speed at ðu; !RÞ ¼ ð0; 0Þ. Trajectories started on, or that approach W b , represent
stable braking, while trajectories that approach W Lb correspond to unstable braking, or
lockup conditions. These invariant manifolds, if they exist, serve two purposes: they
(1) define steady-slip conditions that are invariant under the dynamics and under
which the vehicle decelerates to zero speed; and (2) separate regions of stable and
unstable braking. It would be desirable to quantify these manifolds without having to
perform numerical simulations for several initial conditions.
In what follows, an alternative formulation is considered where wheel slip s
replaces !R as a dynamic state. Figure 4 compares the state space description of the
single-wheel braking model in the (u, !R) phase plane to its description in the (u, s)
phase plane for  ¼ 15 and b ¼ 12. The alternative formulation yields a state space
where, essentially, the point (u, !R) ¼ (0,0) is expanded to represent wheel slip on the
unit interval at u ¼ 0. In doing so, a singularity is introduced at u ¼ 0; but, as indicated
in Figure 4b, the invariant manifolds W b and W Lb are easily identified in the (u, s)
phase space as lines of constant s. It will be shown that a formulation of the equations
 386 B.J. OLSON ET AL.

of motion in terms of u and s allows for the invariant manifolds, and hence steady-slip
conditions and various operating regimes, to be captured by a single function that is
defined in terms of wheel slip and the brake torque.

2.2.3. u and s as Dynamic States

Although it is very natural to cast the equations of motion in terms of the forward
vehicle speed u and the tire/wheel rate of rotation, it is instructive to replace ! with
wheel slip as a state variable. Liu and Sun [6] have developed the equations of motion
for a quarter-car model using u and s as dynamic states, but their investigation focuses
on control algorithms based on gain-scheduling. Here the equations of motion are
Downloaded by [McGill University Library] at 00:55 15 October 2012

developed similarly but with emphasis on a form suitable for a nonlinear dynamic
analysis. Evaluating the time rate of change of wheel slip (for !R  u):
R
s_ ¼ ð!u_  u!_ Þ;
u2
and performing the appropriate substitutions, the equations of motion may be cast in
terms of u and s. The result is:

u_ ¼ ðsÞg
; u > 0; s 2 I: ð14Þ
s_ ¼ gu hb ðsÞ

Since u > 0 by convention, g > 0, and (s) 2 I, it follows that u_ < 0, which is expected.
The function:
hb ðsÞ ¼ ðs  1  ÞðsÞ þ b ð15Þ
2
is nondimensional, where, recall,  ¼ mRJ is the ratio of vehicle inertia to wheel inertia
R
and b ¼ Jg Tb is the dimensionless brake torque. Note that the brake torque term is
additive, so that as b is varied, the shape of hb ðsÞ does not change. It simply shifts up
and down. Also, since typically 
1 (discussed subsequently), the shape of hb ðsÞ is
close to that of ðsÞ.
The general features of the quarter-car model are best demonstrated by treating
Equation (14) as a state-space representation of the single-wheel system and exploring
its behavior in the (u, s) state space. It will be shown that this interpretation of the
single-wheel model yields good insight into its dynamic response during transient and
steady-state braking. The analysis begins by determining the steady-slip conditions
and their local stability characteristics. A more detailed mathematical analysis follows
in a discussion of the global features of the single-wheel braking model.

2.3. Steady-Slip Conditions

Equation (14) shows that, for nonzero u and a slip value s for which hb(s ) ¼ 0, the
time rate of change of slip is identically equal to zero. Correspondingly, wheel slip
 NONLINEAR DYNAMICS OF VEHICLE TRACTION 387

remains constant at s ¼ s , independent of the vehicle speed. This in turn ensures that
the vehicle acceleration u_ ¼ ðs Þg is negative and constant. Here, ðs Þ is the
longitudinal force coefficient corresponding to the fixed slip value. Under these
conditions the vehicle speed monotonically decreases to zero according to the
equation:
uðtÞ ¼ uo  ðs Þgt; u > 0; 0  t  tf ; ð16Þ
where uo > 0 is the initial speed at the instant when s ¼ s , that is, when t ¼ 0, and tf
corresponds to the time when u ¼ 0. Wheel lockup also yields steady-slip conditions
when s ¼ 1. Under lockup conditions, the dynamics of the vehicle are described by
Equation (16), with the coefficient of sliding friction L ¼ ðs ¼ 1Þ replacing ðs Þ.
Downloaded by [McGill University Library] at 00:55 15 October 2012

2.4. Local Stability of Slip Dynamics

Before specifying a quantitative measure of stability, it is convenient to outline and
adopt specific notation. First, recall that constant slip values s denote invariant points
in the slip dynamics. They may be obtained by finding the zeros of hb(s), that is, by
finding the roots of hb(s ) ¼ 0. More precisely, the steady-slip values s , if they exist,
define invariant sets of the system, since once s ¼ s is attained, s remains at that value
for all time, independent of the values of u (for u > 0). Any such value of s ¼ s may
be either stable or unstable and shall be denoted by sþ and s, respectively. Local
stability criteria of wheel slip follows from considering a small perturbation
ðtÞ ¼ sðtÞ  s away from one of these roots. Differentiating with respect to time,
invoking Equation (14), and employing hb ðs Þ ¼ 0, the local slip dynamics near s
can be approximated to leading order by the linearized equation:
g
_ ¼ h0b ðs Þ ; ð17Þ
u
where ðÞ0 ¼ dðÞ g
ds denotes differentiation with respect to s. Since u > 0, Equation (17)
shows that the perturbation grows exponentially fast when h0b ðs Þ > 0 and decays
exponentially when h0b ðs Þ < 0. Thus, the stability of the slip dynamics near s ¼ s are
determined by the slope of hb ðsÞ at s ¼ s :
h0b ðs Þ ¼ 0 ðs Þðs  1  Þ þ ðs Þ: ð18Þ
Stable and unstable steady-slip values are defined to be:

s  ¼ sjhb ðs  Þ ¼ 0; h0b ðs  Þ ? 0 : ð19Þ
The corresponding stable and unstable invariant manifolds of the system in the (u, s)
plane are defined by:
W b ¼ fðu; sÞju > 0; s ¼ s  g: ð20Þ
 388 B.J. OLSON ET AL.

For lockup there is a stable manifold:

W Lb ¼ fðu; sÞju > 0; s ¼ 1g;

defined for parameter conditions under which hb ð1Þ  0. The notation W b shall refer
to either of the invariant manifolds W þ  L
b or W b (W b not included).

2.5. Global Features of the Single-Wheel Braking Model

Equations (17) and (19) hint at the importance of hb ðsÞ, since steady-slip conditions
Downloaded by [McGill University Library] at 00:55 15 October 2012

and the local stability of the slip dynamics are completely determined in terms
of this dimensionless function. In fact, the entire range of vehicle speeds and slip
values are captured by hb ðsÞ under a constant brake torque or slowly varying brake
torque. This is shown in Figure 5, where the function hb ðsÞ versus slip and the
corresponding state space dynamics in u and s are depicted for  ¼ 15 and for various
values of the nondimensional brake torque. The intersection of the function hb ðsÞ
with the line hb ¼ 0 defines the invariant points s [see Equation (19) and Table 1],
and hence the invariant manifolds W b in the (u, s) space, which are defined by
Equation (20).
For b ¼ 7, a small brake torque, only one invariant point exists, which is shown
in Figure 5a as sþ. Since h0b ðs Þ < 0 the steady-slip value sþ is stable and hence
defines the invariant manifold W þ b . All initial conditions ðu; sÞ ¼ fðu; sÞju > 0; s 2 Ig
result in stable braking at this parameter value, hence it is globally stable. This can be
seen by noting that for s  s, s_ ¼ gu hb ðsÞ > 0 so that s increases while for s > s,
s_ ¼ gu hb ðsÞ < 0 and hence s is decreasing. Therefore all initial conditions lead towards
s ¼ sþ .
By increasing the brake torque such that hb ð1Þ ¼ 0, another fixed point is
introduced at s ¼ 1 for which the slope of h0b ðs Þ is negative; hence it is unstable. It is
denoted by s ¼ s and defines the unstable invariant manifold W  b in the (u, s) space.
Note that the creation of the unstable invariant point s corresponds to the
introduction of lockup at s ¼ 1, which defines W Lb . This is essentially a saddle-node
bifurcation creating the steady-slip values s and s ¼ 1. Figure 5b shows this
situation, where W  L L
b and W b are coincident at the brake torque level b ¼ 10:199.
 L
All trajectories lying below W b ¼ W b, that is, in fðu; sÞju > 0; s 2 ½0; 1Þg, are
attracted to the invariant manifold W þ b and result in stable braking conditions. At this
specific parameter value lockup ðs ¼ 1Þ is unstable.
1

Fig. 5. hb ðsÞ versus s and corresponding state space descriptions in u and s for  ¼ 15: (a) b ¼ 7;
(b) b ¼ Lb ¼ 10:199; (c) b ¼ 12; (d) b ¼ cr
b ¼ 15:250; (e) b ¼ 18. See Table 1 for steady
slip values.
Downloaded by [McGill University Library] at 00:55 15 October 2012

NONLINEAR DYNAMICS OF VEHICLE TRACTION

389
 390 B.J. OLSON ET AL.

Table 1. Steady-slip values for the single-wheel braking model.

b s scr sþ

7 – – 0.050
12 0.782 – 0.117
Lb ¼ 10:199 1 – –
cr
b ¼ 15:250 – 0.304 –
18 – – –

As the brake torque is further increased (shown in Fig. 5c for b ¼ 12) s moves
Downloaded by [McGill University Library] at 00:55 15 October 2012

away from s ¼ 1, which stabilizes W Lb and creates a lockup region. All trajectories
above W  
b , that is, those with initial conditions in fðu; sÞju > 0; s 2 ðs ; 1g, tend to
L 
W b (indicating wheel lockup). Trajectories lying below W b are attracted to the
invariant manifold W þ b and result in stable braking conditions.
Further increasing b causes sþ and s to move toward each other. Eventually a
critical brake torque cr b ¼ 15:250 is reached where the stable and unstable invariant
points collide and mutually annihilate in a saddle node bifurcation at the critical slip
value s ¼ scr ¼ 0:304. (See Fig. 5d.) For brake torques greater than cr b the invariant
manifold W Lb is globally stable and wheel lockup occurs for all initial slip conditions.
This situation is shown in Figure 5e where b ¼ 18.
The set of initial conditions for which a trajectory reaches the invariant manifold
W Lb is called the domain of attraction of s ¼ 1 in the (u, s) state space and is denoted
by:
L ¼ fðu; sÞju > 0; s 2 ½s ; 1Þg:

Trajectories started with initial conditions in L (denoted by the shaded regions in

Fig. 5) tend rapidly toward lockup at s ¼ 1, subsequently move along W Lb , and
monotonically approach the point ð0; 1Þ according to Equation (16). This situation
corresponds to braking under lockup conditions which, for wet and dry asphalt
characteristics, is always non-optimum in terms of deceleration since
ðs ¼ 1Þ ¼ L < ðsp Þ. The domain of attraction of sþ, that is, the set of initial
conditions for which a trajectory reaches the stable invariant manifold W Lb , is given by:
S ¼ fðu; sÞju > 0; s 2 I n Lg:

Trajectories started with initial conditions in S rapidly converge toward the stable
invariant manifold and evolve essentially along W þ b , according to Equation (16)
toward the point ðu; sÞ ¼ ð0; sþ Þ, where the vehicle has stopped. The rate at which the
vehicle decelerates under steady-slip depends only on ðs Þ, that is, the particular
friction characteristic and the value of s . Finally, trajectories started on the unstable
invariant manifold remain on W  b and monotonically evolve toward the point ð0; s Þ.

 NONLINEAR DYNAMICS OF VEHICLE TRACTION 391

This situation is physically unattainable, however, since any small perturbation in the
system would cause a trajectory to leave W  b and move into either S or L.
Note that, since s_  1u, the rate at which a trajectory tends toward either W Lb or W þ
b
increases dramatically as u tends toward zero. In fact, from Equation (14), the time rate
of change of wheel slip becomes infinite as u ! 0 with s 6¼ 1; s . Hence, the vehicle
must come to rest under steady-slip conditions for which s ¼ 1 or s ¼ s. There are
only two such physical possibilities: the vehicle decelerates to zero speed (1) at the rate
L g with the wheels locked; or (2) with steady-slip at the absolute rate ðsþ Þg. Peak
steady-braking performance would entail steady slip at s ¼ sp for which the maximum
deceleration is obtained and is equal in magnitude to p g. However, is will be shown in
Downloaded by [McGill University Library] at 00:55 15 October 2012

Section 2.7 that the transition to lockup occurs before s can reach sp . Thus, since sp
cannot be reached under stable braking conditions, optimum braking would entail
steady-slip at s ¼ scr . The corresponding deceleration is equal in magnitude to
cr g ¼ ðs ¼ scr Þg < p g. But since scr is a saddle node in the slip dynamics, any
perturbation in the system could send the braking conditions into wheel lockup. The
critical brake torque needed to sustain optimum braking, and the corresponding lockup
instability at that brake torque value, are discussed in Section 2.7.
It is again stressed that the function hb ðsÞ completely determines the nonlinear
dynamic behavior of the single-wheel system in braking over the entire range of
vehicle speeds and slip values. Given the dimensionless brake torque b , one need
only calculate the zeros of hb ðsÞ to quantify steady-slip values and the corresponding
invariant sets. The slope of hb ðsÞ at these steady-slip values indicates the stability of
the corresponding invariant manifolds. With this knowledge, a complete state space
description of the vehicle dynamics can be constructed for the brake torque of interest
from which information on regions of stable and unstable braking can be easily
extracted.

2.6. Hysteresis in the Single-Wheel Braking Model

The dimensionless function hb ðsÞ and the (u, s) dynamics reveal certain features of the
system that may otherwise be difficult to extract. Referring again to Figure 5, consider
the case when the saddle-node bifurcation has already occurred and that the current
state of the system is that of wheel lockup (Fig. 5e). One may intuitively guess that the
brake torque need only be reduced to a value slightly less than crb ¼ 15:250 in order
for stable braking to again be restored. This, however, is not the case. Although the
stable invariant point sþ reappears, the state of the system remains at (u, 1), or wheel
lockup since that point remains stable as well. In fact, b must be more drastically
reduced below the value Lb such that hb ð1Þ < 0, that is, lockup must be destabilized in
order to restore stable braking conditions. Once this occurs, the system state jumps
from wheel lockup to stable braking conditions. This ‘jump phenomenon’ may be
conveniently captured by a bifurcation diagram.
 392 B.J. OLSON ET AL.
Downloaded by [McGill University Library] at 00:55 15 October 2012

Fig. 6. Bifurcation diagram for the single-wheel braking model.

Figure 6 shows a plot of the invariant points s versus the brake torque. The upper-
and lower-branch solid lines correspond to the stable steady-slip values s ¼ 1 (denoted
by lockup) and sþ (denoted by stable braking), respectively, while the dashed line
corresponds to the values s (denoted by unstable). The nonlinearity in s of the
friction coefficient forms a region where the slip is multivalued. It is this multiplicity
of steady-slip values that is responsible for the jump phenomenon and hysteresis
behavior. As b is increased from zero, the state of the system is governed by the
stable lower branch. Eventually slip destabilizes at a value corresponding to point a
(where b ¼ cr b and s ¼ scr ) and jumps to the upper branch at wheel lockup. In order
to restore stable braking conditions, the brake torque must be reduced to a value
corresponding to the point b (where b < Lb such that hb ð1Þ < 0) where lockup
destabilizes causing slip to jump back to the stable lower branch.

2.7. The Transition to Unstable Braking

When a brake torque is applied to a rubber tire a tractive force is generated at the tire/
road interface, as described in Section 2.1. The standard thinking is that the brake
torque can increase until wheel slip reaches the value sp, beyond which lockup occurs
[1, 2]. Under steady-state braking conditions, the corresponding maximum brake
torque is thus typically taken to be:
Tbp ¼ mgRðsp Þ; ð21Þ
which is the maximum moment that can be provided by the friction force
X ¼ ðsp ÞZ ¼ ðsp Þmg about the wheel center. Thus, it is traditionally assumed
that the critical brake torque at which the lockup instability is impending and the
peak brake torque are the same, and that the transition to wheel lockup occurs
 NONLINEAR DYNAMICS OF VEHICLE TRACTION 393

at s ¼ sp ¼ scr . In what follows, it is shown that the lockup instability does not occur
at the peak value sp corresponding to the maximum of the (s) curve, but at a
condition that is typically nearby. It is subsequently shown that Equation (21) is
actually only an approximation that is accurate when the vehicle to tire/wheel inertia
ratio  is large.

2.7.1. Lockup Instability

Differentiating Equation (15) with respect to s and evaluating the resulting expression
at critical slip yields:
h0b ðscr Þ ¼ 0 ðscr Þðscr  1  Þ þ ðscr Þ ¼ 0: ð22Þ
Downloaded by [McGill University Library] at 00:55 15 October 2012

Recall that scr is the saddle node value in the slip dynamics, where, necessarily, the
slope of hb ðsÞ is zero. Since in Equation (22), ðscr Þ > 0 and ðscr  1  Þ < 0, it
follows that:
ðscr Þ
0 ðscr Þ ¼ > 0: ð23Þ
scr  1  
Equation (23) shows that the slope of the friction characteristic ðsÞ at the critical
value s ¼ scr is positive, rather than zero, whereas 0 ðsp Þ ¼ 0. This means that, for
wet or dry asphalt characteristics, scr is actually smaller than the peak value sp . Hence,
the lockup instability, or the transition to unstable braking, corresponds not to sp , the
peak of ðsÞ, but to the critical value scr , which is the minimum of hb ðsÞ.
These results are consistent with numerical evidence. Recall that the peak value
p ¼ ðsp Þ of Equation (7) occurs at sp ¼ 0:316, whereas scr ¼ 0:304, which is
approximately 4% less than sp . The corresponding critical brake torque is out-
lined next, from which Equation (21) can be obtained by invoking a series of
approximations.

2.7.2. The Critical Brake Torque

Recall from Section 2.5 that the vehicle must come to rest under steady-slip
conditions for which s ¼ 1 (wheel lockup) or s ¼ s. The brake torque corresponding
to a steady-slip value s ¼ s follows from Equation (15) and is given by:
b ¼ ðs  1  Þðs Þ; ð24Þ
which is obtained by invoking the condition hb ðs Þ ¼ 0. The critical brake torque cr
b
corresponding to steady-slip conditions is obtained by maximizing Equation (24) with
respect to s. The result is:

@b 
¼ 0 ðs Þðs  1  Þ þ ðs Þ ¼ 0: ð25Þ
@s s¼s
 394 B.J. OLSON ET AL.

Equation (25) is of the same form as Equation (22), which is the expression that
minimizes hb ðsÞ. Thus, it must be true that the critical brake torque is given by
Equation (24) with the steady-slip value s ¼ scr satisfying Equation (22). Recalling
2
R
that b ¼ Jg Tb and  ¼ mRJ , the critical brake torque can be written in the dimensional
form:
 
cr 1
Tb ¼ mgRðscr Þ 1 þ ð1  scr Þ : ð26Þ

Equation (26) indicates that the typically assumed maximum brake torque given by
Equation (21) follows from two fundamental assumptions: (1) the inertia ratio  of the
Downloaded by [McGill University Library] at 00:55 15 October 2012

vehicle is large relative to unity; and (2) the peak slip value sp can be attained. These
assumptions are generally acceptable so that, for many applications, the true critical
brake torque can be reasonably approximated by the assumed brake torque given by
Equation (21). This is shown next.
Idealizing the wheel as a thin uniform disk of mass mwheel , it follows that:

mR2 m
¼ ¼2 ; ð27Þ
J mwheel

where, recall, m is the mass of the vehicle/wheel combination. In most applications

m
mwheel so that 
1. Moreover, the peak value sp and scr are typically close.
Referring again to Equation (7) and noting that ðscr Þ ¼ ð0:304Þ ¼ 0:972, it follows
that pb ¼ 14:574 for  ¼ 15. Numerical simulations show that pb ¼ 15:250 (see
Fig. 5), rendering the approximation given by Equation (21) in error by less than 5%.

3. THE SINGLE-WHEEL ACCELERATION MODEL

This section describes a single-wheel vehicle acceleration model. Here, the term
acceleration refers to the positive rate of change of velocity in the longitudinal direction
due to an engine torque. Longitudinal acceleration is fundamentally dependent on two
main limitations: engine power and traction [1]. However, the ensuing investigation
assumes that sufficient engine power is available at any given instant to maintain a
constant torque on the wheel. Thus, focus is shifted to understanding tractive
properties, their dynamic characteristics, and how to maximize them.
The single-wheel vehicle acceleration model is physically identical to that of the
single-wheel braking model, consisting of a tire/wheel disk with radius R and polar
moment of inertia J. As depicted in Figure 7, it is constrained to move longitudinally in
the x-direction with its speed denoted as u. The available engine torque, acting in the
positive sense on the wheel, is denoted by Te. The vertical reaction force Z balances the
static weight mg, while the longitudinal force X serves to accelerate the vehicle.
 NONLINEAR DYNAMICS OF VEHICLE TRACTION 395

Fig. 7. Schematic of the single-wheel acceleration model and corresponding free body diagram.
Downloaded by [McGill University Library] at 00:55 15 October 2012

The forward vehicle speed u and longitudinal wheel slip s are chosen as dynamic
states. In vehicle acceleration it is assumed and taken as convention that !R > 0 and
u!R
0  u  !R. Thus, wheel slip s ¼ maxðu;!RÞ ¼ u!R
!R is defined on the interval
I ¼ ½1; 0, taking on the limiting values s ¼ 1 for pure slip (u ¼ 0) and s ¼ 0
for free rolling without slip (u ¼ !R > 0). The former case when u ¼ 0 indicates finite
rotation of the wheel while maintaining zero vehicle speed. This is possible only in the
presence of an externally applied force on the vehicle, which is not present in our
model, to maintain u ¼ 0. The case when u ¼ !R implies the absence of an engine
torque.
The Pacejka tire model could be employed in vehicle acceleration studies by
letting s ! s in the Magic Formula [Equation (6)]. For the present study of
acceleration, we use the slip characteristic
ðsÞ ¼ c1 ð1  ec2 s Þ þ c3 s; ð28Þ
with c1 ¼ 1:18, c2 ¼ 10:0, and c3 ¼ 0:5, which is similar to Equation (7). As
indicated in Figure 3, it is simply the reflection of ðsÞ used for braking about the
vertical axis through s ¼ 0, and thus has a peak value p ¼ 0:972 at sp ¼ 0:316.

3.1. Equations of Motion

Assuming that the friction law given by Equation (5) holds and making the
appropriate substitutions, the equations of motion can be cast in the form:

u_ ¼ ðsÞg
; u > 0; s 2 ð1; 0: ð29Þ
s_ ¼ gu ha ðsÞ

The nondimensional function ha ðsÞ is given by:

ha ðsÞ ¼ ðs þ 1Þ2 ½ðs þ 1Þ1 ðsÞ þ ðsÞ  e ; ð30Þ
2
where, again,  ¼ mRJ is the vehicle/wheel inertia ratio and e ¼ Jg
R
Te is the
dimensionless engine torque.
 396 B.J. OLSON ET AL.

3.2. Steady-Slip Conditions and Local Stability

For nonzero u and a constant slip value s for which ha ðs Þ ¼ 0, Equation (29) shows
that wheel slip remains invariant, independent of the vehicle speed. Correspondingly,
the forward vehicle acceleration is positive and constant, and its speed monotonically
increases according to the equation:
uðtÞ ¼ uo þ ðs Þgt; t  0; ð31Þ
where uo > 0 is the initial speed when s ¼ s , that is, when t ¼ 0. Clearly the vehicle
cannot continue to accelerate indefinitely. Due to aerodynamic drag, saturation of
engine power, and so on, generation of the prescribed engine torque eventually
Downloaded by [McGill University Library] at 00:55 15 October 2012

becomes impossible. In order to quantify this limiting case one must include other
factors in the dynamic model, which will not be considered here.
Local stability of the invariant points s follows in the same manner as discussed
for the single-wheel braking model. (See Section 2.4). Stable and unstable fixed
points are defined to be:

s  ¼ sjha ðs  Þ ¼ 0; h0a ðs  Þ ? 0 ;

and the corresponding invariant manifolds are given by:

W a ¼ fðu; sÞju > 0; s ¼ s  g:
Note that the set fðu; sÞju > 0; s ¼ 1g does not define an invariant manifold, since s_
is singular when u ¼ 0.

3.3. Global Features of the Single-Wheel Vehicle Acceleration Model

The dynamic equations describing the single-wheel acceleration model are of the
same structure as their braking model counterparts, with the only significant
differences appearing in ha ðsÞ. Whereas the brake torque appears simply as an
additive term in the function hb ðsÞ of the single-wheel braking model [Equation (15)],
the engine torque is scaled by the nonlinear term ðs þ 1Þ2 in the function ha ðsÞ of the
single-wheel acceleration model [Equation (30)]. It is this nonlinearity that yields
slightly more complicated dynamics as the engine torque parameter e is varied.
Depicted in Figure 8 is an example plot of the function ha ðsÞ versus slip and the
corresponding state space dynamics in u and s for  ¼ 15 and e ¼ 15:65. At this
parameter value there are two stable invariant points sþ , which define two invariant
manifolds W þ 
a , and a single unstable steady-slip value s that defines the invariant

manifold W a . As with the single-wheel braking model of Section 2, a single function
ha ðsÞ describes completely the nonlinear dynamic behavior of the single-wheel
acceleration model. The full range of dynamic possibilities is captured by the
 NONLINEAR DYNAMICS OF VEHICLE TRACTION 397
Downloaded by [McGill University Library] at 00:55 15 October 2012

Fig. 8. Example plot of ha ðsÞ versus s and corresponding state space descriptions in u and s for  ¼ 15 and
e ¼ 15:65.

Fig. 9. Bifurcation diagram for the single-wheel acceleration model.

bifurcation diagram shown in Figure 9, which shows a plot of the invariant points s
versus the engine torque. The upper- and lower-branch solid lines correspond to the
stable steady-slip values sþ (denoted by stable acceleration and break-loose,
respectively) while the dashed line corresponds to the values s (denoted unstable).
As the engine torque is increased from zero, the state of the system is governed by the
stable upper branch. Eventually slip destabilizes at value corresponding to point a and
jumps to the lower branch, on which the slip is quite large, implying that there is
significant wheel spin with relatively little forward acceleration. Physically, this
corresponds to a very sudden increase in wheel spin, known in the colloquial as
‘burning out’ or ‘peeling out,’ and labeled as ‘break-loose.’ By further increasing the
engine torque s approaches, but never reaches, pure slip at s ¼ 1. In order to
reestablish low-spin stable acceleration, the engine torque must be decreased to a
level corresponding to point b, where slip jumps back to the stable upper branch.
 398 B.J. OLSON ET AL.

4. CONCLUSIONS AND DIRECTIONS FOR FUTURE WORK

The results presented here offer new insight into the behavior of vehicles during
longitudinal braking and acceleration. In each case considered, the unique features of
the modeling approach allow one to capture the full range of dynamic behavior of
single-wheel models in a simple geometrical manner. By choosing the forward
vehicle speed and longitudinal wheel slip as dynamics states, the dynamic equations
of motion lend themselves to a relatively simple investigation and interpretation using
the tools from nonlinear dynamics. This choice of dynamic states, where wheel slip
plays a central role, allows the dynamics for the entire range of vehicle speeds and slip
Downloaded by [McGill University Library] at 00:55 15 October 2012

values to be captured by a single function (one each for braking and acceleration).
These functions completely describe the tractive behavior of a given vehicle on a
given road surface in terms of slip and the brake or engine torque. The relative
simplicity of the analyses described herein is a consequence of the choice of dynamic
states and the interpretation of the resulting equations of motion.
Perhaps the most important conclusion from this work is the fact that the lockup
instability in the single-wheel model does not occur when the brake torque leads to the
maximum point on the slip curve, but at a lower brake torque. The traditional
assumption – that attaining the maximum coefficient of friction leads to lockup – is
shown to be an approximation that is accurate only when the tire/wheel inertia is small
compared to the vehicle inertia. Since this ratio is typically small, the approximation
is quite good. However, when considering a light vehicle with relatively large tire/
wheel inertia, the approximation becomes less accurate. In either case, it is of interest
to note that the commonly held view of lockup for the single-wheel model is only an
approximation.
This analysis is the first step in a new direction for the modeling of longitudinal
ground vehicle traction, and much remains to be done. Two-wheel traction models are
considered in a companion paper [7], although the two-wheel acceleration model has
not been studied in detail. Some other lines of future work include more detailed
parameter studies for specific vehicles under various road conditions, and the
incorporation of these models into ABS/TCS development, where slip also plays a
central role.

ACKNOWLEDGEMENTS

This work was supported in part by the Department of Applied Mechanics at the
Budapest University of Technology and Economics, the Department of Mechanical
Engineering & the Institute for Global Engineering Education at Michigan State
University, the US-Hungarian Joint Fund for Technological Development, and by a
grant from the National Science Foundation. The authors wish to thank Mr. Craig
 NONLINEAR DYNAMICS OF VEHICLE TRACTION 399

Gunn of Michigan State University and Mr. Jim Schramski of ACRA, Incorporated
for reviewing this manuscript. Thanks are also due to Professor Brian Feeny of
Michigan State University for his insight and helpful comments.

REFERENCES

1. Gillespie, T.: Fundamentals of Vehicle Dynamics. Warrendale, Society of Automotive Engineers

(SAE), Inc, 1992.
2. Wong, J.: Theory of Ground Vehicles. Wiley, New York, 1978.
3. SAE Standard: Antilock Brake System Review. SAE Paper No. Iaa46, 1992, pp. 90–102.
Downloaded by [McGill University Library] at 00:55 15 October 2012

4. Bakker, E., Nyborg, L. and Pacejka, H.: Tyre modelling for use in vehicle dynamics studies. SAE Paper
No. 870421, 1987, pp. 190–204.
5. Bakker, E., Pacejka, H. and Lidner, L.: A new tire model with and application in vehicle dynamics
studies, SAE Paper No. 890087, 1989, pp. 101–113.
6. Liu, Y. and Sun, J.: Target slip tracking using gain-scheduling for antilock braking systems.
Proceedings of the American Control Conference, 2 (June 1995), pp. 1178–1182.
7. Olson, B., Shaw, S. and St
epán, G.: Stability and bifurcation of longitudinal vehicle braking. Preprint,
Department of Mechanical Engineering, Michigan State University, USA.
8. Kobayashi, K., Cheok, K.C. and Watanabe, K.: Estimation of absolute vehicle speed using fuzzy logic
rule-based kalman filter. Proceedings of the American Control Conference 5 (1995), pp. 3086–3090.
9. Goodenow, G., Kolhoff, T. and Smithson, F.: Tire-road friction measuring system – a second
generation. SAE Paper No. 680137, 1968, pp. 571–579.
10. Harned, J., Johnston, L. and Scharpf, G.: Measurement of tire brake force characteristics as related to
wheel slip (antilock) control system design. SAE Paper No. 690214, 1969, pp. 909–925.
11. Sitchin, A.: Acquisition of transient tire force and moment data for dynamic vehicle handling
simulations. SAE Paper No. 831790, 1983, pp. 1098–1110.