Professional Documents
Culture Documents
DETC2008-49304
ABSTRACT economy, and result in the loss of more than 40,000 lives
A new control algorithm and the adaptation laws required annually in the United States [1].
for estimation of unknown vehicle parameters have been In recent years, electronic control systems have found their
developed for vehicle stability control (VSC). This algorithm is way into automotive applications, resulting in significant
based on the Lyapunov Direct Method. A vehicle model with improvements in vehicle active safety. The active vehicle
two degrees of freedom (DOF) was used to develop the control control systems aim to improve the vehicle stability by
algorithm. In developing the equations of motion for this simple intervening when the vehicle is at the physical limit of its
model, a new approach for introducing the needed stabilizing performance. They assist the driver to keep control of the
forces and moments was developed. In addition, an eight DOF vehicle and at the same time provide the needed flexibility for
model was developed for control algorithm evaluation. The the vehicle to adapt to road variations. Various types of active
model includes lateral, longitudinal, yaw, and roll motions of control systems have been developed in the past to enhance the
the body plus the rotational DOFs for all of the four wheels. stability and handling characteristics of the vehicle [2–7].
Also included in the model is a transient tire model taking into Active front steering (AFS), active rear steering, four-wheel
account the tire lateral relaxation length. Using the validated 8 steering (4WS), active suspension, and vehicle stability control
DOF simulation model, the new control algorithm was (VSC) are some of the options widely explored. Electronic
evaluated and the results show the advantages of using such an control of vehicle dynamics has seen considerable effort in the
approach for enhancing vehicle stability during emergency integration of individual active control systems.
steering maneuvers. It has been known that vehicle stability
(understeer/oversteer) is influenced by tire normal load
variations during cornering [8]. This behavior can be
INTRODUCTION compensated for by use of active or semi-active suspension
Despite the fact that the United States has one of the best where the normal load variation on a tire can be controlled.
roadway systems in the world, congestion is increasing and The adaptive control algorithm developed here is based on
safety remains a serious problem. Congestion takes its toll in the Lyapunov Direct Method. A simple 2 DOF model has been
lost productivity, costing the nation an estimated $40 billion used to develop the control algorithm. This methodology could
each year. Trucks and buses travel over 100 billion miles be extended to higher order systems of vehicle equations of
annually and face the same congestion, delays, and inherent lost motion, linear or nonlinear.
productivity experienced by the daily commuter. Vehicle To evaluate the control strategy developed here, a nonlinear
crashes represent another $150 billion in financial burden to the 8 DOF vehicle model along with the Magic Formula tire model
M b ra ke Brake moment on wheel where the generalized forces Q’s are given as
F xi C i i (6)
u i rei
i
u (7)
Q r aF x 1 sin 1 aF y 1 cos 1
4
bF x 2 sin 2 bF y 2 cos 2 M zn
n 1
Q K (2)
v ar
i i front w heels
u
i i
u v br rear w heels
u
i
F r M
I p M i ,brake
i xi e i , drive (5)
2
shown in Figure 3 was used. The equations are modified and
simplified for the purpose of developing the control algorithm. (10)
U v and U r have been introduced in the right hand side of the
where x x x d . Choosing adaptation gain matrix to be a
equations to account for the counter force and moment inputs
required for improving vehicle stability. positive diagonal matrix, we know that the above candidate
Assuming small angles and considering lateral tire force as function is always positive. V ( x , t ) can be found by
a linear function of side slip angle, the equations of motion can differentiating V ( x , t ) with respect to time as shown in
be developed as follows:
equation (11).
C f C r bC r aC f
1
mu v V ( x , t ) x A x x A x p p p p x B x
T T T T T
m 0 v u u
2
0 I z r bC r aC f b C r a C f r
2 2
(11)
u u
For diagonal matrices A and , we have
C f f U v
aC f x Ax x Ax
T T
f U r
(9)
p p p p
T T
Considering x v , r
T
the equations are now converted (12)
into the standard state variable form shown in equation (8) Therefore,
where the matrices A, B, C, and U can be easily derived from
equation (9).
1
V ( x , t ) 2 x A x 2 p p x B x
T T T
2
x A x p p x B x
T T T
(13)
A x A x A x d
B x C U A x d (14)
V ( x , t ) x B x C U A x d p p x B x
T T T
x B x B x C U A x d p p
T T
x B x d C U A x d p p
T T
(15)
U A x d B x d C D x (16)
FIGURE 3- SIMPLIFIED TWO DOF VEHICLE MODEL
where D is control gain matrix.
To derive the control algorithm and the adaptation laws, we Substituting equation (16) into equation (15) and defining
consider the following candidate Lyapunov function: new terms we get
(17)
desired values of yaw rate and lateral velocity. Since the current
paper deals with yaw control, desired lateral velocity was
assumed to be the same as lateral velocity. The desired yaw rate
and was derived from the linear model and is given as:
A Aˆ A , B Bˆ B , C Cˆ C u
rd
u
2
where ^ indicates the estimated value and ~ is used for the error. a b 1
We define V ch a r (24)
Bx
C All simulation runs were done at a constant vehicle speed of 80
H p Ax d d (18) Km/h and step steering input of 80 degrees. In addition, the
control parameters used are defined below:
And substituting into equation (17) we get
V ch a r 5 7 .3
V ( x , t ) x H p D x p p
T T
(19) d ia g 1 0 e 9,1 0 e 9
Or, D d ia g 1 0 e 6,1 0 e 6
C f 0
T T
V ( x , t ) x D x p p H x
T
(20)
p ,
C r
pˆ ( t 0 )
0
Asymptotic stability is insured if V (x ,t) is negative The results are shown in Figures 4-8. Figure 4 indicates
that yaw rate is trying to follow the desired yaw rate for 80
definite. In order to have a negative definite V (x ,t) , we set the degrees step steering input.
second term in equation (20) equal to zero; therefore,
T
p p H x 0
T
(21)
p
1
H x
T
(22)
V ( x , t ) x D x 0
T
(23)