Sie sind auf Seite 1von 22

July 8, 2011 15:20 Vehicle System Dynamics Draft4

Vehicle System Dynamics


Vol. 00, No. 00, January 2008, 122
RESEARCH ARTICLE
Variable Stiness Suspension Mechanism:
Design, Analysis and Control
Olugbenga Moses Anubi and Carl D. Crane III

Center for Intelligent Machines and Robotics,


Department of Mechanical and Aerospace Engineering,
University of Florida,
Gainesville, Florida, USA
Email: {anubimoses, carl.crane}@gmail.com.
(March 2011)
A new variable stiness suspension system based on a recent variable stiness mechanism is
proposed. The overall system is composed of the traditional MacPherson-type passive sus-
pension system augmented with a variable stiness mechanism. The main idea is to improve
suspension performance by varying stiness in response to road disturbance. The system is
analyzed using a quarter car model which incorporates the kinematic details of the mechani-
cal design. The passive case shows much better performance over the traditional counterpart
which implies a good failsafe behavior. An L
2
-gain analysis based feedback controller was
designed using the concept of energy dissipation.
Keywords: Suspension, Variable Stiness, Vehicle, Nonlinear H

, dissipative, L
2
-gain
analysis
1. Introduction
The task of improving the drawbacks of passive suspension designs has been a
subject of many researches in the past [13]. The past approaches utilize one of
three techniques [4], adaptive, semi-active[3, 5] or fully active suspension[6, 7]. An
adaptive suspension utilizes a passive spring and an adjustable damper with slow
response to improve the control of ride comfort and road holding. A semi-active
suspension is similar, except that the adjustable damper has a faster response and
the damping force is controlled in real-time. A fully active suspension replaces the
damper with active elements, such as hydraulic actuators or air springs, which can
achieve optimum vehicle control, but at a high cost. The fully active suspension
is also not fail-safe in the sense that a very bad performance degradation results
whenever the control fails, which may be due to either mechanical, electrical or soft-
ware damage. In recent times, research in semi-active suspensions has continued to
advance with respect to their capabilities, narrowing the gap between semi-active
and fully active suspension systems. However, most semi-active systems can only
change the viscous damping coecient of the shock absorber while keeping the
stiness constant. Meanwhile, in passive suspension designs, both the damping

Address all correspondence to this author


ISSN: 0042-3114 print/ISSN 1744-5159 online
c 2008 Taylor & Francis
DOI: 10.1080/0042311YYxxxxxxxx
http://www.informaworld.com
July 8, 2011 15:20 Vehicle System Dynamics Draft4
2 Taylor & Francis and I.T. Consultant
coecient and the spring rate of the suspension elements are usually used as opti-
mization arguments. Therefore, a semi-active suspension system that varies both
the stiness and damping of the suspension element could provide more exibility
in balancing competing design objectives.
An alternative to lower the power consumption of active suspension elements was
introduced in [8, 9]. The variable geometry actuator was rst presented as a 3D
concept called the Delft Active Suspension (DAS). The basic idea behind the DAS
concept is based on a wishbone which could be rotated over an angle and is linked
to a pretensioned spring at a variable length. The spring pretension generates an
eective actuator force at the end of the wishbone. The force was controlled by
varying the position of spring via an electric motor. While the DAS has inherent
stiness variation, it is disadvantageous in the sense that it might be a source of
discomfort to the driver or even lead to instability in the absence of a secondary
spring. Also, there is a possibility of failsafe issues as typical of most active suspen-
sion systems[10]. In [11], a modied xed spring design with optimized geometry,
called the electromechanical low-power active suspension (eLPAS), was presented
and expected to overcome some challenges posed by the DAS concept. The moving
mass was reduced. Packaging and stiness variation issues were overcome by the
optimized geometry. However, ensuring a good failsafe behavior remains an open
problem. [10].
This paper presents new variable stiness suspension system which is based on a
recent variable stiness mechanism [12, 13]. The overall system is composed of the
traditional passive MacPherson suspension system augmented ( in parallel) with a
variable stiness mechanism. The main idea is to improve suspension performance
by varying stiness in response to road disturbance. The system is analyzed using
a quarter car model[14]. The passive
1
case shows superior performance over the
traditional counterpart. As a result, the problem of ensuring a good failsafe behav-
ior is addressed. A state feedback control was designed. The feedback control gains
were obtained by solving a system of Linear Matrix Inequalities which results from
an L
2
-gain analysis based on the energy dissipation concept.
The rest of this paper is organized as follows. In section 2, the variable stiness
mechanism is described and the basic forward analysis results from [12] are pre-
sented. In section 3, the variable stiness suspension system is introduced and the
stiness characteristics is examined. Section 4 details both the nonlinear dynam-
ics modeling and linearization about the equilibrium conguration. In section 5,
the performance of the passive system was evaluated by comparing the frequency
response with the traditional MacPherson counterpart. In section 6, an L
2
-gain
analysis-based feedback controller was designed. Time domain simulation results
were presented in section 7 and the performance of the resulting closed loop system
is compared with both the passive case and the MacPherson suspension. Conclusion
follows in section 8.
2. Variable Stiness Mechanism
In [12], a new variable stiness mechanism was presented. The mechanism is a
simple arrangement of two springs, a lever arm and a pivot bar. The eective
stiness is varied by changing the horizontal position d of the pivot while keeping
the point of application of the external load constant. It was shown that the eective
stiness is a rational function of d. A variation of this mechanism is shown in
1
Passive refers to the case where only pure mechanical elements like springs and dashboards are used in
the design. In otherwise no active force generating element is used or any control algorithm implemented.
July 8, 2011 15:20 Vehicle System Dynamics Draft4
Vehicle System Dynamics 3
Figure 1. Variable Stiness Mechanism Schematic.
Figure 1. The Lever arm PAC is pinned to and free to rotate about a xed point
C. The spring AB is constrained to remain vertical and can translate horizontally
as indicated by the double-headed arrow. Point A slides inside the Lever arm while
point B remains at the ground level. Without loss of generality, the external force
F is assumed to act vertically downwards at a point P on the lever which has a
horizontal distance L from C. is the clockwise inclination of the lever arm from
the horizontal plane. H is the height of the xed point C from the ground. d is the
length of the segment AC of the lever arm. The idea is to vary the overall stiness
of the system by adjusting d accordingly. This is achieved through the control force
u which will be designed in a subsequent section of this paper. Let k and l
0
be the
spring constant and the free length of the spring AB respectively, and the vertical
deection of the point P. The overall free length
0
of the mechanism is dened
as the vertical displacement of point P when no external force is acting on the
mechanism. Summing moments about C and equating to zero yields
F =
kd
Lcos
(l
0
H d tan )
=
kd
2
L
2
cos
2

_
(l
0
H)Lcos + (H 2Lsin )d
d

_
(1)
Hence, the force equation of the system becomes
F = K
e
(
0
)
(2)
where
K
e
=
kd
2
L
2
cos
2

(3)
July 8, 2011 15:20 Vehicle System Dynamics Draft4
4 Taylor & Francis and I.T. Consultant
Figure 2. Variable Stiness Suspension Mechanism.
and

0
= H (H l
0
)
L
d
cos 2Lsin (4)
are the overall stiness and free length of the mechanism respectively. In other
words, the whole system can be represented as a single spring whose spring constant
is determined by d according to (3). When the spring AB is on the left side of the
force F, the resultant spring is stier than the spring AB. In the same vein, when
the spring AB is on the right side of the force F, the resultant spring is softer. How
softer or stier the resultant spring is depends on how far right/left of force F the
spring AB is. Hence, by using the control force u to drive d, the overall stiness of
the mechanism can be varied dynamically.
3. System Description
Figure 2 shows the variable stiness suspension mechanism considered in this pa-
per. The sprung and unsprung masses are isolated by a parallel combination of a
July 8, 2011 15:20 Vehicle System Dynamics Draft4
Vehicle System Dynamics 5
Figure 3. Quarter Car Model of Variable Stiness Suspension Mechanism.
MacPherson strut and a variable stiness mechanism. The idealized quarter car
model shown in Figure 3 is composed of a quarter car body (sprung mass), a
MacPherson strut (spring and damper), a lower strut, lower wishbone and wheel
assembly (unsprung mass). The sprung mass is suspended from the unsprung mass
by the MacPherson and lower struts. The lower wishbone is free to rotate about
point O. Points A and B are xed on the car body. Points D and F are such that
the line DF always remains perpendicular to line CE. Point D can slide along the
line CE. The generalized coordinates are chosen as the vertical displacement y
s
of
the sprung mass, the horizontal displacement d of the lower strut and the rotation
angle of the lower wishbone. Assuming that the MacPherson strut is near vertical,
the eective stiness between the sprung and unsprung masses is approximately
k
eff
(d) = k
s
+
kd
2
l
2
G
cos
2

. (5)
July 8, 2011 15:20 Vehicle System Dynamics Draft4
6 Taylor & Francis and I.T. Consultant
where k
s
and k are the spring constants of the MacPherson and lower struts re-
spectively and l
G
is the length of the lower wishbone. k
eff
(d) varies from a min-
imum value of k
s
when d = 0 to a maximum value of k
s
+ ksec
2
as d l
G
.
This shows that the variable stiness suspension mechanism is can continuously
vary the eective stiness between the sprung and unsprung mass in the range
[k
s
, k
s
+ 2k], [

2
,

2
] in response to road disturbance.
4. Modeling
This section presents a detailed dynamic modeling of the system using the La-
grange equation of motion. The assumptions adopted in Figure 3 are summarized
as follows:
(1) The horizontal movement of the sprung mass is neglected, i.e only the
vertical displacement y
s
is considered.
(2) The sprung and unsprung masses are lumped at some points, hence are
assumed to be particles.
(3) The coil spring deection, the tire deection and the deection and the
damping forces are in the linear regions of their operating ranges.
Let
y
u
the vertical displacement of the unsprung mass (wheel assembly),
l
u
the distance between points X and X,
l
3
the distance between F and the center of mass of the lower strut along DF,
h
u
the distance between points G and X,
l
s
, l

s
the equilibrium and non-equilibrium lengths of the MacPherson strut respectively,
l
3
, l

3
the equilibrium and non-equilibrium lengths of the lower strut respectively,
(x
B
, y
B
) the coordinates of point B measured from the reference frame at point O on the chassis,
l
G
the length of the lower wishbone ,
l
F
the length of the segment OF of the lower wishbone,
k
c
, b
c
the spring constant and damping coecient of the control strut (Figure 2),
k
t
, b
t
the tire spring constant and damping coecient ,
b
s
, b
3
the damping coecients of the MacPherson and lower struts respectively,
m
s
, m
u
the sprung and unsprung masses ,
m
3
the mass of the lower strut ,
I the moment of inertia of the lower wishbone,
e
F
the unit vector from O to F along the lower wishbone.
The position vectors ,P
B
, P
D
, P
F
, P
X
of the points B,D,F and X are given
by:
P
B
= P
O
+
_
x
B
y
B
_
(6)
P
D
= P
O
+
_
d
y
C
_
(7)
P
F
= P
O
+ l
F
e
F
, l
F
= d sec (8)
P
X
= P
O
+ l
G
e
F
+
_
l
u
h
u
_
, (9)
July 8, 2011 15:20 Vehicle System Dynamics Draft4
Vehicle System Dynamics 7
where
P
O
=
_
0
y
s
_
(10)
is the position vector of point O and the unit vector e
F
is given by
e
F
=
_
cos
sin
_
. (11)
The change in length, l
3
, of the lower strut is given by
l
3
= l

3
l
3
= (P
D
P
O
)
T
_
0
1
_
= y
C
+ d tan . (12)
Similarly, the change in length, l
s
, of the MacPherson strut is given by
(l
s
)
2
= P
B
P
X

2
= AB cos C sin (13)
where
A = (x
B
+ l
u
)
2
+ (y
B
h
u
)
2
+ l
2
G
B = 2(x
B
+ l
u
)l
G
C = 2(y
B
h
u
)l
G
.
and the operator . denotes the euclidean norm operator which is dened for the
vector v =
_
v
1
v
2
v
n

T
as
v =
_
v
2
1
+ v
2
2
+ + v
2
n
_1
2
.
The respective position vectors of the centers of mass of the lower strut and the
unsprung mass are
P
3
= P
O
+ l
F
e
F
+ l
4
_
0
1
_
(14)
P
6
= P
O
+ l
G
e
F
. (15)
Dierentiating (12),(13),(14) and (15) yields

l
3
=

d tan + d

sec
2

l
s
=

C cos + B sin
2

AB cos C sin
(16)
July 8, 2011 15:20 Vehicle System Dynamics Draft4
8 Taylor & Francis and I.T. Consultant
_
_
_

P
3
_
_
_
2
=

d
2
+
_
y
s

_

d tan d

sec
2

__
2
= y
s
2
+

d
2
sec
2
+ d
2

2
sec
4
2
_
d

d

tan sec
2
+ y
s

d tan + d y
s

sec
2

_
(17)
_
_
_

P
6
_
_
_
2
= y
s
2
+

2
l
2
G
2 y
s

l
G
cos , (18)
Let the generalized coordinate vector be dened as
q =
_
y
s
d

T
(19)
and T(q, q), V (q), D(q, q) be the kinetic energy, potential energy and Rayleigh
dissipation functions of the system respectively, then
T(q, q) =
1
2
m
s
y
s
2
+
1
2
m
u
_
_
_

P
6
_
_
_
2
+
1
2
m
3
_
_
_

P
3
_
_
_
2
+
1
2
I

2
, (20)
V (q) =
1
2
k
s
(l
s
)
2
+
1
2
k
3
(l
3
)
2
+
1
2
k
c
(d
0
d)
2
+
1
2
k
t
(y
u
r)
2
m
s
gy
s
m
u
gy
u
m
3
g(y
s
d tan + l
4
), (21)
D(q, q) =
1
2
b
s
(

l
s
)
2
+
1
2
b
3
(

l
3
)
2
+
1
2
b
c

d
2
+
1
2
b
t
( y
u
r)
2
, (22)
where
y
u
= P
6
T
_
0
1
_
= y
s
l
G
sin .
Substituting eqns (16) through (18) into (20),(24) and (25) yields
T(q, q) =
1
2
(m
s
+ m
u
+ m
3
) y
s
2
+
1
2
m
3

d
2
sec
2
+
1
2
_
m
u
l
2
G
+ I + m
3
d
2
sec
4

2
m
u
y
s

l
G
cos m
3
_
d

d

tan sec
2
+ y
s

d tan + d y
s

sec
2

_
, (23)
V (q) =
1
2
k
s
(AB cos C sin ) +
1
2
k
3
(y
C
+ d tan )
2
+
1
2
k
t
(y
s
l
G
sin r)
2
+
1
2
k
c
(d
0
d)
2
(m
s
+ m
u
+ m
3
)gy
s
m
u
gl
G
sin + m
3
gl
3
m
3
gd tan
(24)
and
D(q, q) = b
s

2
(B sin C cos )
2
8 (AB cos C sin )
+
1
2
b
3
_

d tan + d

sec
2

_
2
+
1
2
b
c

d
2
+
1
2
b
t
( y
s
l
G

cos r)
2
. (25)
July 8, 2011 15:20 Vehicle System Dynamics Draft4
Vehicle System Dynamics 9
Dene the Lagrangian L(q, q) as
L(q, q) = T(q, q) V (q), (26)
then, the Lagrange equation of motion is given by
d
dt
_
T
q
_

T
q
+
V
q
+
D
q
= 0. (27)
which implies that
M(q) q + V
m
(q, q) q + B(q) q +K(q) G(q) = (k
t
r + b
t
r)W. (28)
Where the inertia matrix M(q) is given by
M(q) =
_
_
m
s
+ m
3
+ m
u
m
3
tan
m
3
tan m
3
sec
2

_
m
3
d sec
2
+ m
u
l
G
cos
_
m
3
d tan sec
2

_
m
3
d sec
2
+ m
u
l
G
cos
_
m
3
d tan sec
2

m
u
l
2
G
+ I + m
3
d
2
sec
4

_
_
, (29)
the Coriolis/Centripetal matrix V
m
(q, q) is given by
V
m
(q, q) =
_
_
0 m
3

0 m
3

tan sec
2

0 m
3
d

sec
4
2m
3

d tan sec
2

_
m
3

d sec

+2m
3
d

tan sec

m
u
l
G

sin
_
m
3

d tan sec
2
2m
3
d

(tan
2
sec
2
+ sec
4

m
3
d

d sec
4
+ 2m
3
d
2

tan sec
4

_
, (30)
the damping matrix B(q) is given by
B(q) =
_
_
b
t
0 b
t
l
G
cos
0 b
c
+ b
3
tan
2
b
3
d tan sec
2

b
t
l
G
cos b
3
d tan sec
2

1
4
b
s
f() + b
3
d
2
sec
2
+ b
t
l
2
G
cos
2

_
_
(31)
where
f = f() =
(B sin C cos )
2
4 (AB cos C sin )
, (32)
the stiness vector K(q) is given by
K(q) =
_
_
k
t
(y
s
l
G
sin )
k
c
(d
0
d) + k
3
(y
C
+ d tan ) tan
k
s
2
(B sin C cos ) + k
3
d(y
C
+ d tan ) sec
2
k
t
l
G
(y
s
l
G
sin ) cos
_
_
,
(33)
July 8, 2011 15:20 Vehicle System Dynamics Draft4
10 Taylor & Francis and I.T. Consultant
the gravity vector is given by
G(q) =
_
_
m
s
+ m
u
+ m
3
m
3
tan
m
u
l
G
cos m
3
d sec
2

_
_
, (34)
and the input vector W(q) is given by
W(q) = k
t
_
_
1
0
l
G
cos
_
_
. (35)
4.1. Dynamics Properties
This section explains the properties of the dynamics given in (28) which are ex-
ploited in the subsequent control design.
4.1.1. Inertia Matrix
The inertia matrix M(q) is symmetric, positive denite
x
T
M(q)x > 0, x R
3
(36)
4.1.2. Coriolis/Centripetal Vector
C(q, q) is bounded as follows
C(q, q) c q q
2
, c R
+
. (37)
Also, there exist a matrix V
m
(q, q) such that C(q, q) = V
m
(q, q) q and
x
T
_
1
2

M(q) V
m
(q, q)
_
x = 0, x R
3
(38)
The property (38) is the usual skew symmetric property of the Coriolis/centripetal
matrix of Lagrange dynamics [15].
4.1.3. Damping Matrix
The damping matrix B(q) is symmetric positive denite
x
T
B(q)x > 0, x R
3
. (39)
Also, there exists a positive denite matrices B and

B such that
x
T
Bx x
T
B(q)x x
T

Bx > 0, x R
3
. (40)
4.2. Linearization
For the purpose of frequency analysis, the dynamics of (28) is linearized about the
equilibrium conguration as
M
0
q + B
0
q +
_

K

G
_
q = W
0
r (41)
July 8, 2011 15:20 Vehicle System Dynamics Draft4
Vehicle System Dynamics 11
where
M
0
= M(q
0
), B
0
= B(q
0
), W
0
= W(q
0
)

G =
G
q

q=q
0
=
_
_
0 0 0
0 0 m
3
sec
2

0
0 m
3
sec
2
m
u
l
G
sin
0
2m
3
d
0
tan
0
sec
0
_
_
, (42)

K =
K
q

q=q
0
=
_
_
k
t
0 k
t
l
G
cos
0
0 k
c
+ k
3
tan
2

0
k
23
k
t
l
G
cos
0
k
23
k
33
_
_
(43)
where
k
23
= k
3
(y
C
+ 2d tan
0
) sec
2

0
,
k
33
=
1
2
k
s
(B cos
0
+ C sin
0
) + k
3
(y
C
+ 2d
2
tan
2

0
sec
2

0
+ d
2
sec
4

0
) + 2k
t
l
2
G
cos 2
0
.
Let
=
_
q
q
_
, (44)
then, the linearized dynamics of (41)is written in state space form as
=
_
0
33
I
33
M
1
0
_

K

G
_
M
1
0
B
0
_
+
_
0
31
M
1
0
W
0
_
r. (45)
5. Performance Evaluation
The performance of suspension systems is usually evaluated using the following
measures:
Ride Comfort
This is the amount of vibration experienced by the passenger on the vehicle. It
is quantied using the force transfered to the chassis. Since force is proportional
to acceleration, the ride comfort will be given in terms of the chassis acceleration
or car body acceleration, y
s
.
Suspension Deection
This characterizes the amount of deection undergone by the suspension ele-
ment. It is important because it provides a metric to determine the tendency of
the suspension element exceeding its physical deection limit. The suspension
deection is given in terms of the angular displacement of the lower wishbone.
Road Holding
This is the extent to which the road-tire contact is maintained. It is characterized
by the deection undergone by the tire. Hence, the road holding or dynamic tire
force is given in terms of the deection y
u
r undergone by the tire.
July 8, 2011 15:20 Vehicle System Dynamics Draft4
12 Taylor & Francis and I.T. Consultant
Thus, the controlled-output equation is given by
z
c
=
_
_
z
1
z
2
z
3
_
_
=
_
_
y
s
(l
s
)
2
y
s
l
G
sin r
_
_
which is linearized as
z
c
=
_

_
e
T
1
M
1
0
_

K

G
_
e
T
1
M
1
0
B
0
0 0 B sin
0
C cos
0
1 0 l
G
cos
0
0
23
_

_
+
_
_
0
0
1
_
_
r. (46)
where
e
1
=
_
1 0 0

T
5.1. Performance Weights
The performance measure used in this paper is the weighted H

norm of the
Linear-Time-Invariant system given by (45) and (46) from the input signal r to the
controlled-output signal z
c
. The following frequency-dependent weighting functions
are used.
Road Roughness Filter
The disturbances acting on the vehicle suspension system include road irregu-
larities, braking forces, acceleration forces, inertia forces on a curved track, and
payload changes. Among them, road roughness is the most important disturbance
to either the rider or the vehicle structure itself[16]. For the purpose of this paper,
the road disturbance input to the vehicle model will be represented by a white
Gaussian noise signal of power spectral density
S
psd
() = G
r

2
passing through a a rst-order lter[17] given by
W
r
(s) =
(2G
r
V )
1/2
s + 2V
0
(47)
where G
r
= 6410
7
is the road-roughness coecient for a road class D (average)
according to the International Organization for Standardization(ISO)[18, 19],
0
=
0.01cycle/m and V is the speed of the vehicle. The simulations in this paper assumes
a constant vehicle speed of 40mph (17.8816 m/s).
Vertical Acceleration Weighting
Ride comfort is measured by a specic index, which depends on the accelera-
tion level, frequency, direction, and location[17]. The ISO 2631 [20] describes a
method of evaluating the eect of exposure of humans to vibration by weighting
the root-mean square (RMS) acceleration with human vibration-sensitivity curves.
The frequency dependent weight
W
1
(s) =
50s + 500
s
2
+ 50s + 1200
(48)
July 8, 2011 15:20 Vehicle System Dynamics Draft4
Vehicle System Dynamics 13
Figure 4. System Interconnection block diagram.
Figure 5. Car Body Acceleration.
has been used[17, 21, 22] to approximate the the ISO weighting curves and is also
used in this work to penalize the car body acceleration as shown in Figure 4.
The other performance weights used are
W
2
(s) =
5
s + 10
(49)
and
W
3
(s) =
500
s + 20
(50)
for the suspension deection and the road holding penalties respectively.
Figures 5 through 7 show the frequency responses for the Traditional MacPherson
July 8, 2011 15:20 Vehicle System Dynamics Draft4
14 Taylor & Francis and I.T. Consultant
Figure 6. Suspension Deection.
Figure 7. Dynamic Tire Force.
Suspension Mechanism [14] and the Variable Stiness Suspension Mechanism.
6. Control Design
The Taylor series expansion of the stiness vector K(q) and the gravity vector
G(q) in (28) about the equilibrium point q
0
are given by
K(q) G(q) = K(q
0
) G(q
0
) +
_

K

G
_
(q q
0
) +O
_
q q
0

2
_
. (51)
July 8, 2011 15:20 Vehicle System Dynamics Draft4
Vehicle System Dynamics 15
where O
_
q q
0

2
_
denotes the higher order terms.
Let
e = q q
0
(52)
be the error between the instantaneous and equilibrium congurations of the sys-
tem. The dynamics of (28) is re-written as
M e + V
m
e + B e +
_

K

G
_
e +w = e
2
u (53)
where
w = O
_
q q
0

2
_
(b
t
r + k
t
r)W. (54)
e
2
=
_
0 1 0

T
. (55)
The arguments in the parenthesis are dropped for clarity sake. The output equation
z
c
= L
1
e + L
2
e (56)
where
L
1
=
_

_
e
T
1
M
1
0
_

K

G
_
0 0 B sin
0
C cos
0
0 0 l
G
cos
0
_

_
, L
2
=
_
e
T
1
M
1
0
B
0
0
23
_
is used to characterize ride comfort, suspension deection and road holding perfor-
mances. The controller is designed using the L
2
-gain analysis based on the concept
of energy dissipation[2325]. the following denition gives a brief review of the
concept of energy dissipation.
Definition 6.1: Consider the nonlinear system
x = f(x, w)
z = h(x, w) (57)
where x R
n
, w R
p
, z R
m
are the state, input and output vectors respectively.
The system is said to be -dissipative, > 0, if there exists an energy storage
function V (x(t)) > 0 x = 0 with V (x(0)) = 0 such that
_
T
0
z
2
dt
2
_
T
0
w
2
dt V (x(t)), T R
+
. (58)
That is, the L
2
-gain of the system is less than or equal . Alternatively, the system
(57) is -dissipative if there exists a nonnegative function V such that the energy
Hamiltonian function dened by
H = z
2

2
w
2
+
dV
dt
(59)
is non positive for all x and w in the domain of interest.
July 8, 2011 15:20 Vehicle System Dynamics Draft4
16 Taylor & Francis and I.T. Consultant
Given the system (53) with the output given by
z =
_
z
c
u
_
(60)
where R
33
, R, > 0, > 0
1
, and following Denition 6.1, the control
objective is to nd a state feedback control u such that the corresponding closed
loop system becomes -dissipative. The design of the state feedback controller is
then summarized in the following theorem.
Theorem 6.2 : Given the state feedback control law
u = F
1
e +F
2
e (61)
If there exists a positive denite matrix Q, together with , such that
B + N
3
+
1
4
2
I +e
2
F
2
+
2
F
T
2
F
2
0, (62)

K +

G + N
2
+ 2F
T
1
F
2
+e
2
F
1
0 (63)
Q
_

K

GN
2
2F
T
1
F
2
e
2
F
1
_
+ N
1
+
2
F
T
1
F
1
0, (64)
where
N
1
= L
T
1

T
L
1
(65)
N
2
= 2L
T
1

T
L
2
(66)
N
3
= L
2

T
L
2
(67)
then the closed loop system
M e + (B e
2
F
1
) e + V
m
e +
_

K

Ge
2
F
2
_
e +w = 0 (68)
with output given by (60) is -dissipative, > 0.
Proof : Let P R
33
be a positive semidenite matrix dened as follows
P = P
0
e
2Qt
, Q = Q
T
> 0 (69)
and consider a positive denite energy function given by
V (e, e) =
1
2
e
T
M e +
1
2
e
T
P
T
e. (70)
Taking time derivative yields

V = e
T
M e +
1
2
e
T

M e + e
T
Pe +
1
2
e
T

Pe, (71)
1
The inequality operator {<, , >, } is dened for matrix argument A R
nn
as: A 0, x
T
Ax
0 x R
n
, x = 0
July 8, 2011 15:20 Vehicle System Dynamics Draft4
Vehicle System Dynamics 17
again dropping arguments in parenthesis for clarity sake. Substituting (68) and
(69) into (71) yields

V = e
T
((B e
2
F
1
) e w) + e
T
_
P

K +

G +e
2
F
2
_
e e
T
QPe
+ e
T
_
V
m

1
2

M
_
e. (72)
Using the skew symmetric property given in (38) yields

V =
2
w
2
z
2

2
w
2
+e
T
_
P

K +

G + N
2
+ 2F
T
1
F
2
+e
2
F
1
_
e
+e
T
_
QP + N
1
+
2
F
T
1
F
1
_
e + e
T
_
B + N
3
+w+e
2
F
2
+
2
F
T
2
F
2
_
e
=
2
w
2
z
2
+ e
T
_
B + N
3
+
1
4
2
I +e
2
F
2
+
2
F
T
2
F
2
_
e
2
_
_
_
_
w
1
2
2
e
_
_
_
_
2
+e
T
_
P

K +

G + N
2
+ 2F
T
1
F
2
+e
2
F
1
_
e +e
T
_
QP + N
1
+
2
F
T
1
F
1
_
e.
(73)
Using the boundedness property of B(q) given in (40), (73) becomes

V
2
w
2
z
2
+ e
T
_
B + N
3
+
1
4
2
I +e
2
F
2
+
2
F
T
2
F
2
_
e
+e
T
_
P

K +

G + N
2
+ 2F
T
1
F
2
+e
2
F
1
_
e +e
T
_
QP + N
1
+
2
F
T
1
F
1
_
e.
(74)
Letting
P =

K

GN
2
2F
T
1
F
2
e
2
F
1
0 (75)
and utilizing the inequalities (62) through (64) given in Theorem 6.2, (74) becomes

V
2
w
2
z
2
.
Integrating the above inequality yields
V (e(t), e(t))
2
_
T
0
w
2
dt
_
T
0
z
2
dt (76)
which implies that the closed loop system (68) is -dissipative.
July 8, 2011 15:20 Vehicle System Dynamics Draft4
18 Taylor & Francis and I.T. Consultant
Table 1. DYNAMIC PARAMETER VALUES
Parameter Value
m
s
453 kg
m
u
71 kg
m
3
20 kg
b
s
100 N/m/s
b
s
2
1500 N/m/s
b
3
1500 N/m/s
b
c
1000 N/m/s
b
t
700 N/m/s
k
s
10000 N/m
k
s
2
20000 N/m
k
3
10000N/m
k
c
10000N/m
k
t
190000 N/m
I 0.001kgm
2
Table 2. KINEMATIC PARAMETER VALUES
Parameter Value
l
G
0.4m
h
u
0.25m
l
u
0.13m
x
B
0.45m
y
B
0.5m
y
C
0.3m
7. Simulation
Using Schur complement, the inequalities given in (62)through (64) are transformed
into a system of Linear Matrix Inequality as follows
_
B + N
3
+
1
4
2
I +e
2
F
2
F
T
2
F
2

1

2
_
0 (77)
_
Q
_

K

GN
2
_
+ N
1
+ Q
_
2F
T
2
+e
2
_
F
1
F
T
1
F
1

1

2
_
0 (78)

K +

G + N
2
+
_
2F
T
2
+e
2
_
F
1
0 (79)
which can be solved using ecient convex optimization algorithms. The values of
the kinematic as well as dynamics parameters used are given in Tables 1 and 2.
The simulation parameters are given in Table 3. In the time domain simulation, the
vehicle traveling at a steady horizontal speed of 40mph is subjected to a road bump
of height 10cm. The responses (Chassis Acceleration, Suspension Deection and
Tire Deection) are compared between the traditional MacPherson-type suspen-
sion, the passive and the active cases of the proposed variable stiness suspension.
Figures 8, 9 and 10 shows the Car body Acceleration, Suspension Deection and
Dynamic Tire Force responses respectively. The proposed variable stiness suspen-
sion system allows getting globally much better performances than the traditional
July 8, 2011 15:20 Vehicle System Dynamics Draft4
Vehicle System Dynamics 19
Table 3. SIMULATION PARAMETER VALUES
Parameter Value
0.1
0.01
F
1
_
100 1000 100

F
2
_
200 1000 200

diag
__
0.2 0.1 0.02
_
Q I
Figure 8. Car Body Acceleration.
MacPherson-type suspension, improving ride comfort and road holding. Figure 11
shows the control authority from which it is seen that the control input is bounded
and smooth.
8. Conclusion
This paper examined the design of a variable stiness suspension system. The main
idea stemmed from a variable stiness mechanism which changes its stiness char-
acteristics by adjusting the horizontal position of a pivot bar. The proposed system
showed improved performances in terms of ride comfort, suspension deection and
road holding. An L
2
-gain analysis based feedback controller was designed to further
improve the performance of the system by compensating for unattainable stiness
and damping in the passive case. The passive case shows much better performance
over the traditional counterpart which implies a good failsafe behavior.
In future works, other variations to the mechanism will be considered with pos-
sible design optimization implemented. Also, possible roll/pitch improvement will
be examined using a half-car model.
July 8, 2011 15:20 Vehicle System Dynamics Draft4
20 REFERENCES
Figure 9. Suspension Deection.
Figure 10. Dynamic Tire Force.
9. Acknowledgment
The authors would like to gratefully acknowledge the support provided by the
Department of Energy via the University Research Program in Robotics (URPR),
grant number DE-FG04-86NE37967.
References
[1] Alkhatib, R., Jazar, G. N., and Golnaraghi, M. F., 2004. Optimal design of passive linear suspension
using genetic algorithm. Journal of Sound and vibration, 275, pp. 665691.
[2] Williams, R. A., 1997. Automotive active suspensions. part 1: basic principles. In Proceedings of
July 8, 2011 15:20 Vehicle System Dynamics Draft4
REFERENCES 21
Figure 11. Control Authority.
IMechE, Vol. 211, pp. 415426.
[3] Butsuen, T., 1989. The design of semi-active suspensions for automotive vehicles. PhD thesis,
Massachussets Institute of Technology.
[4] Ashfak, A., Saheed, A., Rasheed, K. A., and Jaleel, J. A., 2009. Design, fabrication and evaluation
of mr damper. World Academy of Science, Engineering and Technology, 53.
[5] Do, A.-L., Sename, O., and Dugard, L., 2010. An lpv control approach for semi-active suspension
control with actuator constraints. In 2010 American Control Conference.
[6] Fialho, I., and Balas, G. J., 2002. Road adaptive active suspension design using linear parameter-
varying gain-scheduling. IEE Transactions on control systems technology, 10(1), pp. 4354.
[7] Williams, R. A., Best, A., and Crawford, I. L., 1993. Rened low frequency active suspension. In
Proceedings of IMechE International Conference, pp. 285300.
[8] van der Knaap, A., 1989. Design of a low power anti-roll/pitch system for a passenger car. PhD
thesis, Delft University of Technology.
[9] Venhovens, P. J. T., and van der Knaap, A. C. M., 1995. Delft active suspension (das). background
theory and physical realization. Smart Vehicles, pp. 139165.
[10] Evers, W.-J., Teehuis, A., van der Knaap, A., Besselink, I., and Nijmeijer, H., 2011. The electrome-
chanical low-power active suspension: Modeling, control, and prototype testing. Journal of Dynamic
Systems, Measurement, and Control, 133, April.
[11] van der Knaap, A. C. M., Teerhuis, A. P., Tinsel, R. B. G., and Vershuren, R. M. A. F., 2008. Active
suspension assembly for a vehicle. International Patent No. 2008/049845.
[12] Anubi, O. M., Crane, C., and Ridgeway, S., 2010. Design and analysis of a variable stiness mech-
anism. In Proceedings IDETC/CIE2010. ASME 2010 International Design Engineering Technical
Conferences & Computers and Information in Engineering Conference.
[13] Anubi, O. M., and Crane, C., 2011. Semi-global output feedback asymptotic tracking for an under-
actuated variable stiness mechanism. In Proceedings of the 13th IFTOMM World Congress in
Mechanism and Machine Science.
[14] Fallah, M. S., 2010. New dynamic modeling and practical control design for macpherson suspension
system. PhD thesis, Concordia University, Montreal, Quebec, Canada, April.
[15] Lewis, F. L., Dawson, D. M., and Abdallah, C. T., 2004. Robot Manipulator Control, Theory and
Practice, 2nd ed. Marcel Dekker, Inc.
[16] Elbeheiry, E., Karnopp, D., and Abdelraaouf, A., 1995. Advanced ground vehicle suspension system
- a classied biography. Vehicle System Dynamics, 24, pp. 231258.
[17] Zuo, L., and Nayfeh, S. A., 2003. Structured h
2
optimization of vehicle suspensions based on multi-
wheel models. Vehicle System Dynamics, 40, pp. 351371.
[18] Tanghirad, H., and Esmailzadeh, E., 1998. Automobile passenger comfort assured through lqg/lqr
active suspension. Journal of Vibration Control, 4(5), pp. 603618.
[19] Elbeheiry, E., Karnopp, D., and Abdelraaouf, A., 1996. Suboptimal control design of active and
passive suspensions based on a full car model. Vehicle System Dynamics, 26, pp. 197222.
[20] , 1997. International Organization for Standardization, ISO 2631-1:1997, Mechanical Vibration and
Shock Evaluation of Human Exposure to Whole Body Vibration - Part 1: General Requirements.
[21] Zuo, L., and Nayfeh, S. A., 2003. Low order continuous-time lters for approximation of the iso
2631-1 human vibration sensitivity weightings. Journal of Sound and Vibration, 265, pp. 459465.
[22] Muller, P., Popp, K., and Schiehlen, W., 1979. Covariance analysis of nonlinear stochastic guideway-
vehicle-systems. In Proceedings of the Sixth IAVSD Symposium, pp. 337351.
[23] Yim, J., and Park, J. H., 1999. Nonlinear h

control of robotic manipulator. In IEEE Conference


July 8, 2011 15:20 Vehicle System Dynamics Draft4
22 REFERENCES
on System, Man and Cybernetics,SMC, Tokyo, Japan, pp. 960965.
[24] van der Schaft, A. J., 1992. L
2
-gain analysis of nonlinear systems and nonlinear state feedback h

.
IEEE Trans. Automat. Contr, 37(6), pp. 770784.
[25] Isidori, A., and Astol, A., 1992. Disturbance attenuation and h

control via measurement feedback


in nonlinear systems. IEEE Trans. Automat. Contr, 37(6), pp. 12831293.

Das könnte Ihnen auch gefallen