Sie sind auf Seite 1von 2

TP07 4:lO

Proceedings of the 35th


Conference on Decision and Control
Kobe, Japan December 1996

A Sliding Mode Controller for Anti-lock Brake System:


Usage of Sluggish Actuators
Taketoshi Kawabe, Masao Nakazawa, Ikuro Nostu and Yoshito Watanabe
Research Dept. Nissan Diesel Motor Co., Ltd.
1-1 AgecFshi Saitama 362 Japan
1 Introduction

equation (2), the integrator provides a servo mechanism. Without loss of generality, we assume Di(q) is
monic hereafter. We chose the control law [5]:

Anti-lock brake systems (ABS) become popular nowadays as one of the most important automotive technologies for active safety. The main role of ABS is to
prevent wheels from locking by ensuring tire slip r a
tio within a certain range while braking on slippery
road surfaces. It is very important for an ABS controller to increase robustness against uncertainty in v e
hicle dynamics which is substantially changed by road
surface conditions, vehicle mass and other factors. B e
sides, an inherent non-linearity in a braking mech&
nism should be taken into account at the design of a
controller[l, 2, 31. In the case of large commercial vehicles, the brake actuators are more sluggish in response
than these of passenger cars, which make it difficult to
design an ABS controller having a good performance.
To deal with the non-linearity we use the methodology
of sliding mode control. A new method to use sluggish
actuators with sliding mode controller is proposed.

IT

= -(CB)-1K[sat(sl/41) - * - ~ a t ( s , / 4 ~ ) (3)

where K = diag(kl,...,km) > 0 is switching gain


and $i > 0 is a constant to adjust the boundary
layer thickness [4]. sat denotes a saturation function:
sat(si/&) = si/q5i when lsi/q5il < 1 or equivalently si
is inside the boundary layer, sat(si/&) = sgn(si/@i)l
when lsi/q5il 2 1 or equivalegtly si is outside the
boundary layer. Here K is chosen for STS < 0 to
hold outside the boundary layer:

The closed loop system composed by the plant and the


controller is transformed to an equivalent system represented in Figure 1when si is inside the boundary layer.
In the figure Ai indicates the multiplicative uncertainty
at the plant inputs for the i-th channel ( i.e. actuator delay ). Figure l implies that inside the boundary

2 The Sliding Mode Controller for Sluggish

Actuators
Consider the plant

k ( t ) = f ( z ( t ) ) Bu(t) y ( t ) , a(t) = Cs(t)

(1)

where s ( t ) E Rn is a state vector, u(t) E Rm is a


control, y ( t ) E Rn is a bounded signal represents uncertainty or disturbance, B E Rnxmis a constant m*
trix, f(z)is a map x E Rn --+ f(s)E Rn,and t d e
=
is a constant matrix,
notes time.
Ci E R I x m ,i is suffix which take a value from 1 to
m (i = 1,. ,m),a is a control error which should be
zero. We assume without loss of generality that CT is
chosen so that rank(CB) = m and the plant is minimum phase respect to the output Cx. Consider the
switching surface

[cT.

Figure 1: The equivalent closed loop system.


layer, the controller becomes PI, PID or PIDD... type
compensator k&Li(q)/&qDi(q), The control error a
is bounded for bounded f(s)and y when a single integrator l / q with the uncertainty Ai is stabilized by
the compensator. To design stabilizing compensator
for the system in Figure 1, we use the well-known frequency shaping concept which is popular in robust con-

trol theory [SI.


where Ni(q) and Di(q) are polynomial in differentiating
operator q such that Di(q) and M ( q ) = qDi(q)+Ni(q)
are Hurwitz and Ni(q)/qDi(q)
is strictly proper. In
0-7803-3590-2/96 $5.00 0 1996 IEEE

Outside the boundary layer, bang-bang control U =


-Ksgn(S), may bring si to the boundary layer very
fast.

2769

3 ABS Controller Design

Vehicle motions during braking are generally described


as follows [l,a]:

...

R*

G f M -

2ft-I
.. .

Air pressure
modulator

2% 3

Figure 2: Brake system configuration.


1

where Mu : vehicle mass, Jw, : wheel inertia for the


i-th wheel, R, : wheel radius, xu =-V,/R, here
V, is vehicle linear speed , z
,
, : angular speed of
the i-th wheel, U+ : brake torque for the i-th wheel,
pi(&) : adhesion coefficient between the i-th wheel
and road surface, fr,: rolling resistance for the i-th
wheel, faero: aerodynamic force, X i : slip ratio of the
i-th wheel (= {zu- zw,}/zu),F ! , : vertical load on
the i-th wheel, m : number of the wheels. Here f (z)
is highly non-linear and uncertain because the a d h e
sion coefficient pi(&) varia greatly with changes in
road surface conditions[2]. But f(z) is bounded b e
cause 0 < pLi(Xi)< 1, so are the term y and z for
q ( 0 ) 2 z u ( t ) 2 zw,(t) 2 0 is held during braking.
0 ) initial value of zu((t).The actuator
Here ~ ~ (means
delays (in our case approximately e-0~05S/(0.165s 1)
where s denotes Laplace operator) are treated as unmodeled dynamics in this study to simplify the model.

We chose

ai = (1 - Ao)z,

+ zwi

where A0 is reference slip ratio. It is easy to verify that


ai= 0 indicates Xi = Xo. It is known that pi(&) has a
peak value when Xi r? 0.2 in every road surface so we
set A0 = 0.2. A first-order switching surface was chosen
for less execution time and for less tuning parameters.
N I ,N2 axe set to 2 ~ for
5 front wheels, N3, N4 to 2 ~ 7
for rear wheels. The switching gain ki was set at the
value corresponding to the maximum available brake
torque. The wheel acceleration feedback technique was
also

usd[5].

4 Vehicle Test Results

The experiment was conducted on basalt surface and


ice using a medium-size truck ( M , = 3870 [kg]). For
wet basalt and ice, the pi values at wheel lock (p(1))are
approximately 0.16 and 0.07 respectively. The configuration of a brake system is shown in Figure 2. The air
pressure modulated for each wheel is converted to oil

0.8

0.6

* 0.4
0.2
0

10
Time [sec]

15

20

Figure 3: Test results on ice (p(1) = 0.07).


pressure by AOH (Air Over Hydraulic) boosters connected with brake cylinders of each wheels.
The test results are shown in Figure 3. The slip ratio
for a left rear wheel (i = 4) rose rapidly at braking
and is kept around the reference value of 0.2. Though
ununiformity at road surface should disturb the slip ra,
tio, the control error was small enough for a practical
use. These results verifies the effectiveness of the proposed method. The similar results were obtained for
the other wheels and also for the case p(1) = 0.16 [5].

References
Salman: A Robust Servo-Electronic Controller
for Brake Force Distribution, Trans. of ASME Dynamic
Systems, Measurement, and Control, 112 442/447
(1990)
[2] Tan and Chin: Vehicle Traction Control: Variable Structure Control Approach, Trans. of ASME
Dynamic Systems, Measurement, and Control, 113
223/230 (1991)
[3] Lyne, Passino and Yurkovich: Fuzzy Learning
Control for Antiskid Braking Systems, IEEE Trans. on
Control system Technoloo, 1-2 122/129 (1993)
[4] Slotine and Li, Applied Nonlinear Control, Prentice Hall (1991)
[5] T. Kawabe, M. Nakazawa, I. Notsu and Y.
Watanabe: A Sliding Mode Controller for Slip Ratio Control System, Proceedings of AVEC96 Internai
tional Symposium on Advanced Vehicle Control at the
Aachen University of Technology 2 797/804 (1996)
[el Doyle, Francis and Tannenbaum: Feedback Control Theory, McMillan (1992)

2770

[11

Das könnte Ihnen auch gefallen