Vehicle Traction Control and Its Applica
Vehicle Traction Control and Its Applica
Vehicle Traction Control and Its Applica
Transportation Technology
UC Berkeley
Title:
Vehicle Traction Control And its Applications
Author:
Kachroo, Pushkin
Tomizuka, Masayoshi
Publication Date:
01-01-1994
Series:
Research Reports
Permalink:
http://escholarship.org/uc/item/6293p1rh
Keywords:
Motor vehicles--Automatic control, Automobiles--Automatic control, Automobiles--Dynamics
Abstract:
This paper presents a study of vehicle traction control and discusses its importance in highway
automation. A robust control strategy is designed for slip control, which in turn controls the traction.
It is shown how traction control can be used to satisfy different objectives of vehicle control. The
importance of traction is further emphasized by comparing its performance to passive controllers in
a simulation study in which an impulse-like wind disturbance is introduced. The comparative study
shows that the system under traction control is stable in the presence of external disturbances,
whereas the system under passive control may become unstable in the presence of external
disturbances.
Copyright Information:
All rights reserved unless otherwise indicated. Contact the author or original publisher for any
necessary permissions. eScholarship is not the copyright owner for deposited works. Learn more
at http://www.escholarship.org/help_copyright.html#reuse
Pushkin Kachroo
Masayoshi Tomizuka
UCB-ITS-PRR-94-08
The contents of this report reflect the views of the authors who are
responsible for the facts and the accuracy of the data presented herein.
The contents do not necessarily reflect the official views or policies of
the State of California. This report does not constitute a standard,
specification, or regulation.
March 1994
ISSN 1055-1425
Vehicle Traction Control And Its ApplicationszyxwvutsrqponmlkjihgfedcbaZY
Abstract
1. Introduction
Vehicle traction control can greatly improve the performance of vehicle motion and
stability by providing anti-skid braking and anti-spin acceleration. Vehicle traction control
is especially important for automated highway systems as related to longitudinal and lateral
control.
Many companies have developed and used anti-lock braking (ABS) and anti-slip
acceleration control systems [7, 111. A typical commercial ABS system is composed of
sensors, a control unit and a brake pressure modulator. In the prediction stage, the control
logic of such a system uses information from the wheel angular velocity and/or acceleration
to estimate the wheel slip. Wheel slip is a nonlinear function of wheel angular velocity and
vehicle velocity and is described in more detail in section-2. The control command is based
on the estimated slip and wheel acceleration. The wheel slip/acceleration phase plane is
divided into different sectors. Each sector has a corresponding control action (e.g.
APPLY, HOLD, RELEASE). This design process tries to produce an optimal limit cycle
in the phase plane of wheel slip and acceleration. The control stage of the algorithm is
usually referred to as the selection stage. Similar algorithms are designed for anti-spin
acceleration. Although these systems work in practice, their design is experimental rather
than analytical, and their tuning and calibration rely on trial and error. With the recent
wheel slip.
Advanced Technology for Highways (PATH), we can use an alternate method to estimate
wheel slip. As angular wheel velocity is measured directly, we only need vehicle velocity to
estimate slip. For lateral control of vehicles in the PATH program, magnetic markers are
3
installed on the road at specified distances from each other in the longitudinal directions.
Vehicle velocity can thus be estimated by measuring the time elapsed between consecutive
markers.
avoid accumulation of integration error, the initial velocity should be updated (from wheel
angular velocity) every few seconds before acceleration or braking starts. At the initiation
of acceleration or braking, the last initial condition should be used for the integration
process. Additional hardware may also be required to reduce the accumulation of the error
In this paper, it is assumed that vehicle velocity and wheel angular velocity are both
a part of highway automation, longitudinal and lateral guidance of the vehicle should be
addressed. The input forces, which control the vehicle motion come from the road-tire
interaction and have two components, one in the longitudinal direction and one in the lateral
direction. The h-active force in the lateral direction depends on the cornering stiffness and
can be controlled by the steering [8,9]. The tractive force in the longitudinal direction, on
the other hand, is a nonlinear function of the wheel slip and can be controlled by
maintaining the wheel slip at some required value. The throttle and the brakes ultimately
control the longitudinal tractive force. Controlling the longitudinal traction can achieve
various control objectives while assuring ride quality and passenger comfort. A few of
these are:
(4) Make vehicles move longitudinally in a platoon following the vehicles in front in
an automated highway system.
(5) Make a platoon of vehicles follow a desired lateral and longitudinal path
simultaneously in an automated highway system.
A vehicle system could also use different traction control algorithms at different times
after assessing which control law is appropriate at that time instant. A supervisory control
could be devised to make such decisions. For instance, in a platoon of vehicles which is
following a curved path using control law designed for objective (5), if a vehicle starts
veering out and thus creating an emergency or abnormal condition, then changing the
control law to satisfy objective (3) might prove to be a good way to solve the problem.
In the studies for longitudinal control and platooning, it is generally assumed that the
road can provide necessary forces as determined by the controller. This assumption
implies that either the vehicle is operating in a range such that the vehicle can respond using
The dynamics for the system are highly nonlinear and time varying, which motivates
the use of sliding mode control strategy [13] to follow a target slip. Lyapunov stability
theorem based [ 15, 161 and sliding mode based [ 17, 18, 191 controllers have been assessed
by researchers. The sliding mode controller designed for vehicle traction control is made
adaptive to reduce the control discontinuity around the switching surface of the sliding
mode. Sliding mode based scheme is also used to estimate the road tire conditions for
maximum acceleration and maximum deceleration. The main problem with sliding mode
control is the high frequency chattering across the switching surface[ 131. A boundary layer
is introduced around the switching surface and an appropriate function is used in the
Traction controllers, on the other hand, handle disturbances well. Simulations were
performed on various road conditions to compare the performances of the two types of
controllers. A simulation study was also performed to compare the effects of wind
5
disturbance on the two types of controllers for longitudinal control of vehicles. The results
of the study are given in section-10 of this paper, which confirm the advantage of using
traction control.
2. Background
needed. In this section, a mathematical model for vehicle traction control is described [S,
6, 17, 18, 191 for analysis of the system, design of control laws, and computer
simulations. Although, the model considered here is relatively simple, it retains the
stability of the system, described in this section, is analyzed by linearizing the system
described in this sub-section. The model identifies wheel speed and vehicle speed as state
variables and wheel torque as the input variable. The two state variables in this model are
associated with one-wheel rotational dynamics and linear vehicle dynamics. The wheel
where J, is the moment of inertia of the wheel, o, is the angular velocity of the wheel, the
over dot indicates differentiation with respect to time, and the other quantities are as defined
in Table 1. The total torque acting on the wheel divided by the moment of inertia of the
wheel equals the wheel angular acceleration. The total torque consists of shaft torque from
the engine, which is opposed by the brake torque and the torque components due to the tire
tractive force and the wheel viscous friction force. The wheel viscous friction force, a
function of the wheel angular velocity, is the friction force developed on the tire-road
contact surface. The tractive force developed on the tire-road contact surface is dependent
on the wheel slip, the difference between the vehicle speed and the wheel speed, normal&d
by the vehicle speed for braking and the wheel speed for acceleration (see Eq. (2)). The
engine torque and the effective moment of inertia of the driving wheel depend on the
Th Brake torque
F. Tractive force
Linear portion
of the curve
(Deceleration)
-1 0
Wheel Slip (h)
force at the tire-ground contact patch [14,20]. The driving torque produces compression at
the tire tread in front of and within the contact patch. Consequently, the tire travels less
distance than it would if it were free rolling. In the same way, when a braking torque is
applied, it produces tension at the tire tread within the contact patch and at the front.
Because of this tension, the tire travels more distance than it would if it were free rolling.
This phenomenon is referred to as the deformation slip or wheel slip [14, 17, 18, 19,201.
The adhesion coefficient p(k) is a function of wheel slip h. Figure 1 shows a typical p-h
curve. References [17, 18 and 191 are the sources for the typical curve and [8] gives a
more mathematical description of the tire model. Mathematically, wheel slip is defmed as
h=(Ow-&)/U,OfO (2)
8
(3)
which is equal to the linear vehicle velocity, V, divided by the radius of the wheel. The
variable o is defined as
0, for 0,2 0, \ (4)
O=max(0,,0,)=
1 WV for ow < ov \
which is the maximum of vehicle angular velocity and wheel angular velocity.
The tire tractive force is given by
where the normal tire force, N,, depends on vehicle parameters such as the mass of the
vehicle, location of the center of gravity of the vehicle, and the steering and suspension
dynamics. The adhesion coefficient, which is the ratio between the tractive force and the
normal load, depends on the road-tire conditions and the value of the wheel slip [5, 171.
For various road conditions, the curves have different peak values and slopes, as shown in
parameters like speed and vertical load. The average peak values for various road surface
I Gravel 0.6
Ice 0.1
1 .o W
Pavement
Wet Asphalt
The model for wheel dynamics is shown in Figure 3. The parameters in this figure
are defined in Table 1. The figure shows the acceleration case for which the tractive force
and wheel viscous friction force are directed toward the motion. The wheel is rotating in
the clockwise motion and slipping against the ground, i.e. o, > w. The slipping produces
the n-active force towards right causing the vehicle to accelerate towards right. In the case
of deceleration, the wheel still rotates in the clockwise motion but skids against the ground,
i.e. q < %. The skidding produces the tractive force towards left causing the vehicle to
decelerate.
1 0
ground
Traction Force F,
zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
The vehicle model considered for the system dynamics is shown in Figure 4. The
M, : Vehicle mass
(during braking).
The linear acceleration of the vehicle is governed by the tractive forces from the
wheels and the aerodynamic friction force. The n-active force, Ft, is the average friction
1 1
force of the driving wheels for acceleration and the average friction force of all wheels for
deceleration. The dynamic equation for the vehicle motion is
The linear acceleration of the vehicle is equal to the difference between the total tractive
force available at the tire-road contact and the aerodynamic drag on the vehicle, divided by
the mass of the vehicle. The total tractive force is equal to the product of the average
friction force, Ft and the number of relevant wheels, NW. The aerodynamic drag is a
nonlinear function of the vehicle velocity and is highly dependent on weather conditions. It
The dynamic equation of the whole system can be written in state variable form by
x2 = 0, (8)
and denoting x = max ( x1,x2 ), we can rewrite Equations (1) and (6) as
il = -fdXd + blN CL (h > (9)
22 = -f2(x2) - b2N l.t (h ) + b3T (10)
where
T=T,-Tb
h = (x2 - x1)/x)
flh) = [Fv@wxdl/(MvRw)
blN = NvNw/(M,Rw)
f2@2) = Rw&vCQYJW
bzN = RwNv/J,
b3 = l/J, (11)
1 2
The combined dynamic system can be represented as shown in the Figure 5. The control
input is the applied torque at the wheels, which is equal to the difference between the shaft
torque from the engine and the braking torque. During acceleration, engine torque is the
primary input where as during deceleration, the braking torque is the primary input.
x = max(o, V/R,)
Wheel slip is chosen as the controlled variable for traction control algorithms because
of its strong influence on the tractive force between the tire and the road. Wheel slip is
calculated from Equation (3) by using the measurements of wheel angular velocity and the
estimated value of the vehicle velocity from either the accelerometer data or the magnetic
marker data, as explained in the Introduction. By controlling the wheel slip, we control the
tractive force to obtain desired output from the system. In order to control the wheel slip, it
is convenient to have system dynamic equations in terms of wheel slip. Since the
functional relationship between the wheel slip and the state variables is different for
acceleration and deceleration, the equations for the two cases are described separately in the
following sub-sections.
1 3
Acceleration Case
caused by the following factors: (1) the relationship between wheel slip and wheel velocity
and vehicle velocity is nonlinear, (2) the p--h relationship is nonlinear, (3) there are
multiplicative terms like (1 - h)b&t/x2 and (1 - h&T/x2 in the equation, and (4) functions
Deceleration Case
During deceleration, condition x2 I x1, (x1+0) is satisfied, and therefore wheel slip
is defined as:
Substituting Equations (9), (10) and (15) into Equation (16), we obtain
h = [ [(l + h)fl(xl) - f2(x2)] - [b +(l + h) blN]jt + bJl/xl (17)
This gives the wheel slip dynamic equation for deceleration case. This equation is also
The local stability of a nonlinear system can be studied by linearizing the system
around its equilibrium point. Therefore, in this section, the vehicle nonlinear system
equations are linearized around the equilibrium point in order to study the system stability.
The equilibrium point (x10,x20) of the vehicle system described by Equations (9) and (10)
can be obtained by equating the right hand sides of the two equations to zero.
Acceleration Case
-
[
ap
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2
X20 1
Notice that dfl/dxl, df2/dx2, x1, x2, blN and b2N are all positive. When acl/ah is
positive, the eigenvalues of A have negative real parts, and when aCL/ah is negative, the
dxl
dfl,XlO) + &x20)
df2
w9
[
b&x20 + b&@ >
40 1 CL
h
H
Hence, under the condition (20) the system is unstable.
1 5
Deceleration Case
For the deceleration case, the Jacobian matrix at the equilibrium is:
df 1 acL (21)zyxwvutsrqponmlk
- -+x10) -bm-ho,xzo)%
dxlzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
ah XT0
A=
acL
bm-@o,x2o~~
L
ah XT0
2
X10 1
Here also dfl/dxI, df2/dx2, x1, x2, blN and b2N are all positive, so when aCL/ah is
positive, the eigenvalues of A have negative real parts. When a@h is negative, the
dxl
df 1
+x10> + %x20)
dx2 (22)
x20
h + bN/xlO
40 1 >
CL
Hh
3. Slip Control
strategy based on sliding mode, which is robust to parametric uncertainties, is chosen for
slip control.
The following is the derivation of the sliding mode control law for wheel slip
1 6
regulation.
The slip dynamic equation for acceleration (13) can be written as
i=f+bu (23)
where
f = $dxl) - (1 - a)(fdXd + bi?N/@)) - bmCl(~)l (24)
u = (l - ‘1, (25)
x2 ’
and
b = b3 (26)
Define the switching surface S(t) by equating the sliding variable, s defined below to zero.
s=h, (27)
where h, = h - & and b is the desired slip. The nonlinear function f is estimated as ‘i, and
and its bounds can be time varying or state dependent. Since the control input is multiplied
by the control gain in the dynamics, the geometric mean of the lower and upper bounds of
the gain, G = V bmmbmin, is taken as the estimate of b. The bounds can also be written as
CC-~ I g/b I a, where a = 2’bmaJbmin. The controller is designed as
T--=-u
(1-V (28)
where
the system is guaranteed with an exponential convergence once the switching surface is
chattering [13]. Chattering is caused by non-ideal switching of the s variable around the
switching surface. Delay in digital implementation causes s to pass to the other side of the
surface S(t), which in turn produces chattering. A practical method for avoiding chattering
is to introduce a region around S(t) so that s changes its value continuously [13]. In the
present work, we define a boundary of a fixed width $ around the switching surface and
j = +W<xd, (33)
and u is changed to
u = g-‘G - k int(aj,s,+)] (34)
The bandwidth of the filter for variable s is given by ‘y.
Notice that the second term in (34) acts as a PI controller in the region Isl < 0. The
first term is given by (30) which attempts to cancel the nonlinear term in (23) and further
adds the desired dynamics. If the cancellation of the nonlinear term is perfect, i.e.
f - (bi)T = 0, Equations (23), (30), (32) and (34) will result in a linear error equation with
no forcing term, which implies that the slip error and the integral of the slip error as well as
the sliding variable, s, all converge to zero. However, the cancellation can never be perfect,
which can be easily understood by the presence of p,(h) in Equation (24). The integrator
can absorb the error due to imperfect cancellation and assures a superior performance.
In the later portion of this report, sliding mode control will be utilized for longitudinal
platoon control, and the associated control variable will include a term proportional to the
integral of the error, introduced by the int(.) function. It should be noted that the
For deceleration, we can obtain the system in the form of Eq. (23) by defining f, u
andbas
and
b = b3 (37)
Following the same steps as for the acceleration case, the control law for deceleration is
define
A S = s - p(.)int(a,j,s,$>, (38)
where
t sdt] (39)
(41)
when
kZa(F+q) (43)
and
A
b&s
(44
b
The adaptation law is given by Equation (44). The block diagram of the controller is
shown in Figure 6. In this figure, D refers to the Laplace operator and the dashed line
- Aa
-T-- ----- JI 1-1 I--
+-@ 1
I I
-t
f
The tire/road surface contact can be characterized by the local slope of the lt - h curve.
Maximum u-active force occurs at the peak of the curve where acl/ah is zero. Maximum
positive tractive force is achieved at the positive peak of the curve and is desirable in order
achieved at the negative peak of the curve and is desired for maximum deceleration. The
2 0
acceptable operating zone for the wheel slip is between the peak slips where the slopes are
positive; outside this zone, the slopes are negative. To produce the fastest acceleration or
deceleration, which can be seen as minimum-time control, the wheel slip should be
regulated where the adhesion coefficient attains the peak value: a positive peak value for
acceleration and a negative peak value for deceleration. The derivation of this optimal
formulation is given in the appendix section of Tan’s thesis [17]. A method to estimate the
slope of the l.t - h curve, based on a single parameter sliding mode parameter estimation
scheme is described below. The algorithm is based on estimating the slope of the l.~ - h
curve and then moving the target slip in the direction of the peak slip, until the peak slip is
reached.
The minimum time acceleration and minimum time deceleration control can be
employed in a single vehicle system In an automated highway system where a platoon of
vehicles is to be controlled, minimum time acceleration and minimum time deceleration are
4.1 Estimation
This section explains how to use the sliding mode parameter estimation to identify the
local slope of the l.t - h curve, &./ah, from the estimations of l.t and h. In practice, the
adhesion coefficient, p,, can be regarded as a function of time (road changes) and of slip
(operating condition variations), i.e. l.t = l.t(h, t). Under the assumption that the road
(45)
for virtually all time instants tk, where Al.t and Ah are the differences between two adjacent
To obtain the value of Ap, we differentiate Equation (9) and rearrange terms.
2 1
(46)
Since the second term inside the bracket is usually insignificant compared with the first
term, and the first term can be approximated by the difference of vehicle velocities at two
consecutive sampling time instants divided by the sample time interval, we can approximate
Ai aszyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Ai; = bh(tk) - ‘h(tk-I)$+--&
(47)
An alternate estimation scheme could be based on differentiating Equation (10) instead of
Equation (9).
The slope is estimated by using the sliding mode parameter estimation scheme. The
where k’ is the estimator gain and $’ is the boundary width across the switching surface.
variable a’ is designed to provide a constant bandwidth filter which removes the high
where y is the desired constant bandwidth of the filter, and almax is the maximum value of
2 2
We can utilize the sign of the current &L/&L, which is the output of the sliding mode
parameter estimation process, to reach the peak slip. This information reveals the direction
of the peak slip from the present position of h. For example, if the slope is positive at
h(tk), we know that the present slip is within the positive slope region. We therefore
propose a simple and practical method for updating the estimated peak slips based on the
sign of the estimated slope. The proposed method contains the following basic steps: (1)
Move the target slip in the estimated direction towards the peak slip. (2) Steer the wheel
slip toward the new target slip via the sliding mode slip tracking algorithm. (3) At the
same time, estimate the new local slope in the p - h curve and then return to step 1. The
algorithm designed to move the target slip toward the peak slip (step 1) is called the target
update algorithm.
The target slip is updated according to the following algorithm. The target slip is
moved in the direction of the peak according to the sign of the slope of the p - h curve.
The step size remains the same if the sign of the slope does not change, and the step size is
where,if the sign of the estimated slope at the (k+l)st instant is different from the one
estimated at the kth instant, then
step(O) = a (56)
2 3
For varying road conditions, either the step size should have a lower bound or it should be
re-initialized if the value of the estimated slope varies substantially after a steady value has
been reached. This is important because otherwise the estimation process would cease.
Notice in Equation (53) that n = 1 is the acceleration case and n = 2 is the deceleration case.
Chattering is observed when using the combined algorithm. This chattering should
not be confused with the sliding mode chattering for which a solution has been proposed.
This chattering is due to rapid updating of the target slip. In order to smooth the control,
the algorithm is modified such that the target is updated only when Isl c E, where s is the
sliding variable used in slip control and e is chosen to be some small positive number
For anti-spin acceleration and anti-skid braking, the wheel slip should be maintained
at the positive slope region of the jt- h curve. To accomplish this, we can use the
estimation scheme described in the last section. To obtain anti-spin acceleration, if the
current wheel slip is in the negative slope region, the target slip should be decremented, and
if the wheel slip is in the positive slope region, the target slip is not changed. Similarly, to
obtain anti-skid deceleration, if the current wheel slip is in the negative slope region, the
target slip should be incremented, and if the wheel slip is in the positive slope region, the
When the aim of a traction control system is only anti-skid braking and anti-spin
acceleration without any demands on velocity or time, the ultimate control objective is not
the regulation of the wheel slip itself. The wheel slip should not enter the negative slope
2 4
region. From this point of view, a fuzzy rule-based control approach would also be
Although maximum traction is desirable for optimum time control while decelerating
in straight line braking, a trade off between stability and stopping distance may be
necessary during combined hard braking and severe steering maneuvers. In these type of
maneuvers, understeered vehicles - even those equipped with ABS - will oversteer and spin
out. A previous study [15, 161 shows that for lateral control of a vehicle while
decelerating, the best cornering performance and steerability of the vehicle can be attained
by maintaining the longitudinal wheel slip at the values given by Equations (57) and (58).
The cornering performance and steerability of the vehicle is measured by the smallest
turning radius without loss of stability, combined with the shortest stopping distance.
ha = aeb16sl (57)zyxwvutsrqponmlkjihgfed
= cedl&l (58)
hrT
In the equations, 6s is the forward steering angle; a, b, c and d are constants, the optimum
values of which can be obtained by simulations, and finally, hi and &T are the front and
the rear target slip values. These equations were obtained from analysis of simulation
results, conducted by Taheri [ 161, where different combinations of desired front and rear
slip values were used. It was found that the best combination of desired slip values, in
functions of the front wheel steering angle. To control the wheel slips at these values, the
I I direction of motion I
Ir-y- jqzyxwvutsrqponmlkjihgfedcbaZYXWV
A2 4
Longitudinal control strategies are necessary in order to regulate the spacing and
(see Figure 7). The longitudinal control algorithm must maintain the spacing policy under
normal maneuvers such as acceleration, deceleration, turning, and merging. The controller
must also insure good performance over a variety of operating points and external
longitudinal platoon is obtained by using two sliding surfaces, one for tracking the vehicle
velocity, and the other to obtain the input torque. This ” two surface” method [4] is used
for systems with relative degree greater than one. The technique is discussed below.
Sl = ti + ClE (59)
where E is the spacing error between the vehicle being controlled and the vehicle in front.
The spacing error is defined as the difference between the actual distance between the two
vehicles and the desired distance between them. When the desired distance between the
two vehicles is a constant, i: becomes the velocity error between the vehicle to be controlled
Sl = ii + Cli (60)
..
where a = it - xl&s, and Xl&s is the desired xl, which is equal to the velocity of the front
vehicle for a constant desired spacing. For the first sliding surface, p(X) is the control
input, which is further controlled by the second sliding surface using the system control
input T. Using the sliding mode design procedure and introducing - ksgn(sl) term for
robustness, we obtain
$1 = il - k&s + cl& = - ksgn(sl) (61)
Substituting xl by the estimated quantities in the right hand side of Equation (9) yields
-&(x1) + i&;(h) = i&s - C1i - ksgn(sl)
From the desired adhesion value, the desired slip value h&s can be calculated using the
p, - h curve. Since the actual curve is not known, we use a nominal p - h curve. The error
in the estimated and actual p value is denoted by p&h). The sliding gain of the first surface
can overcome this error by utilizing the bounds on blNj&(h). The estimated value of
ft(xl) in the Equations (62) and (63) is shown by Tl(xl). For chattering reduction, the
function sgn(.) is replaced by int(.). Here also, the control law can be made adaptive to
reduce the discontinuity across the switching surface. To obtain the desired slip, we try to
control the wheel slip directly at the desired value using the algorithm for slip-control, by
s2 = h, w
The design of the control law, using the second sliding surface, is the same as described in
section 3.
In the platoon control problem, a decentralized control law is used for the special
vehicle. The position errors can be obtained by integrating the velocity errors. Thus, for
2 7
each subsystem containing a vehicle in front and a following vehicle, sliding mode control
guarantees the convergence of the spacing errors with a sum of finite and exponentially
convergent time periods, so that the overall system is convergent with a rate of the slowest
dynamical subsystem.
The longitudinal traction control developed in this paper can be combined with some
appropriate lateral control [8, 91 to satisfy the objective of building a complete motion
control system. In a combined platoon vehicle system for PATH, the longitudinal spacing
error between vehicles is controlled by the longitudinal controller, while the lateral
deviation and the yaw rate are controlled by the lateral controller.
It is important to use traction control when longitudinal and lateral controllers are
being implemented simultaneously, because wheel slip not only controls traction in the
longitudinal direction, but also in the lateral direction. At high wheel slip values, there is
less adhesion in the lateral direction and therefore, for lateral stability of the vehicle, the
9 Passive Control
Some studies for longitudinal control and platooning assume that either the road can
provide necessary forces for the controller in the operating range or that a traction control is
in place. This assumption implies that if the road can react sufficiently, passive or
tractionless controllers can be used to satisfy the control objectives. Passive controllers do
not explicitly take adhesion availability of the road-tire interaction into account and therefore
their range of operation is limited as compared to the traction controllers. A simple PID
control and a sliding mode based passive control are derived next.
2 8
(65)
tries to minimize the longitudinal spacing error E between two vehicles without taking
traction into account. Different weights are given to the proportional, derivative and
integral terms based on experimental data. The tuning of the PID gains relies heavily on
trial and error and the design is experimental rather than analytical. Hence, there is no
stability proof for this control on the highly nonlinear model of the system.
A tractionless sliding mode controller, which tries to maintain the longitudinal spacing
between vehicles, can be designed by differentiating Equation (9), so that the input variable
Xl = F1 + F2T (67)
where
+ blN*l il+ blN$(l _ $f2(x2) - b2NpL(h)1
ahX2 1 (68)
ahzyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGF
x2
and
F2 = blN--ap (l - ‘lb, (69)
ah X2
If we let
s = i + Cl& (70)
where & = x1 - Xl&s , the control law takes the form
T = -&[-@I + xldes - cl& - k sgn(s)]
F2
(71)
where Fl is the estimate of Fl, and ^Fz is the estimate of F2. Here also sgn(.) can be
replaced by an appropriate function for chattering reduction and the sliding gain can be
A simulation study was performed for the acceleration case without any measurement
noise. A typical lo - h curve was assumed fixed for the whole test road surface. During
acceleration, only the two wheels attached to the engine were considered, because the
inertia of the other two is negligible. The vehicle parameters for simulation during
Parameters Values
mass of the vehicle (MV) 1000 Kg
c = 0.595(R,/Mv)
f2@2) 0 rad/sec2
In the simulations, velocities and distances are expressed in radians/set and radians,
and can be easily expressed in meters/set and meters respectively by multiplying the
value of 0.175. The nominal road condition is utilized in designing the active sliding mode
30
traction controller. Our intent is to simulate performance under wide varying road surface
conditions. We chose nominal road condition for the control design purposes such that the
l.~ value would be the average of the values for extreme conditions. The nominal l.~ - h
curve corresponds to an earth road. However, note that this nominal condition does not
reflect the standard highway condition. Simulations are performed on dry concrete road
and a slippery road to show extreme road conditions. It is important to analyze the
performance in the whole range of road conditions, for the presence of substances like
grease and dirt can cause the surface conditions to vary. We have chosen to use the
varying surface conditions for evaluating external disturbances, because they have a
pronounced effect on the performance of vehicles in terms of their stability. For dry
concrete road the maximum ~1 value of 0.8 is chosen and the corresponding h is 0.2,
whereas, for a slippery road, the maximum l.t value of 0.2 is chosen and the corresponding
blN = 14.7548
b2N = 35.2284
b3 = 0.0497 (73)
Parametric uncertainty is taken to be about 25% for all parameters. The sampling
The simulation results for the slip control (34) for desired slip equal to 0.15 are
shown in Figure 8 and Figure 9. Figure 8 shows the simulation performed on a slippery
road and Figure 9 shows the simulation performed on dry concrete. Both figures show
good wheel slip tracking in spite of 25% modeling errors in the parameters.
3 1
5 o.j~ 1
0.05 , I 0.2 , I
i+.l.~~
1 0 0.1 0.2 0.3 0 0.1 0.2 0.3
Time(sec) Time (set)
:w
1
Time Response Controller
5, I
F-0.05 -
2
5; -0.1 .
0 0.1 0.2 0.3 0 0.1 0.2 0.3
Time(sec) Time (set)
The simulation results for the passive PID control law (65) are shown in Figure 10.
From the plot of the results, the performance is satisfactory. The disadvantage of this law
is that the gains of the controller have to be tuned based on trial and error, or by ignoring
f 0
$ 0 0.1 0.2 0.3 0 0.1 0.2 0.3
Time(sec) Time (set)
Simulation results for the longitudinal traction control of a vehicle trying to maintain a
constant spacing between itself and a vehicle in front while accelerating, using the two
sliding surface method described in section-7, are shown in Figure 11 and Figure 12.
Figure 11 shows the simulation on a slippery road, while Figure 12 shows the simulation
on dry concrete. The desired adhesion in the longitudinal direction is provided by the first
sliding surface, as given by Equation (63). The torque input is computed using Equation
(34), where the second sliding surface is defined in Equation (64). The calculation of the
desired wheel slip is based on the nominal p-h curve. The actual and desired velocities
follow each other very closely even in the presence of parametric uncertainties. The
spacing errors (not shown in the plots) are however smaller than those obtained by using
3 3
the passive PID control law. Notice that wheel slip is higher on the slippery road than on
dry concrete.
3-
01 I
0 0.1 0.2 0.3
Time(sec) Time (xc)
1000 Controller
F
g 500-
s
iz
E 0
u 0 0.1 0.2 0.3
3 Time(sec) Time (xc)
0
0 0.1 0.2 0.3 0 0.1 0.2 0.3
Time (xc)
Time(sec)zyxwvutsrqponmlkjihgfedcbaZYXWVUTSRQPONMLKJIHGFEDCBA
Results of the traction control which was synthesized with the sliding mode
estimation scheme for time optimal control while accelerating a vehicle, are shown in
Figure 13 and Figure 14. Figure 13 shows the simulation on a slippery road, while Figure
14 shows the simulation on dry concrete. The peak slip for the slippery road is 0.15 and
that for dry concrete is .2. The simulation results are satisfactory as the peak slips are
0.1 0.2
5 Time(sec) Time (xc)
2-
03
A tractionless single surface sliding mode controller (71) is used for longitudinal
control and the results are shown in Figure 15. The tracking is satisfactory.
Time Response 600. Controller
F
5 400-
$ 200-
”g 0
1 2 0 1 2 3
Time(sec) Time (set)
0 1 2 3 1 2
Time(sec) Time (set)
robustness comparison. In the longitudinal vehicle tracking while using the sliding mode
passive controller, when a wind force disturbance of 0.2g (g is the acceleration due to
gravity) is given at 0.3 seconds from the start of the simulation, for a period of 0.05
seconds, the slip becomes uncontrollable, as shown by Figure 16. The disturbance forces
the slip to attain values greater than 0.15, thereby causing it to be in the negative p - h slope
region. This makes the system unstable. Similar instability is seen in Figure 17 when the
same disturbance is given to the system, which is being controlled by the PID passive
control law, at 0.5 seconds from the start. However, using the control law with the
traction controller, as illustrated in Figure 18, the system remains stable when it encounters
the same disturbance at 0.3 seconds from the start. Although the disturbance used in these
simulations is not practical for the actual vehicle system, it gives insight into the stability of
1 :_---‘i
-9 0 1
Time(sec)
2 3
1
Time (set)
2
s
-5
0 1 2 3
Time(sec) Time (set)
Controller
2ooo-
1000 r
I-y
O*
-loo0 ;
0 0.5 1 1.5
Time(sec) Time (xc)
11. Conclusions
It was shown that traction control is important for safety and highway automation of
vehicles. A robust control strategy was designed for slip control, which in turn controls
the traction. It was shown how traction control can be used to satisfy different objectives
of vehicle control. The importance of traction control was further emphasized by
under traction control is stable in the presence of external disturbances, whereas the system
under passive control may become unstable in the presence of external disturbances.
vehicle with a complete set of sensors and controller integrated in a single system, or a
platoon of vehicles, where the sensors and the controllers are distributed within the vehicles
and the roadway, It can be used to accelerate or decelerate a single vehicle in the minimum
time, or it can be used to enhance the maneuvering ability of a vehicle, especially during
severe steering actions. Traction control also improves the performance of platoon of
vehicles in terms of stability and achieving a tighter control. Traction control makes the
system robust to external disturbances and provides a better control especially during
12. Acknowledgments
This work was performed as part of the PATH Program of the University of
facts and the accuracy of the data presented herein. The contents do not necessarily reflect
the official views or policies of the State of California. This report does not constitute a
standard, specification, or regulation.
References
PI Birch, S., Vehicle Sensor, Automotive Engineering, Vol. 97, No. 6, 91-92,
June 1989
Nov.-Dec. 1980.
[41 Green, J., and Hedrick, J. K., Nonlinear Torque Control for Gasoline Engines,
[51 Harned, J. L., et al., Measurement of Tire Break Force Characteristics as related
to Wheel Slip Control System Design, SAE Trans. Vol. 78, pp. 909-925, No.
690214, 1969.
[61 Leiber, H. et al., Anti-Skid System (ABS) for Passenger Cars, Bosch Technical
l-71 Leiber, H., and Czinczel, A., Four Years of Experience with 4-Wheel Antiskid
181 Peng, H., and Tomizuka, M., Vehicle Lateral Control for Highway Automation,
r91 Matsumoto, N., and Tomizuka, M., Vehicle Lateral Velocity and Yaw Rate
Control with Two Independent Control Inputs, Trans. of the ASM E , 114, pp.
606-613, 1992.
II101 Rohrs, C. E., Valavani, L. S., Athans, M., and Stein, G., Robustness of
II111 Schurr, H., and Dittner, A., A New Anti-skid-Brake System for Disc and Drum
WI Slotine, J.-J. E., and Coetsee, J. A., Adaptive Sliding Controller Synthesis for
[I31 Slotine, J.-J. E., and Li, Weiping, Applied Nonlinear Control, Prentice Hall,
[I51 Taheri, S., A Feasibility Study of the Use of a New Nonlinear Control Law for
Automobile Anti-lock Braking Systems, ACC Proceedings, vol. 2 of 3, 1990.
[161 Taheri, S., A, and Law, E. H., Investigation of Integrated Slip Control Braking
and Closed Loop Four Wheel Steering Systems for Automobiles during
Combined Hard Braking and Severe Steering, ACC Proceedings, ~01.2, 1990.
[I71 Tan, H. S., Adaptive and Robust Controls with Application to Vehicle Traction
1181 Tan, H. S., and Tomizuka, M., A Discrete-Time Robust Vehicle Traction
1191 Tan, H. S., and Tomizuka, M., An Adaptive Sliding Mode Vehicle Traction
Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.
Alternative Proxies: