Shock and Vibration 15 (2008) 493–503 493

IOS Press

Optimal control of an 8-DOF vehicle active

suspension system using Kalman observer
S. Hossein Sadati, Salar Malekzadeh ∗ and Masood Ghasemi
Department of Mechanical Engineering, K.N.Toosi University of Technology, 4th Tehranpars square, Tehran, Iran

Received 6 December 2006

Abstract. In this paper, an 8-DOF model including driver seat dynamics, subjected to random road disturbances is used in order
to investigate the advantage of active over conventional passive suspension system. Force actuators are mounted parallel to the
body suspensions and the driver seat suspension. An optimal control approach is taken in the active suspension used in the
vehicle. The performance index for the optimal control design is a quantification of both ride comfort and road handling. To
simulate the real road profile condition, stochastic inputs are applied. Due to practical limitations, not all the states of the system
required for the state-feedback controller are measurable, and hence must be estimated with an observer. In this paper, to have the
best estimation, an optimal Kalman observer is used. The simulation results indicate that an optimal observer-based controller
causes both excellent ride comfort and road handling characteristics.

Keywords: Active suspension system, 8-DOF model, optimal control, passenger dynamics, ride comfort, Kalman observer, road
disturbance

1. Introduction

Demands for better ride comfort and controllablity of road vhicle motivate many automotive industries to consider
the use of active suspension. The active suspension system made it possible to surpass the compromise between ride
comfort and vehicle stability, which was not possible with passive suspension [9]. The idea of active suspension is to
insert a force actuator between the sprung and unsprung masses in addition to the spring and the damper, replacing
the damper or replacing both spring and the damper. The advantage of the active suspension is due to its capability
to control the attitude of the vehicle, to reducethe effects of braking and to reduce the vehicle roll during cornering
maneuvers in addition to increase the ride comfort and vehicle road handling.
The actuator is secured in parallel with a spring and shock absorber. Active suspension requires sensors to be
located at different points of the vehicle to measure the motions of the body, suspension system and/or the unsprung
mass. This information is used in the online controller to cammand the actuators in order to provide the exact amount
of force required [2,3,7].
In any vehicle suspension system, there are a veriety of performance parameters which need to be optimized.
These parameters include, among others, four important quantities which should be considered carefully in designing
a suspension system: ride comfort, body motion, road handling, and suspension travel. The trade-off between ride
comfort and road handling characteristics is usually a trial and error procedure. Moreover, no suspension system
can simultaneously minimize all four of the above mentioned parameters. The advantage of controlled suspension
is that a better set of design trade off is possible rather than with passive systems. Linear optimal control theory
provides a systematic approach to design the active suspension controller [5].
In this study, we intend to emphasize the effective methodogy of controller design in order to satisfy a preasigned
set of design criteria.

∗ Corresponding author. Tel.: +98 911 3329757; Fax: +98 21 8882997; E-mail: s [email protected].

ISSN 1070-9622/08/$17.00  2008 – IOS Press and the authors. All rights reserved
494 S.H. Sadati et al. / Optimal control of an 8-DOF vehicle active suspension system using Kalman observer

This paper is arranged as follows: the full 8-DOF model of a vehicle, including passenger dynamics, is first
presented followed by a discussion of the proposed control strategy. A simulation is then made in order to apply the
proposed control on the system. Hence, we continue with an exposition of an 8-DOF model of a vehicle including
passenger dynamics.
This paper illustrates the importance of passenger dynamics when the objective would be to improve the passenger
ride comfort as well as to satisfy the road handling requirements.

2. Mathematical modeling

Mathematical modeling of vehicle, including the passenger dynamics and road disturbance is presented in this
section. A linear model is considered to represent the vehicle-passenger dynamics, while a normal random profile
is used to model the road roughness.

2.1. Vehicle-passenger model

The dynamical 8 degrees of freedom model of the vehicle with a driver is presented in Fig. 1. The model consists
of a body (sprung mass), connected by the suspension systems to four wheels (unsprung masses) and a passenger.
Body motions are considered to be bounce, roll, and pitch with every wheel having its own bounce. The passenger is
considered to have only vertical oscillations. The suspension, tires and passenger seat are modelled by linear springs
in parallel to viscous dampers. The system variable notations with their corresponding values are presented in the
nomenclatures at the end of the paper.

Fig. 1. Linear model of vehicle, passenger and road with 8 degrees of freedom.

The values given for the different variables and parameters in the nomenclature are related to a real Samand
vehicle. This vehicle is considered as the “National Vehicle of Iran” whose manufacture began around 2003.
The parameters related to the tires are denoted with subscript t, while the passenger parameters have subscript
d. The eight degrees of freedom are: front right wheel displacement, z 1 ; front left wheel displacement, z 2 ; rear
left wheel displacement, z 3 ; rear right wheel displacement, z 4 ; body bounce, z b ; body roll, θ r ; body pitch, θ p and
passenger bounce, z d [1].
The actuators are considered to be a source of controllable force, and located in parallel to the suspension spring
and shock absorber, passenger seat suspension.
The equation of motion are linear and can be written in matrix form as [12]
ẋ = Ax + Bu + Gw (1)
 S.H. Sadati et al. / Optimal control of an 8-DOF vehicle active suspension system using Kalman observer 495

where the state vector x is defined as

x = [z1 z2 z3 z4 zb θp θr zd ż1 ż2 ż3 ż4 żb θ̇p θ̇r żd ]T (2)
The input vector u represents the five actuator forces, while the disturbance vector w consists of road disturbances.
u = [u1 u2 u3 u4 u5 ]T (3)

w = [zf r zf l zrl zrr ]T (4)

The matrix reprentation of Eq. (1) would be the basis for the linear optimal controller design.

2.2. Road profile model

In the early days of studying the performance of vehicles on rough roads, functions such as sine waves, step
functions or triangular waves were generally applied as disturbances from the ground. While these inputs provide a
basic idea for comparative evaluation of designs, it is recognized that the road surface is usually not represented by
these simple functions, and therefore the deterministic irregular shapes cannot serve as a valid actual road disturbance
for studying the true behavior of the vehicle.
In this paper, a real road surface taken as a random exciting function, is used as the input to the vehicle-road
model. Power spectral density (PSD) analysis is used to describe the basic properties of random data. Several
attempts have been made to classify the roughness of a road surface. In this study, classifications are based on the
international organization for standardization(ISO). The ISO has proposed road roughness classification using the
PSD values(ISO, 1982).
Table 1 illustrates the stochastic characteristics of the final random input design, which corresponds to the poor
road condition as being classified by ISO.

Table 1
Stochastic characteristic of road random disturbance
Min(m) Max(m) PSD(m3 /cycle)
−0.029 0.029 0.0004

3. Optimal controller design

The performance characteristics which are of most interest when designing the vehicle suspension are passenger
ride comfort, body motion, suspension travel and road handling [5].
The linear time-invariant system (LTI) is described by Eq. (1). To design the controller, it is assumed that all the
states are available and that they can be measured exactly. First, let us consider a state variable feedback regulator
of the type
u = −k x (5)
where k is the state feedback gain matrix. The optimization procedure consists of determining the control input
u, which minimizes the performance index. The performance index J represents the performance characteristic
requirement as well as the controller input limitations.
In the conventional method, the performance index J penalizes the state variables and the inputs, thus it has the
standard form of

∞
J = (xT Q x + uT R u) dt (6)
0

where Q and R are posotive definite, known as the weighting matrices.

496 S.H. Sadati et al. / Optimal control of an 8-DOF vehicle active suspension system using Kalman observer

Using linear optimal control theory [11] for the controller design, the performance index can be written as a
weighted sum of the mean square values of output performance variables, including suspension relative travel,
passenger relative bounce, passenger acceleration, and body pitch [6]. The passenger acceleration is used here as an
indicator of ride comfort. The controller should minimize all the quantities.
The performance index written in terms of the variables used in this study can be written as

∞
2
J= [ q1 (zb1 −z1 )2 +q2 (zb2 −z2 )2 + q3 (zb3 −z3 )2 +q4 (zb4 −z4 ) +q5 (zd −zb5 )2 +q6 θp2 +q7 .z̈d2 ] dt (7)
0

where (zb1 − z1 ), (zb2 − z2 ), (zb3 − z3 ) and (zb4 − z4 ) are related to the suspension relative travel, and (z d − zb5 )
is the driver relative bounce.
Changing Eq. (7) into a general matrix format, the optimal matrix will be
 
ee1 0 0 0 e15 e16 e17 0 0 0 0 0 0 0 0 0
 0 ee2 0 0 e25 e26 e27 0 0 0 0 0 0 0 0 0 
 
 0 0 ee3 0 e35 e36 e37 0 0 0 0 0 0 0 0 0 
 
 0 0 0 ee4 e45 e46 e47 0 0 0 0 0 0 0 0 0 
 
 e15 e25 e35 e45 ee5 e56 e57 e58 0 0 0 0 e513 e514 e515 e516 
 
 e16 e26 e36 e46 e56 ee6 e67 e68 0 0 0 0 e613 e614 e615 e616 
 
 e17 e27 e37 e47 e57 e67 ee7 e78 0 0 0 0 e713 e614 e715 e716 
 
 0 0 0 0 e58 e68 e78 ee8 0 0 0 0 e813 e614 e815 e816 
Q=  
 0 0 0 0 0 0 0 0 ee9 0 0 0 0 0 0 0  
 0 0 0 0 0 0 0 0 0 ee10 0 0 0 0 0 0 
 
 0 0 0 0 0 0 0 0 0 0 ee11 0 0 0 0 0 
 
 0 0 0 0 0 0 0 0 0 0 0 ee12 0 0 0 0 
 
 0 0 0 0 e513 e613 e713 e813 0 0 0 0 ee13 e1314 e1315 e1316 
 
 0 0 0 0 e514 e613 e713 e813 0 0 0 0 e1314 ee14 e1415 e1416 
 
 0 0 0 0 e515 e613 e713 e813 0 0 0 0 e1315 e1415 ee15 e1516 
0 0 0 0 e516 e613 e713 e813 0 0 0 0 e1316 e1416 e1516 ee16

 
R1 0 0 0 0
 0 R2 0 0 0 
 
R= 
 0 0 R3 0 0 
 0 0 0 R4 0 
0 0 0 0 R5

Linear optimal control theory [11] provides the controller gain of Eq. (5) using the performance index relation of
Eq. (6). This procedure results in the gain matrix k as

k = −R−1 BP (8)

where the matrix P is optained by solving the algebraic Riccati equation

AT P + P A − P BR−1 B T P + Q = 0 (9)

Equation (1) for the optimal closed-loop system may be written in the form

ẋ = (A − BK)x + Gw (10)

Which is the form ultimately used in the simulation performed.

 S.H. Sadati et al. / Optimal control of an 8-DOF vehicle active suspension system using Kalman observer 497

4. Optimal observer design

We have previously assumed that the state vector is available for use in the optimal state feedback law. It might
appear that the state feedback method is not applicable in practice, since the whole state is not available. However, it
is possible to estimate the states of the system, provided the system is detectable. In order to be able to provide the
necessary states used in the feedback law, a reduced order state-space observer is designed to estimate those states
that are not directly measurable. It is possible to recover much of the behavior of the state feedback law by using
state estimates instead of states in the feedback law.
The input to the state observer is the system output measurment y, which generally could be expressed in the form
y = Cx + v (11)
where x is the system states and v is the measurement noise. A typical sensor arrangement for passenger cars can
provide the relative suspension travel and the relative passenger bounce. For this case we have
 
1 0 0 0 −1 b2 a1 0 0 0 0 0 0 0 0 0
 0 1 0 0 −1 −b1 a1 0 0 0 0 0 0 0 0 0 
 
C=  0 0 1 0 −1 b2 −a2 0 0 0 0 0 0 0 0 0 
 (12)
 0 0 0 1 −1 −b1 −a2 0 0 0 0 0 0 0 0 0 
0 0 0 0 1 bd −ad −1 0 0 0 0 0 0 0 0
The observer structure is in the form of
x̂˙ = Ax̂ + Bu + L(y − C x̂) (13)
where L is the optimal observer gain matrix that produces an LQG optimal estimate of x, denoted as x̂ [6]. The
optimal gain matrix L can be evaluated using the relation
L = P C T V −1 (14)
where P is the positive definite solution of the Riccati equation given by
AT P + P A − P C T V −1 CP + W = 0 (15)
In this equation, W = E[ww ], V = E[vv ] are the plant distribution and measurement noise covariances and
T T

E stands for the usual statistical expectation. Notice that in this equation, x̂ is a function of the output y and input u,
but the input u itself is generated by a state feedback law as
u = −k x̂ (16)
Therefore, for simulation purposes, we can augment the system dynamic equation (Eq. (1)) with that of the
observer (Eq. (13)), and solve the augmented system altogether as illustrated by the following equation.
     
ẋ A −BK ẋ x G0 w
= + + (17)
x̂˙ LC A − BK − LC x̂˙ x̂ 0L v

5. Simulation results

In this section the performance of Kalman optimal observer in estimating system states is demonstrated. Also,
the effects of active system to remedy the drawbacks of the passive system are examined. The simulation is done in
Matlab 7.1 software.

5.1. Optimal Kalman observer performance

Figure 2 shows the comparison made between the first four system states and their estimations. These states are
front right tire bounce, front left tire bounce, rear right tire bounce, and rear left tire bounce. Also, Fig. 4 shows
the comparison made between the second four system states and their estimations. These states are body roll, body
pitch, body bounce, and driver bounce. To better detect the observer performance, the error related to the observer
estimation is shown in Figs 3 and 5. As it is clear in Figs 2 to 5, despite the random input, the estimations are quite
close to the true system states and this performance is achieved because the Kalman observer works as an optimal
estimator.
498 S.H. Sadati et al. / Optimal control of an 8-DOF vehicle active suspension system using Kalman observer

Fig. 2. Comparison of real and estimated states for the front and rear suspension bounces.

Fig. 3. Error of the estimation for the first four states.

5.2. Comparison between active and passive system states

Figure 6 shows the comparison made between the passive and active system states including front tires bounce,
rear tires bounce, body bounce, and driver bounce. Also, in Fig. 7, a comparison made between body pitch and
driver acceleration from passive and active systems is presented.
From Fig. 6, it is clear that front and rear tire bounces in the active system are smaller than those of the passive
system. Also, body pitch and driver bounce are reduced. In Fig. 7, it can be observed that these two states are
reduced considerably.
It is also clear that the tire deflecion is a factor of road handling. Moreover, body pitch and driver acceleration
are important criteria for the ride comfort. It is therefore concluded that the active system retains better ride comfort
characteristics as compared with the passive system.
 S.H. Sadati et al. / Optimal control of an 8-DOF vehicle active suspension system using Kalman observer 499

Fig. 4. Comparison of real and estimated states including body roll, body pitch, body bounce, and driver bounce.

Fig. 5. Error of the estimation for the second four states.

6. Conclusions

The objective of this paper is to examine the use of optimal state-feedback controller to improve the ride comfort
in a passenger car. In this paper, the driver seat was included in the full vehicle, so that the response of the driver
due to road disturbances can be observed. The potential for improved vehicle ride comfort and road handling that
can be achieved from the controlled actuator forces is examined.
The performance characteristics of such a suspension system is evaluated and compared with that of a passive
suspension system.
The Kalman observer used to estimate the system unmeasurable states showed a good performance due to the fact
that Kalman observer is an optimal estimator in nature.
500 S.H. Sadati et al. / Optimal control of an 8-DOF vehicle active suspension system using Kalman observer

Fig. 6. Comparison between passive and active systems for the front and rear tires bounce, body bounce, and driver bounce.

Fig. 7. Comparison between passive and active systems for the body pitch and driver acceleration.

The suspension designs which have emerged from the use of optimal state-feedback control theory proved to be
effective in controlling the vehicle vibrations and achieving a better performance as compared with the conventional
passive suspension.

Acknowledgment

We would like to express our gratitude to Dr. Shahram Azadi from the Dept. of Mechanical Engineering at the
K.N.Toosi University of Technology for his sincere assistance throughout this research work.

Nomenclature and data

The definition of the nomenclature and their actual values taken from a real Samand vehicle is followed in Table 2.
This vehicle is considered as the “National Vehicle of Iran” whose manufacture began around 2003.
 S.H. Sadati et al. / Optimal control of an 8-DOF vehicle active suspension system using Kalman observer 501

Table 2
Nomenclature and data
Not. Discription Value
Ixx Body inertia around x 398.4 kg.m2
Iyy Body inertia around y 1872.4 kg.m2
m1 Front right suspension mass 95 kg
m2 Front left suspension mass 95 kg
m3 Rear left suspension mass 90 kg
m4 Rear right suspension mass 90 kg
mb Body mass 1161.9 kg
md Passenger mass 90 kg
a1 Dimension 1039.5 mm
a2 Dimension 1635.1 mm
b1 Dimension 724 mm
b2 Dimension 724 mm
ad Dimension 650 mm
bd Dimension 500 mm
kf r Front right suspension stiffness 20040 N/m
kf l Front left suspension stiffness 20040 N/m
krl Rear left suspension stiffness 24960 N/m
krr Rear right suspension stiffness 24960 N/m
kd Passenger seat stiffness 16000 N/m
ktf Front tires stiffness 177500 N/m
ktr Rear tires stiffness 177500 N/m
kar Antiroll bar stiffness 43000 N/m
cf r Front right damping 965 Ns/m
cf l Front left damping 965 Ns/m
crl Rear left damping 4000 Ns/m
crr Rear right damping 4000 Ns/m

Appendix A: Equations of motion for the 8-DOF vehicle

Using the Lagrange method, the dynamic equations of motion for the different coordinates of the 8-DOF vehicle
have been derived and presented in their final simplified forms here in Appendix A.
Front right tire:
m1 z̈1 + kf (z1 − zb + b2 θr + a1 θp ) − ktf (zf r − z1 )
(A1)
+kar [(z1 − z2 )/(b1 + b2 )2 + θr /(b1 + b2 )] + cf (ż1 − żb + b2 θ̇r + a1 θ̇p ) = −u1
Front left tire:
m2 z̈2 + kf (z2 − zb − b1 θr + a1 θp ) − ktf (zf l − z2 ) − kar [(z1 − z2 )/(b1 + b2 )2 + θr /(b1 + b2 )]
(A2)
+cf (ż2 − żb − b2 θ̇r + a1 θ̇p ) = −u2
Rear right tire:
m3 z̈3 + kr (z3 − zb − b1 θr − a2 θp ) − ktr (zrl − z3 ) + cr (ż3 − żb − b1 θ̇r − a2 θ̇p ) = −u3 (A3)
Rear left tire:
m4 z̈4 + kr (z4 − zb − a2 θp + b2 θr ) − ktr (zrr − z4 ) + cr (ż4 − żb − a2 θ̇p + b2 θ̇r ) = −u4 (A4)
Body bounce:
mb z̈b − kf (z1 − zb + b2 θr + a1 θp ) − kf (z2 − zb − b1 θr + a1 θp ) − kr (z3 − zb − b1 θr − a2 θp )
−kr (z4 − zb − a2 θp + b2 θr ) + kd (zb − ad θp + bd θr − zd ) − cf (ż1 − żb + b2 θ̇r + a1 θ̇p )
(A5)
−cf (ż2 − żb − b1 θ̇r + a1 θ̇p ) − cr (ż3 − żb − b1 θ̇r − a2 θ̇p ) − cr (ż4 − żb − a2 θ̇p + b2 θ̇r )
+cd (żb − ad θ̇p + bd θ̇r − żd ) = u1 + u2 + u3 + u4 − u5
502 S.H. Sadati et al. / Optimal control of an 8-DOF vehicle active suspension system using Kalman observer

Body roll:
Ixx θ̈r + b2 kf (z1 − zb + b2 θr + a1 θp ) − b1 kf (z2 − zb − b1 θr + a1 θp ) + b1 kr (z3 − zb + b2 θr − a2 θp )
−b1 kr (z4 − zb − a2 θp − b1 θr ) + bd kd (zb − ad θp + bd θr − zd ) + kar (θr + zb11 −z +b2 )
2

(A6)
+b2 cf (ż1 − zb + b2 θ̇r + a1 θ̇p ) − b1 cf (ż2 − żb − b1 θ̇r + a1 θ̇p ) − b2 cr (ż3 − żb + b2 θ̇r − a2 θ̇p )
−b1 cr (ż4 − żb − b1 θ̇r + a2 θ̇r ) + bd cd (żb − ad θ̇p + bd θ̇r − żd ) = −b2 u1 + b1 u2 − b2 u3 + b1 u4 − bd u5
Body pitch:
Iyy θ̈p + a1 kf (z1 − zb + b2 θr + a1 θp ) + a1 kf (z2 − zb − b1 θr + a1 θp ) − a2 kr (z3 − zb − b1 θr − a2 θp )
−a2 kr (z4 − zb − a2 θp + b2 θr ) − ad kd (zb − ad θp + bd θr − zd ) + a1 cf (ż1 − żb + b2 θ̇r + a1 θ̇p )
(A7)
+a1 cf (ż2 − żb − b1 θ̇r + a1 θ̇p ) − a2 cr (ż3 − żb − b1 θ̇r − a2 θ̇p ) − a2 cr (ż4 − żb − a2 θ̇p + b2 θ̇r )
−ad cd (żb − ad θ̇p + bd θ̇r − żd ) = −a1 u1 − a1 u2 + a2 u3 + a2 u4 + ad u5
Driver bounce:
md z̈d − kd (zb − ad θp + bd θr − zd ) + cd (żb − ad θ̇p + bd θ̇r − żd ) = u5 (A8)

Appendix B: The elements of the Q matrix used in the optimal control method

r5 = A(16, 5), r6 = A(16, 6), r7 = A(16, 7), r8 = A(16, 8), r13 = A(16, 13),
r14 = A(16, 14), r15 = A(16, 15), r16 = A(16, 16), rd = 1/md
ee1 = q1 , ee2 = q2 , ee3 = q3 , ee4 = q4 , ee5 = q1 + q2 + q3 + q4 + q5 + r52 .q7
ee6 = b22 q1 + b21 q2 + b22 q3 + b21 q4 + b2d q5 + r62 .q7
ee7 = a21 q1 + a21 q2 + a22 q3 + a22 q4 + a2d q5 + q6 + r62 .q7
ee8 = q5 + r82 .q7 , ee13 = r13 2
.q7 , ee14 = r142 2
.q7 , ee15 = r15 2
.q7 , ee16 = r16 .q7
e15 = − q1 , e16 = b2 q1 , e17 = a1 q1 , e25 = − q2 , e26 = −b1 q2 , e27 = a1 q2
e35 = − q3 , e36 = b2 q3 , e37 = −a2 q3 , e45 = − q4 , e46 = −b1 q4 , e47 = −a2 q4
e56 = −b2 q1 + b1 q2 − b2 q3 + b1 q4 + bd q5 + r5 .r6 .q7 e57 = −a1 q1 − a1 q2 + a2 q3 + a2 q4 + ad q5 + r5 .r7 .q7
e58 = −q5 + r5 .r8 .q7 , e513 = r5 .r13 .q7 , e514 = r5 .r14 .q7 , e515 = r5 .r15 .q7 , e516 = r5 .r16 .q7
e67 = a1 b2 q1 − a1 b1 q2 + a2 b2 q3 + a2 b1 q4 − ad bd q5 + r6 .r7 .q7
e68 = −bd q5 + r6 .r8 .q7 , e613 = r6 .r13 .q7 , e614 = r6 .r14 .q7 , e615 = r6 .r15 .q7 , e616 = r6 .r16 .q7 ,
e78 = ad q5 + r7 .r8 .q7 , e713 = r7 .r13 .q7 , e714 = r7 .r14 .q7 , e715 = r7 .r15 .q7 , e716 = r7 .r16 .q7 ,
e813 = r8 .r13 .q7 , e814 = r8 .r14 .q7 , e815 = r8 .r14 .q7 , e816 = r8 .r16 .q7
e1314 = r13 .r14 .q7 , e1315 = r13 .r15 .q7 , e1316 = r13 .r16 .q7
e1415 = r13 .r15 .q7 , e1416 = r13 .r16 .q7 , e1516 = r13 .r14 .q7

References

[1] S. Malekzadeh, Analytical Scrutiny and Computer Simulation of Vehicle in Response to road Ramps, B.S. Thesis, Tabriz Universiy, Tabriz,
Iran, 2004.
[2] K. Hyo-Jun, S.H. Yang and P. Young-Pil, Improving the vehicle performance with active suspension using road-sensing algorithm, Elsevier,
Computers and Structures (2002), 80-1569–1577.
[3] K. Jeong-Hoon and L. Chong-Won, Semi-active damping control of suspension systems for specified operational response mode, Journal
of Sound and Vibration 260 (2003), 307–328.
[4] T.D. Gillespie, Fundamentals of vehicle Dynamics, SAE Inc, 1992.
[5] H.D. Taghirad and E. Esmailzadeh, Passaenger ride comfort through observer based control, In proceeding of the 15th ASME biennial
conference on Mechanical Vibration and Noise, 1995.
[6] F. Yu, J.-W. Zhang and D.A. Crolla, A study of kalman filter active vehicle suspension system using correlation of front and rear wheel
road inputs, Proceeding institute Mechnical Engineering 214 (2000), part D.
[7] T.J. Gordon and R.S. Sharp, On Improving the performance of automotive semi-active suspension systems through road preview, Journal
of Sound and Vibration 217(1) (1998), 163–182.
 S.H. Sadati et al. / Optimal control of an 8-DOF vehicle active suspension system using Kalman observer 503

[8] Y.J. Li, Optimal preview control design of active and semi-active suspension systems including jerk, society of automotive enginners, Inc,
1996.
[9] R.A. Williams, Automotive active suspensions-Part 1:Basic principles, Engineering center, Jaguar Cars, Coventry, Proceeding institute
Mechanical Engineering 211 (1997), part D.
[10] R.G.M. Huisman, An optimal continuous-time control strategy for active suspensions with preview, Journal of Vehicle System Dynamics
22 (1993), 43–55.
[11] J.B. Burl, Linear Optimal Control, Michigan Technological University, 1998.
[12] T.E. Fortman and K.L. Hitz, An Introduction to Linear Control Systems, Marcel Dekker, New York, 1997.