WHEEL SLIP CONTROL IN ABS BRAKES USING GAIN SCHEDULED CONSTRAINED LQR

Idar PetersenW , Tor A. JohansenW*WW , Jens KalkkuhlWWW and Jens LdemannWWW

SINTEF Electronics and Cybernetics, N-7465 Trondheim, Norway. e-mail: Idar.Petersen@ecy.sintef.no Department of Engineering Cybernetics, Norwegian University of Science and Technology, N-7491 Trondheim, Norway. DaimlerChrysler, Research and Technology, Alt-Moabit 96a, D-10559 Berlin, Germany.
Keywords: Automotive control, gain scheduling, nonlinear systems, robustness, stability

Abstract
A wheel slip controller for ABS brakes is formulated using an explicit constrained LQR design. The controller gain matrices are designed and scheduled on the vehicle speed based on local linearizations. A Lyapunov function for the nonlinear control system is dervied using the Riccati equation solution in order to prove stability and robustness with respect to uncertainty in the road/tyre friction characteristic. Experimental results from a test vehicle with electromechanical brake actuators and brake-by-wire show that high performance and robustness are achieved.

Figure 1: Quarter car forces and torques. acceleration. The equations of motion of the quarter car are py @ I{ b M $ @ u I{ b where y $ I} I{ We u M We vljq+$, (1) (2)

1 Introduction
An anti-lock brake system (ABS) controls the slip of each wheel to prevent it from locking such that a high friction is achieved and steerability is maintained. ABS controllers are characterized by robust adaptive behaviour with respect to highly uncertain tyre characteristics and fast changing road surface properties [1]. The introduction of advanced functionality such as ESP (electronic stability program), drive-by-wire and more sophisticated actuators offers both new opportunities and requirements for higher performance in ABS brakes. The contribution of the present work is a study of a model-based design of ABS controllers, see also [2], [3]. We consider electromechanical actuators [4] rather than hydraulic actuators, which allow continuous adjustment of the clamping force. Here, we design the wheel slip controller based on a recently developed explicit LQR design method that takes into account input and state constraints [5], see also [6]. The wheel slip dynamics are highly nonlinear. Despite this, our control design relies on local linearization and gain-scheduling. In order to analyse the effects of this simplication, we develop a Lyapunov based stability and robustness analysis. Results from experiments using a test vehicle are also included. Other studies of model-based wheel slip control in ABS can be found in [7], [8], [9], [10].

horizontal speed at which the car travels angular speed of the wheel vertical force tyre friction force brake torque wheel radius wheel inertia

The tyre friction force I{ is given by I{ @ I} +> K > , where the friction coefcient  is a nonlinear function of  K  tyre slip friction coefcient between tyre and road slip angle of the wheel y y $u (3)

and the longitudial slip  dened by @ (4)

2 Modelling
In this section, we review a mathematical model of the wheel slip dynamics, see also [1], [8] and [7]. The problem of wheel slip control is best explained by looking at a quarter car model as shown in Figure 1. The model consists of a single wheel attached to a mass p. As the wheel rotates, driven by inertia of the mass p in the direction of the velocity y> a tyre reaction force I{ is generated by the friction between the tyre surface and the road surface. The tyre reaction force will generate a torque that initiates a rolling motion of the wheel causing an angular velocity $. A brake torque applied to the wheel will act against the spinning of the wheel causing a negative angular

 describes the normalised difference between horizontal speed y and the speed of the wheel perimeter $u. The slip value of  @ 3 characterises the free motion of the wheel where no friction force I{ is exerted. If the slip attains the value  @ 4> then the wheel is locked which means that it has come to a standstill. The friction coefcient  can span over a very wide range, but is differentiable with the property +3> K > , @ 3 and +> K > , A 3 for  A 3. Its qualitative dependence on slip  is shown in Figure 2. The friction coefcient  increases with slip  up to a value where it attains its maximum K .

=0

0.8

= 10

0.6

0.4

0.2

Proof. Note that +w, is a continuous trajectory. Hence, the possibile escape points are  @ 3 and  @ 4. Consider rst u b  @ 3. Since +3, @ 3, it follows from (5) that  @ yM We  3 due to We  3. Hence, +3,  3 implies +w,  3 for all w  3. Consider next  @ 4. Then, $ @ 3 and from (2) it follows that $  3 due to the discontinuity vljq+$, in (2). From (4), we b b conclude that   3 and +3,  4, which implies +w,  4 for all w  3. Finally, note that y  3 from (1) because I{  3 for b  5 ^3> 4`.
0.4 0.6 0.8 1

0
1 0.9 0.8 0.7 0.6

0.2

3 Control problem
The control problem is essentially to control the value of the longitudinal slip  to a given setpoint  that is either constant or commanded from a higher-level control system such as ESP (electronic stability program). The controller must be robust with respect to uncertainties in the tyre characteristic and variations in the road surface conditions. The control input is the clamping force Ie that is related to the brake torque as We @ ne Ie . A maximum or a minimum force can be applied to the brake pads by the actuator during braking. The (small) minimum is to ensure that the pads are positioned close to the brake disk (no air-gap). Maximum is what the actuator is capable of. There is also a rate limit at how fast the torque can be changed by the actuator. Integral action or adaptation must be incorporated to remove errors due to model inaccuracies, in particular the road/tyre friction coefcient K . It is essential that the controller maintains a high performance and is robust w.r.t. to any road/tyre friction curve that can be encountered cf. Figure 2.

H =0.95 H =0.8

()

0.5 0.4 0.3 0.2 0.1 0

H =0.5

H =0.3

H =0.1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

x (longitudinal slip)

Figure 2: Tyre slip/friction curves +> K > , For higher slip values, the friction coefcient will decrease to a smaller value where the wheel is locked and only the sliding friction will act on the wheel. The dependence of friction on the road condition is shown in the bottom part of Figure 2. For wet or icy roads, the maximum friction K is small and the right part of the curve is atter. The tyre friction curve will also depend on the brand of the tyre. In particular, for winter tyres, the curve will cease to have a pronounced maximum. If the motion of the wheel is extended to two dimensions, then the lateral slip of the tyre must also be considered. The slip angle  is where the wheel moves with velocity y{ in the longitudinal direction and with a velocity y| in the lateral direction. In this case, the longitudinal slip { @ y{ $u and the lateral y slip | @ vlq  are distinguished as well as the corresponding friction coefcients { and | . The top part of Figure 2 shows the dependence of the friction coefcient  on the side slip angel = The side force I| depends greatly on the side slip angle . For large side slip angles, the force gets smaller. In the sequel, for simplication purposes unless otherwise stated, the side slip angle will be considered to be zero with { @  and y{ @ y= Using (1)-(4) we get for y A 3 and $  3   5 b @ 4 4 +4 , . u I} +> K > , . 4 u We  y p M yM

4 Gain-scheduled LQRC controller design and analysis

The dynamics of the wheel and car body are given by (5) and (6), respectively. Due to large differences in inertia, the wheel dynamics and car dynamics will evolve on signicantly different time scales. The speed y will change much more slowly than the slip  and is therefore a natural candidate for gain scheduling. Thus, for the control design, we consider only (5) and regard y as a slow time-varying parameter. A constrained LQR controller [5] is applied. This requires a nominal linearized model for design. 4.1 Linearized dynamics a Let ( > We ) be an equilibrium point for (5) dened by the con stants  > I} and  , where  is the desired slip K   M a +4  , . u I} + >  >  , We @ K pu The linearized slip dynamics are given by 4 b + @ y  , . 4 +We y a We , . h.o.t. (7)

(5)

4 I} +> K > , (6) y@ b p Note that when y $ 3, the dynamics of the open loop system (from We to ) becomes innitely fast with innite gain. This leads to a loss of controllability and the slip controller must be switched off for small y. Proposition 1 Consider the system (5)-(6) with We +w,  3 for all w  3. If y+3, A 3 and +3, 5 ^3> 4`> then +w, 5 ^3> 4` and y+w,  3 for all w  3 where y+w, A 3= b

where 4 and 4 are linearization constants given by   4 +4  , . u5 C +>  > , 4 @ I} K p M C 4 . I} + >  >  , K p u 4 @ A 3 M

(8) (9)

The wheel slip dynamics (5) can be written in the form {5 @ b where !+{5 , @ .  !+{5 , 4 . +We y y  u M  5

namely  (10) S4>4 +y, @  5 5 . 4 U4 T5>5 +y, . 4

We ,

and {5 @ 

 and  M We @ +4 pu

u W M e

4 +4 p

{5 , .

I} +{5 .  > K > , (11)

  , . u I} + > K > ,

+T4>4 4 y 4@5 S4>5 +y, @ +T4>4 +y,U, 4 4 S5>5 +y, @ 5 4 y 4 U   4@5 4@5 + 5 5 4 . 4 U4 T5>5 +y, . 5+T4>4y,U, y 4 . y 5 4 4 U and the controller gains are N4 +y,@ N5 +y,@
4@5 T4>4 +y,U4   5 4 . 5 . 4 U4 T5>5 +y, . 4

T4>4 +y,U,4@5 y 4 4@5  +y,U,

5+

4@5

(12)

It can be seen that this system has an equilibrium point given by {5 @ 3, We @ We and !+3, @ 3, and the linearized slip model (7) with a perturbation term written as follows {5 @ b 4 4 {5 . +We y y 4 {5 .

5+

T4>4 +y,U,4@5 y 4

4@5

We , .

 +{5 , y

(13)

4

where  +{5 , @ !+{5 ,

4.2 Wheel slip control with integral action Let the system dynamics be augmented with a slip error integrator {4 @   @ {5 b     {4 b { @ D+y, {4 . E+y, +x We , . Z +y, +{5 , {5 b 5 (14) where D+y, @   > E+y, @   > Z +y, @  

We introduce the saturated control + pd{ pd{ We > x A We a plq plq x@ We > x ? We a x> a otherwise

(19)

pd{ plq for some We A We  3. Furthermore, we dene the error (note that both x and x depend on y and {) a

v +{> y, @ 4 +x

x, a

(20)

3 3

Since We is assumed unknown, we have dened x @ We , and the equilibrium point to be { @ +{ > 3,W and x @ We where 4 { is still unspecied and we get 4
{ @ D+y, +{ b { , . E+y, +x x , . Z +y, +{5 , (15) Next, dene the quadratic cost function for the purpose of local LQ design based on the nominal part of (15): ] R W M+{+w,> x^w> 4,, @ ++{+ , { , T+y, +{+ , { , w . +x+ , x ,W U +x+, x ,,g (16) Assuming constant y, the optimal control law is given by x @ N+y,{ a (17) where the gain matrix N+y, @ U4 E W +y,S +y, and we choose to neglect the unknown { and x which will be accounted for due to the integral action. The symmetric matrix S +y, A 3 is dened by the solution to the algebraic Riccati equation for the design: S +y,D+y, . DW +y,S +y, S +y,E+y,U4 E W +y,S +y, @ T+y,

4 4 y

3 4 y

3 4 y

With the denition +{> y, @ v +{> y, .  +{5 , and assuming { @ We @N4 +y,, the closed loop dynamics can be written as 4 b { @ +D+y, . E+y,N+y,, { . Z +y,+5 > y, { with { @ { { . (21)

Proposition 2 Consider the system (5) with controller (19). Assume U A 3 and the smooth matrix-valued function T satises T4>5 +y, @ T5>4 +y, @ 3> T4>4 +y, A 3> T5>5 +y, A 3> gT4>4 +y,  3> gT5>5 +y,  3 for all y A 3. Moreover, suppose gy gy S5>4 +y,F +4 , (22) T4>4 +y,A # y $ 5 +5 ,S5>5 +y,5 S5>4 +y,5 +5 , { {  { T5>5 +y,5A . {5 +4 , y yF (23) are satised for all y A 3, {5 5 ^  > 4  ` and some F A 3 pd{ A W A W plq , the equilibrium { @ 3 and  5 +3> 4,. If We e e is uniformly exponentially stable. Proof. Consider a neighbourhood [ 3 of the origin in U5 such that v +> y, @ 3 for all { 5 [ 3 and y A 3. Let a Lyapunov { function candidate be Y +, @ {W S +y, { { G+y,@S4>4 +y,S5>5 +y,
3 3

S4>5 +y,S5>4 +y, A 3

(24)

(18)

Its time-derivative along trajectories of (21)   CS +y, b bW b y { . { S { . {W S { b Y @ {W Cy is found by substituting for (15), (17) and (18) in (25): CS +y, b y {W T+y, b{ { Y @ {W Cy { { .  +5 ,+Z W +y,S +y, . {W S +y,Z +y,,

x 10

Left hand side

2.5 2

(25)

1.5

0.5

Right hand sides

-0.5

-1

0.1

0.2

0.3

0.4

0.5

(26)
2.5

0.6

0.7

0.8

0.9

(longitudinal slip)

x 10

From Proposition 1, it is clear that y  3. Thus, the negab   W CS +y, y { requires S Q +y, @ CS +y, A 3 for all tivity of { Cy b Cy y A 3= For S +y, to be positive denite, it is sufcient that 3 S4>4 +y, A 3 and G+y, @A 3 which follows due to (24). Note that since T4>4 +y,> T5>5 +y,> 4 A 3, it follows immediately 3 that S4>4 +y, A 3. Then, (26) becomes: b { { Y   +5 ,+Z W S +y, . {W S +y,Z +y,, 5 @ T4>4 +y,4 T5>5 +y,5 { {5 5 .  +5 ,+S5>5 +y,5 . S5>4 +y,4 , { { { y {W T+y, { (27)
3

Left hand side

2

1.5

0.5

Right hand sides

0

-0.5

-1

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

(longitudinal slip)

Figure 3: Illustration of the robust stability requirement, i.e. the left and right hand sides of eq. (23). The upper part is for y @ 4 m/s and the bottom part is for y @ 65 m/s.

To obtain all {4 -terms in (27) in a quadratic form, we apply Youngs inequality 5de  d5 @F . Fe5 . Hence, b Y  S5>4 +y, 5 {4 F T5>5 +y,5 {5 y 5 S5>4 +y, 5 +5 ,  { .  +5 ,S5>5 +y,5 . { { y y F T4>4 +y,5 . {4 b Y  taking S +y, from the solutions of the Riccati equation are possibly conservative. (28) The controller gain N+y, depends on the speed (gain scheduling). From a practical point of view, a useful gain schedule is gT +y, A 3 and gT5>5 +y, A 3, as this reachieved by letting 4>4 gy gy duces the gain as y $ 3. As mentioned earlier, this is necessary to avoid instability due to the unmodelled (actuator) dynamics since these tend to dominate as y $ 3. An important aspect is the initialization of the integrator state {4 . Note that {4 is a controller state that can be initialized arbitrarily, while { is unknown since it depends on the road/tyre 4 friction coefcient K . Hence, an intelligent initialization of {4 +3, based on any a priori information on K and possibly on the known {5 +3, may signicantly improve the transient per formance [2]. 4.3 An idealized design example We consider a design example with the following parameters p @ 783 nj, I} @ 7747 Q, u @ 3=65 p, M @ 4=3 nj p5 and the friction model in the right part of Figure 2. Assuming  @ 3=47 and the nominal design  @ 3=; and  @ 3, K we get 4 @ 43=5 and 4 @ 3=65. We choose U @ 4 and T+y, @ Ty6@5 with T4>4 @ 9 43< and T5>5 @ 73 439 . Note the scaling due to the different magnitudes of We and . The choice for T+y, leads to a gain schedule with reduced gain as y $ 3, that is useful to avoid instability due to unmodelled acatuator dynamics as y $ 3. Figure 3 shows that the robust stability requirement (23) is satised for all {5 5 ^  > 4  ` for all the friction curves in the right part of Figure 2. Although curves are shown only for y @ 4 m/s and y @ 65 m/s, we have veried that (24) is fullled for intermediate values of y. The control design is also satised the stability requirements (22) and (24), and we conclude robust uniform exponential stability of the equilibrium.

Due to (22) and (23) it follows that {4 T4>4 +y,5 T5>5 +y,5 {5

and we conclude that the equilibrium is uniformly exponentially stable using Corollary 4.2, [11]. Inequality (22) states that the error weight T4>4 +y, must be sufciently large, leading to a sufeciently large controller gain. Note that S5>4 +y, depends on T4>4 +y,> but it is evident that s S5>4 +y, increases with T4>4 +y, such that (22) will indeed hold for a sufciently large T4>4 +y,, except when y $ 3= Inequality (23) states that the error weight T5>5 +y, must also be sufciently large, leading to a sufciently high s to stabilize gain the system. Note that S5>5 +y, increases with T5>5 +y, such that this is also possible, except for y $ 3. { Note that T5>5 +y, m5 m is essentially chosen in (23) to dominate the perturbation +5 ,. Inequality (23) should be checked with { respect to the perturbations  that are generated by all possible friction curves + , to ensure robust stability. Inequality (24) can be seen to be non-restrictive since it will always be satised for 4 close to zero [3]. This corresponds to generating the nominal model by linearizing near the peak of the friction curve. Experience shows that high performance is indeed achieved this way. Note that for 4 @ 3, no information on the friction curves is actually utilized in the control design. The constant F A 3 should be chosen to minimize conservativeness. However, the choice T4>5 +y, @ T5>4 +y, @ 3 and

5 Implementation
0.9

slip (lambda)
1

The experimental vehicle is a Mercedes E220 equipped with electromechanical disc brake actuators and a brake-by-wire system. A previous version of this system is described in [4]. The system consists of four independent servo controllers, one for each brake, and a central electronic control unit (ECU) where the wheel slip controllers run. With reference to the theoretical study presented in previous sections, we note the following implementation details:  The actual control implementation is based on the discretization of the models and a discrete-time control design. The discrete-time model also takes into account communication delays on the brake-by-wire electronics. The sampling rate for the wheel slip control loop is 7 ms.  The electromechanical brake actuator is assumed to be linear, b  Ie @ Ie . Ie (29)

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

0
x 10
4

0.5

1.5

2.5

t, [s]
clamping force (Fb, N) from driver (dashed) and ABS (solid)
2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0

0.5

1.5

2.5

t, [s]

speed (v, m/s) and wheel speed (w r, m/s)

25

where  is the time-constant and Ie is the commanded clamping force. This is reasonable since it contains a local feedback loop. The model used for control design is augmented with a discretized version of (29) and the measured clamping force Ie is used for feedback.  The controller actually computes the change of the clamping force, rather than the clamping force itself. This is implemented in order to enjoy the benets of velocity-based gain scheduling [12], as well as it easily account for the actuator constraint (see [5]) b b pd{ b pd{  Ie  Ie Ie (30)

20

15

10

0.5

1.5

2.5

t, [s]

Figure 4: Experimental results with braking on dry asphalt. clamping force does not oscillate, we conclude that this is actually sensor noise that is known to increase at the speed goes to zero. We also note that the initial transient is not satisfactory as the claming force does not increase fast enough so that the slip is too low and the resulting friction force is too low in the interval w 5 +3=5> 3=:, leading to increased braking distance. This is due to the signicant model inaccuracy in the low-slip region, and redesign of the slip controller is necessary for this region. The second test, Figure 5, is braking on snow, starting at y+3, @ 55=3 m/s and without any steering manoeuvers. The slip setpoint is  @ 3=3: and we note that the regulation is satisfactory. There are some small oscillations that we believe are due to unmodeled actuator phenomenon (non-linearities) that are expected to be more pronounced when operating at low clamping force levels due to low friction on snow. The third test, Figure 6, is braking on a wet inhomogeneous surface (asphalt that is partially covered by a plastic coating and water) without any steering manoeuvers. The intial speed is y+3, @ 55=3 m/s and the slip setpoint is  @ 3=3<. We note that in this case there are signicant transients in the slip and claming force. We still conclude that the regulation performance is highly satisfactory, since the surface is characterized by very large variations in the friction coefcient that cause large disturbances on the system.

Anti-windup is implemented on the integrator state {4 .  Gain scheduling is implemented by computing gain matrices for a nite number of operating points and then switching gain matrices. To achieve bumpless transfer the integrator is reset at the switching instants. The gain matrices are scheduled on both speed and slip. The scheduling on slip only has an effect for very small slip values, and will improve the transient performance when the wheel slip controller is activated.  The slip  and speed y are estimated online using an extended Kalman lter based on wheel speed ($) and acceleration measurements.  The ABS system monitors the commands given by the pd{ @ driver using the brake pedal. Essentially, we set We where I is the clamping force commanded by the ne Ie e driver using the brake pedal. The slip controller is deactivated when the speed is below 1 m/s.  The tuning of the implemented controller is similar to the tuning in the idealized set-up in section 4.3. More specically, the dominant pole of the nominal linear closed loop is almost the same in both cases (about 12.0 for a typical y).

6 Experimental results
The rst test, Figure 4, is braking on dry asphalt, starting at y+3, @ 54=8 m/s and without any steering manoeuvers. The slip setpoint is  @ 3=3< and we note that the regulation is highly accurate and satisfactory. When the speed approaches zero, signicant variability in the slip emerges. Since the

7 Discussion and Conclusions

Using Lyapunov analysis and experimental verication we have shown good performance and robustness of a modelbased nonlinear wheel slip controller for ABS. In order to

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

slip (lambda)

0
4

0
x 10
4

t, [s]
2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 x 10

t, [s]
clamping force (Fb, N) from driver (dashed) and ABS (solid)
2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0

clamping force (Fb, N) from driver (dashed) and ABS (solid)

t, [s]

t, [s] speed (v, m/s) and wheel speed (w r, m/s)

25

speed (v, m/s) and wheel speed (w r, m/s)

25

20

20

15

15

10

10

t, [s]

t, [s]

Figure 5: Experimental results with braking on snow. achieve the robustness, the approach does not rely on knowledge of the tyre/road friction curve. Static uncertainty (due to unknown K ) is eliminated using integral action, while dynamic uncertainty (due to unknown shape of + ,) is handled by a robust design with sufcient stability margin. The results are somewhat prelimenary, and can be possibly be extended and improved. In particular, the analysis might take into consideration some of the details that are present in the implementation, but omitted from the current analysis. Also, the use of the Riccati-solution to dene the Lyapunov function is not always the best choice. Finally, means for improving the transient performance of the controller (especially for low slip) are under investigation. Although a detailed comparison with existing off-the-shelf ABS has not been conducted, the present results are encouraging, in particular when taking into account the modest time taken to design, tune and comission this model-based approach.

Figure 6: Experimental results with braking on a wet inhomogeneous suface.

J. Ldemann, Wheel slip control in ABS brakes using gain scheduled constrained LQR. http://www.itk.ntnu.no/ansatte/Johansen_Tor.Arne/ecc2001full.pdf, 2001. R. Schwarz, Rekonstruktion der Bremskraft Bei Fahrzeugen mit Elektromechanisch Bettigten Radbremsen. PhD thesis, Institut fr Automatisierungstechnik der TU Darmstadt, 1999. T. Johansen, I. Petersen, and O. Slupphaug, Explicit suboptimal linear quadratic regulation with state and input constraints, in IEEE Conference on Decision and Control, Sydney, 2000. A. Bemporad, M. Morari, V. Dua, and E. Pistikopoulos, The explicit solution of model predictive control via multiparametric quadratic programming, in Proc. American Control Conference, Chicago, pp. 872876, 2000. R. Freeman, Robust slip control for a single wheel, Tech. Rep. CCEC 95-0403, University of California, Santa Barbara, 1995. S. Drakunov, . zgner, P. Dix, and B. Ashra, ABS control using optimum search via sliding modes, IEEE Trans. Control Systems Technology, vol. 3, pp. 7985, March 1995. C. C. de Wit, R. Horowitz, and P.Tsiotras, Model-based observers for Tire/Road contact friction prediction, in In New Directions in Nonlinear Observer Design (H. Nijmeijer and T. Fossen, eds.), pp. 2342, Springer Verlag, May 1999. J. Yi, L. Alvarez, R. Horowitz, and C. C. de Wit, Adaptive emergency braking control using a dynamical Tire/Road friction model, in IEEE Conference on Decision and Control, Sydney, 2000. H. K. Khalil, Nonlinear Systems. Prentice Hall, 1996. I. Kaminer, A. M. Rascoal, P. P. Khargonekar, and E. E. Coleman, A velocity algorithm for the implementation of gainscheduled controllers, Automatica, vol. 31, pp. 11851191, 1995.

[4] [5] [6]

[7] [8] [9]

Acknowledgements
The work was sponsored by the European Commission under the ESPRIT LTR-project 28104 H5 C.

[10]

References
[1] M. Burckhardt, Fahrwerktechnik: Radschlupf-Regelsysteme. Wrzburg: Vogel Verlag, 1993. [2] T. A. Johansen, J. Kalkkuhl, J. Ldemann, and I. Petersen, Hybrid control strategies in ABS, in American Control Conference, (Washington D.C.), 2001. [3] I. Petersen, T. A. Johansen, J. Kalkkuhl, and [11] [12]