A Benchmark for Control of

Anti-lock Braking Systems

Stefan Solyom1, Anders Rantzer1 and Jens Kalkkuhl2

The following is a simplified model of wheel dynamics subject to brake torque

dω (t)
= α µ (λ (t)) − β Tb (t − τ )µ b(ω (t)), Tb ≥ 0
dt
dv(t)
= −γ µ (λ (t))
dt
v(t) − ω (t)r
λ ( t) =
v(t)

where:
ω - angular velocity of the wheel
v - velocity over ground of the car
λ - longitudinal tire slip
Tb - brake torque. It is the input signal of the model.
µ - road-tire friction coefficient. Dependence on the longitudinal tire slip (λ ) for
four different surfaces is shown in Figure 1.
µ b - friction coefficient in the brakes. This is used in the modeling of wheel lock.
For simplicity use µ b = min(ω /ε , 1), for some small ε > 0.
τ - time delay (an appropriate value is 14 miliseconds)
r - wheel radius (a suitable value for a passenger vehicle is 0 .3 meters)
α , β , γ - positive constants, resulting from physical parameters of the vehicle
(appropriate values for a passenger vehicle are respectively 1500, 1 and
10).
The following signals are available from the plant:
• ω - angular velocity of the wheel
• v - velocity over ground of the car
• µ H - maximum road-tire friction coefficient
These are estimated or measured and available for feed-back in the control al-
gorithm. In this benchmark, all of these signals can be considered measurable.

1
Specifications
The control objective is to maintain a desired tire slip level λ , by adjusting the
brake torque (Tb). The ABS has to fulfill the following requirements [8]:
• no wheel lock allowed to occur for speeds above 4 ms
• wheel lock for a period of less than 0.2 seconds is allowed for speeds in the
range of 0.8 . . . 4 ms
• the control system should be robust with respect to other unmodeled dy-
namics, e.g. actuator dynamics. A reasonable model of the actuator dynam-
ics, that can be used in the robust design is:

.0091s + 3.9545
.
0.0001s2 + 0.0402s + 3.9545

It is of special importance that the above mentioned specifications are fulfiled

0.9
Dry asphalt

0.8

0.7
firction coefficient (µ)

0.6 Wet asphalt

0.5

0.4

Snow
0.3

0.2

Ice
0.1

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
tire slip (λ)

Figure 1 Tire friction curves

2
Comments on the model
The benchmark model is derived from the well known quarter car model [6], [4].
This consists of a single wheel attached to a mass, as shown in Figure 2.

Fz

v
ω
Tb
Fx

Figure 2 Quarter car

The benchmark describes the equations of motion of the quarter car in case
of braking. The constants given in the model have the following physical inter-
pretation:
rFz 1 Fz
α = , β = , γ =
J J m
where,
m - mass of the quarter car
Fz - vertical force
r - wheel radius
J - wheel inertia
An additional time delay (τ ) is added due to sampling and communication
between the different modules of the system.
The longitudinal tire slip (λ ) definition will imply that a locked wheel (ω = 0)
is described by λ = 1, while the free motion of the wheel (ω r = v) is described
by λ = 0.
The tire friction force, is determined by Fz µ (λ , µ H , α , Fz, v) where µ (λ , µ H , α , Fz)
is the road-tire friction coefficient. This is a nonlinear function with a typical de-
pendence on the slip shown in Figure 1. The most common tire friction model
used in the literature is the “Magic Formula” [1], or Pacejka model. This model
uses static maps to describe dependence between slip and friction and it can
depend on the vehicles velocity (v). This function depends also on the normal
force ( Fz), steering angle (α ), road surface ( having different maximum values
µ H for different road conditions). For ease of writing, the model equations high-
light only the dependence on the longitudinal tire slip (λ ). In Figure 1 there
are shown tire friction curves, generated by the Pacejka model, for four different
kind of surfaces.
Notice that this model contains a quite simple description of the slip dynamics
for a wheel. It does not capture pitching motion of the car body while braking,
suspension dynamics, actuator dynamics, tire dynamics nor camber angle (in
the above given model, the tire is consider perpendicular on the road surface).
However, it captures the major control challenges of the problem.

3
1. References
[1] E. Bakker, H. B. Pacejka, and L. Lidner. “A new tire model with application
in vehicle dynamics studies.” SAE, 890087, 1989.
[2] P. A. Bliman, T. Bonald, and M. Sorine. “Hysteresis Operators and Tire Fric-
tion Models: Application to vehicle simulation.” In Proceedings of ICIAM’95,
Hamburg, Germany, 1995.
[3] C. Canudas de Wit and P. Tsiotras. “Dynamic Tire Friction Models.” In
Proceedings of the IEEE Conference on Decision and Control, Phoenix,
pp. 3746–3751, 1999.
[4] S. Drakunov, Ü. Özgüner, P. Dix, and B. Ashrafi. “ABS control using
optimum search via sliding mode.” In IEEE Transactions on Control System
Technology, vol. 3, pp. 79–85, 1995.
[5] F. Gustafsson. “Slip-based Tire-road Friction estimation.” Automatica, 33:6,
pp. 1087– 1099, 1997.
[6] T. Johansen, J. Kalkkuhl, J. Lüdemann, and I. Petersen. “Hybrid Control
Strategies in ABS.” Proceedings of the American Control Conference, Arling-
ton, 2001.
[7] Y. Liu and J. Sun. “Target slip tracking using gain-scheduling for braking
systems.” In Proceedings of the American Control Conference, Seatle, Wash-
ington, pp. 1178–1182, 1998.
[8] S. Solyom. “Synthesis of a model-based tire slip controller.” Technical Report
Licentiate thesis ISRN LUTFD2/TFRT–3228–SE, Department of Automatic
Control, Lund Institute of Technology, Sweden, June 2002.
[9] P. Wellstead and N. Pettit. “Analysis and redesign of an antilock brake system
controller.” In IEE Proceedings - Control Theory Appl., vol. 144, pp. 413–425,
1997.