TS fuzzy model with nonlinear consequence and PDC controller

for a class of nonlinear control systems

R. Rajesh
*
, M.R. Kaimal
Department of Computer Science, University of Kerala, Kariavattom Campus, University of Kerala, Thiruvananthapuram-695581, India
Received 30 December 2003; received in revised form 7 September 2005; accepted 16 January 2006
Available online 18 April 2006
Abstract
In this paper a new TakagiSugeno (TS) fuzzy model with nonlinear consequence (TSFMNC) is presented which can approximate a class of
smooth nonlinear systems, nonlinear dynamical systems and nonlinear control systems. It is also proved that TakagiSugeno fuzzy controller with
nonlinear consequence (TSFCNC) can be used to approximate a class of nonlinear state-feedback controllers using the so-called parallel
distributed compensation (PDC) method. The inverted pendulum problem has been simulated with TSFCNC and compared with TakagiSugeno
fuzzy controller with linear consequence (TSFCLC) and the results show that TSFCNC performs better than TSFCLC. A real-life example of
dynamic positioning of ship is simulated and the results also show that TSFCNC performs better than TSFCLC.
# 2006 Elsevier B.V. All rights reserved.
Keywords: TS fuzzy model; TS fuzzy model with nonlinear consequence; Parallel distributed compensation; Inverted pendulum; Fuzzy control; Feedback control
1. Introduction
The essential element for the study of a nonlinear control
problem is to get a tractable model of a dynamical system for
use in control system design. The design should be simple
enough to work with, but must retain the essential features of
the system. Various schemes using fuzzy logic models [1
4,9,12,13,15,18] have been developed in the last few decades
since the Zadehs seminal paper on fuzzy sets [19], one of
which is the TakagiSugeno fuzzy model [15]. Wang et. al.
proposed the so-called PDC [17] as a design frame work and
also modied Tanakas stability theorem to include a control
algorithm. Recently in [16] it has been proven that TS fuzzy
model with linear rule consequence and PDC controller as a
universal framework for nonlinear systems.
This paper presents TakagiSugeno fuzzy model with
nonlinear consequence (TSFMNC). It is proved that TSFMNC
can approximate a class of nonlinear systems, nonlinear dynamic
systems and nonlinear control systems. TakagiSugeno fuzzy
controller with nonlinear consequence (TSFCNC) is designed
using parallel distributed compensation (PDC) method which
can better approximate a class of nonlinear state-feedback
controllers. A famous benchmarking problem, namely, the
inverted pendulum problem has been simulated using both
TSFCNCand TSFCLC, and the results showbetter performance
with TSFCNC. A real-life example of dynamic positioning of
ship is simulated and the results also show that TSFCNC
performs better than TSFCLC.
This paper is organized as follows. Section 2 deals with the
approximation of nonlinear functions using TakagiSugeno
fuzzy models with nonlinear consequence. Section 3 deals with
the applications to modeling and control of nonlinear dynamic
systems. Section 4 deals with a real-life example of dynamic
positioning of ships. Section 5 concludes the paper.
2. Approximation of nonlinear functions using Takagi
Sugeno fuzzy models with nonlinear consequence
2.1. TakagiSugeno fuzzy models with nonlinear
consequence
Suppose that the nonlinear function f x : R
n
!R is dened
over a compact region D&R
n
with the following assumptions:
A
1
. f 0 0.
www.elsevier.com/locate/asoc
Applied Soft Computing 7 (2007) 772782
* Corresponding author at: Department of Computer Science and Engineer-
ing, Bharathiar University, Coimbatore-641046, India.
E-mail addresses: kollamrajeshr@ieee.org (R. Rajesh),
mrkaimal@yahoo.com (M.R. Kaimal).
1568-4946/$ see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.asoc.2006.01.014
A
2
. f 2C
2
1
. Therefore, f ;
@ f
@x
and
@
2
f
@x
2
are continuous and there-
fore bounded over D.
A
3
. f x is expressable in the form of f x axcx and
can be approximated as f
0
x a
0
cx in the region D
0

fx=jx
i
j < e
0
g and f
j
1
j
2
;...; j
n
x a
j
1
j
2
;...; j
n
cx in the region
D
j
1
j
2
;...; j
n
fx=x 2D; j
i
e x
i
j
i
1e 8 ig where j
i
are
integers, e
0
and e are small positive numbers. x x
1
; x
2
;
. . . ; x
n

T
. ax is continuous and a
0
ax
00...0
, a
j
1
j
2
;
. . . ; j
n
ax
j
1
j
2
;...; j
n
2R
1n
. cx : R
n
!R
n
is dened over
compact region D&R
n
. cx c
1
x
1
; c
2
x
2
; . . . ; c
n
x
n

T
and c
i
x
i
is either equal to sin x
i
or equal to x
i
.
Then f x can be approximated by TSFMNC. The rules of
TSFMNC is of the following form:
Rule 0: IF x
1
is about 0 . . . and x
n
is about 0 THEN

fx
a
0
cx.
Rule j
1
j
2
; . . . ; j
n
: IF x
1
is about j
1
e. . . and x
n
is about j
n
e
THEN

f a
j
1
j
2
;...; j
n
cx.
For Rule 0, choose the possibilityof ringh
0
x as, 1inside D
0
and, 0 outside. The possibility of ring for the j
1
j
2
; . . . ; j
n
th rule
is given by the product of all membership functions associated
with the j
1
j
2
; . . . ; j
n
th rule and is h
j
1
j
2
;...; j
n
xt

n
i1
M
j
i
x
i
t where the membership functionfor x
i
is givenby(1). It
is assumed that h
j
1
j
2
;...; j
n
x have already been normalized, i.e.
h
j
1
j
2
;...; j
n
x !0 and

j
1
j
2
;...; j
n
h
j
1
j
2
;...; j
n
x 1.
M
j
i
x
i

1
jx
i
j
i
ej
e
jx
i
j
i
ej < e
0 else where
_
(1)
Then by using center of gravity method for defuzzication,
the TSFMNC can be represented as:
y

fx h
0
xa
0
cx

j
1
j
2
;...; j
n
h
j
1
j
2
;...; j
n
xa
j
1
j
2
;...; j
n
cx
(2)
2.2. Analysis of approximation
In this subsection, we will prove the fact that any smooth
nonlinear function satisfying the Assumptions A
1
A
3
can be
approximated, to any degree of accuracy, using TS fuzzy
model with nonlinear consequence. In the following discus-
sions, we will only concentrate on one of such regions D
j
1
j
2
;...; j
n
by assuming that x 2D
j
1
j
2
;...; j
n
. In the following, for simplicity,
f is substituted instead of j
1
j
2
; . . . ; j
n
.
Consider ex, the approximation error between f x and

fx:
kexk
_
_
_
_
_
f x

f
h
f
xa
f
cx
_
_
_
_
_

_
_
_
_
_
f x

f
h
f
xa
f
cx
f

f
h
f
xa
f
cx cx
f

_
_
_
_
_

_
_
_
_
_
f x

f
h
f
x f x
f

f
h
f
xa
f
cx
cx
f

_
_
_
_
_

f
h
f
xk f x f x
f
k

f
h
f
xka
f
cx cx
f
k max
f
k f x
f x
f
k max
f
ka
f
cx cx
f
k
Since x 2D
j
1
j
2
;...; j
n
, the distance between x and any vertex point
of D
j
1
j
2
;...; j
n
is less than

n
p
e, i.e. jx x
j
1
j
2
;...; j
n
j

n
p
e, and
since cx is continuous over x, i.e. jcx cx
j
1
j
2
;...; j
n
j <d
where d is an arbitrarily small number, it is possible to make
ex arbitrarily small by just reducing e.
Now consider the approximation of
de
dx
. Before nding the
approximation of
@e
@x
, consider the following two facts:
Fact 1. [16] Dene
@h
j
1
j
2
;...; j
n
@x

@h
j
1
j
2
;...; j
n
@x

x
1
@h
j
1
j
2
;...; j
n
@x

x
2
; ;
@h
j
1
j
2
;...; j
n
@x

x
n
_ _
where it exists, then

j
1
j
2
;...; j
n
@h
j
1
j
2
;...; j
n
@x

x
0 (3)
Fact 2. [16]

j
1
j
2
;...; j
n
x x
j
1
j
2
;...; j
n

@h
j
1
j
2
;...; j
n
@x

x
I (4)
Now consider the approximation of
@e
@x
, the difference
between
@ f
@x
and
@

f
@x
. In the following, for simplicity, f is
substituted instead of j
1
j
2
; . . . ; j
n
.
R. Rajesh, M.R. Kaimal / Applied Soft Computing 7 (2007) 772782 773
@e
@x
_
_
_
_
_
_
_
_

@ f
@x

f
h
f
xa
f
cx
_ _
@x
_
_
_
_
_
_
_
_
_
_
_
_

@ f
@x

f
a
f
cx
@h
f
x
@x

f
h
f
xa
f
@cx
@x

x
_
_
_
_
_
_
_
_
_
_

_
_
_
_
_
@ f
@x

f
a
f
cx
f

@h
f
x
@x

f
a
f
cx
f

@h
f
x
@x

f
a
f
cx
@h
f
x
@x

f
h
f
xa
f
@cx
@x

x
_
_
_
_
_

_
_
_
_
_
@ f
@x

f
f x
f

@h
f
x
@x

f
a
f
cx
cx
f

@h
f
x
@x

f
h
f
xa
f
@cx
@x

x
_
_
_
_
_
Since x 2D
j
1
j
2
;...; j
n
, the distance between x and any vertex point
of D
j
1
j
2
;...; j
n
is less than

n
p
e, i.e. jx x
j
1
j
2
;...; j
n
j

n
p
e, and
since cx and ax are continuous over x, i.e. jcx
cx
j
1
j
2
;...; j
n
j <d and jax ax
j
1
j
2
;...; j
n
j <d
1
where d and d
1
are arbitrarily small numbers, it is possible to make ex
arbitrarily small by just reducing e.
Next consider region D
0
. In region D
0
, it is known from
Taylor series that ex and
de
dx
can also be made arbitrarily small
by reducing e
0
. Therefore, the following theorem is obtained by
summarizing the results above:
Theorem 1. For a smooth nonlinear function f x : R
n
!R
satisfying Assumptions A
1
, A
2
and A
3
can be approximated, to
any degree of accuracy, by a TS fuzzy model with nonlinear
consequence. Furthermore, its derivative can be approximated
to any degree of accuracy, except for a nite number of points.
2.3. Example
An example is given in this subsection for illustration.
Consider the approximation of two-dimensional nonlinear
function f x
1
; x
2
cos
2
x
1
sin x
1
x
1
x
2
, where x
1
; x
2

p=2; p=2. The plot of the function is shown in Fig. 1.
TSFMNC is constructed where j
1
j
2
th rule is of the
form:
Rule
j
1
j
2
: IF x
1
is j
1
e and x
2
is j
2
e THEN
f x
1
; x
2
cos
2
j
1
e; j
2
e
sin x
1

x
2
_ _
TSFMLC is constructed where j
1
j
2
th rule is of the form:
Rule
j
1
j
2
: IF x
1
is j
1
e and x
2
is j
2
e THEN
f x
1
; x
2

cos
2
j
1
e sin j
1
e
j
2
e
; j
2
e
_ _
x
1
x
2
_ _
In both the cases e 10p=180 and j
i
9; 8; . . . ;
1; 0; 1; . . . ; 8; 9 for i 1; 2. Fig. 2 shows the plot of TSFMNC
and Fig. 3 shows the plot of TSFMLC. Fig. 4 shows the
difference of the function and TSFMNC and Fig. 5 shows
the difference of the function and TSFMLC. Fig. 6 shows the
difference of TSFMNC and TSFMLC. 1024 data points
(32 for x
1
32 for x
2
) are taken for simulation and found the
R. Rajesh, M.R. Kaimal / Applied Soft Computing 7 (2007) 772782 774

_
_
_
_
_
@ f
@x

f
f x
@ f
@x

x
x
f
x Oe
2

@h
f
x
@x

f
a
f
cx cx
f

@h
f
x
@x

f
h
f
xa
f
@cx
@x

x
_
_
_
_
_

_
_
_
_
_
@ f
@x

f
@ f
@x

x
x
f
x
@h
f
x
@x

f
a
f
cx
cx
f

@h
f
x
@x

f
h
f
xa
f
@cx
@x

x
_
_
_
_
_
Oe
From Fact1

_
_
_
_
_

f
a
f
cx cx
f

@h
f
x
@x

f
h
f
xa
f
@cx
@x

x
_
_
_
_
_
Oe
From Fact2

_
_
_
_
_

f
a
f
cx cx
f

@h
f
x
@x

x
ax
@cx
@x

f
h
f
xa
f
@cx
@x

f
h
f
xax
@cx
@x

x
_
_
_
_
_
Oe
_
_
_
_
_

f
a
f
axcx
cx
f

@h
f
x
@x

x
@cx
@x

x
_
_
_
_
_

f
h
f
xa
f
ax
@cx
@x

x
_
_
_
_
_
_
_
_
_
_
Oe
From Fact2
Fig. 2. Plot of TSFMNC.
Fig. 1. Plot of f x
1
; x
2
cos
2
x
1
sin x
1
x
1
x
2
.
following error in modeling:

x
1

x
2
f x
1
; x
2
TSFMLCx
1
; x
2

2
0:0088 (5)

x
1

x
2
f x
1
; x
2
TSFMNCx
1
; x
2

2
0:0079 (6)

x
1

x
2
TSFMLCx
1
; x
2
TSFMNCx
1
; x
2

2
1:0078 10
4
(7)
The errors show that TSFMNC better approximates the func-
tion than TSFMLC.
3. Applications to modeling and control of nonlinear
dynamic systems
3.1. Approximation of nonlinear dynamic systems using
TakagiSugeno fuzzy models with nonlinear consequence
The TSFMNC is used to describe dynamic systems. It is of
the following form:
Rule 0: IF x
1
is about 0 . . . and x
n
is about 0 THEN
x A
0
cx.
Rule j
1
j
2
; . . . ; j
n
: IF x
1
is about j
1
e . . . and x
n
is about j
n
e
THEN x A
j
1
j
2
;...; j
n
cx.
where x x
1
; x
2
; . . . ; x
n

T
are the system states and
cx c
1
x
1
; c
2
x
2
; . . . ; c
n
x
n

T
.
By using center of gravity method for defuzzication,
TSFMNC can be represented as Eq. (8) where h
j
1
j
2
;...; j
n
x is
the possibility for the j
1
j
2
; . . . ; j
n
th rule to re:
x h
0
xA
0
cx

j
1
j
2
;...; j
n
h
j
1
j
2
;...; j
n
xA
j
1
j
2
;...; j
n
cx (8)
Consider the nonlinear system (9) where f x is a vector eld
dened over compact region D&R
n
satisfying Assumptions
A
1
, A
3
and the following Assumption A
4
:
A
4
. f 2C
2
n
. Therefore, f,
@ f
@x
and
@
2
f
@x
2
are continuous and there-
fore bounded over D:
x f x (9)
Suppose f x can be written as f
1
x; . . . ; f
n
x
T
. What we
mean by approximation is nding a TSFMNC given by

fx

f
1
x; . . . ;

f
n
x
T
such that k f x

fxk is small. Since
k f x

fxk is small iff each of its components (which are
nonlinear functions) is small, then by applying Theorem 1
proven in the previous section, the following corollary is
obtained:
R. Rajesh, M.R. Kaimal / Applied Soft Computing 7 (2007) 772782 775
Fig. 5. Difference of f x
1
; x
2
and TSFMLC.
Fig. 6. Difference of TSFMNC and TSFMLC.
Fig. 4. Difference of f x
1
; x
2
and TSFMNC.
Fig. 3. Plot of TSFMLC.
Corollary 2. Any smooth nonlinear system(9) satisfying the
Assumptions A
1
, A
3
and A
4
can be approximated, to any degree
of accuracy, by a TSFMNC(8).
Similarly, smooth nonlinear control system x f x
gxu can also be approximated using a TSFMNC given by
x h
0
xA
0
cx B
0
u

j
1
j
2
;...; j
n
h
j
1
j
2
;...; j
n
x
A
j
1
j
2
;...; j
n
cx B
j
1
j
2
;...; j
n
u
3.2. Approximation of nonlinear state-feedback controllers
using TakagiSugeno fuzzy controller with nonlinear
consequence
TakagiSugeno fuzzy controller with nonlinear conse-
quence (TSFCNC) is designed using parallel distributed
compensation (PDC) method. The j
1
j
2
; . . . ; j
n
th rule of the
TSFCNC is given by
Rule 0: IF x
1
is about 0 . . . and x
n
is about 0 THEN
u K
0
cx.
Rule j
1
j
2
; . . . ; j
n
: IF x
1
is about j
1
e . . . and x
n
is about j
n
e
THEN u K
j
1
j
2
;...; j
n
cx.
The output of the TSFCNC is
u h
0
K
0
cx

j
1
j
2
;...; j
n
h
j
1
j
2
;...; j
n
xK
j
1
j
2
;...; j
n
cx (10)
Following similar argument as in the above sections, the
following theorem is obtained:
Theorem 3. For smooth nonlinear state feedback controller
u Kcx where x is dened over a compact region
can be approximated, to any degree of accuracy, by TSFCNC
(10).
Remark 4. Most of the real-life control system applications
like ball and beam system, truck and trailer docking system,
rotational pendulum, etc., are having some of its state variables
in terms of sin function and hence can be better approximated
using this TSFCNC.
3.3. Inverted pendulum an application
A famous benchmark problem namely the inverted
pendulum control problem is chosen for study. The nonlinear
and non-stable behavior of the inverted pendulum problem
renders the use of conventional method very difcult.
The control objective is to balance the inverted pendulum
for the approximate range of vertical angle, namely
x
1
2p=2; p=2. The equations of motion of the pendulum
[17] are
x
2

g sin x
1
amlx
2
2
sin 2x
2
=2 a cos x
1
u
4l=3 aml cos
2
x
1

(11)
where x
1
denotes the angle (in radians) of the pendulum from
the vertical, x
2
x
1
the angular velocity, g the gravity constant,
m the mass of the pendulum, M the mass of the cart, 2l is the
length of the pendulum, u the force applied to the cart and
a 1=m M. The values used in the simulation are,
g 9:8 m/s
2
, m 2 kg, M 8 kg, a 0:1, 2l 1 m.
By neglecting the higher order terms Eq. (11) can be written
as
x
2

g sin x
1

4l=3 aml cos

2
x
1

a cos x
1
u
4l=3 aml cos
2
x
1

(12)
3.4. Modeling of TSFMLC and design of TSFCLC
The system (12) is approximated by the following two-rule
TakagiSugeno fuzzy model with linear consequence:
R
1
: If x
1
is about 0 then
x
1
x
2
_ _

0 1
g
4l=3 aml
0
_ _
x
1
x
2
_ _

0
a
4l=3 aml
_ _
u
R
2
: If x
1
is about p=2 then
x
1
x
2
_ _

0 1
2g
p4l=3 aml cos
2
88

0
_ _
x
1
x
2
_ _

0
a cos 88

4l=3 aml cos

2
88

_ _
u
By utilizing the concept of parallel distributed compensation
(PDC), the following two rules are designed for the controller:
R
1
: If x
1
is about 0 then u
1
K
1
x.
R
2
: If x
1
is about p=2 then u
2
K
2
x.
3.5. Modeling of TSFMNC and design of TSFCNC
Eq. (12) can also be written as
x
1
x
2
_ _

0 1
g
4l=3 aml cos
2
x
1

0
_ _
sin x
1

x
2
_ _

0
a cos x
1

4l=3 aml cos

2
x
1

_ _
u (13)
It is clear from the Eq. (13) that Eq. (12) is expressable in the
form of x Axcx Bxu. Now the system (13) is
approximated by the following two-rule TakagiSugeno fuzzy
model with nonlinear consequence:
R
1
: If x
1
is about 0 then
x
1
x
2
_ _

0 1
g
4l=3 aml
0
_ _
sin x
1

x
2
_ _

0
a
4l=3 aml
_ _
u
R. Rajesh, M.R. Kaimal / Applied Soft Computing 7 (2007) 772782 776
R
2
: If x
1
is about p=2 then
x
1
x
2
_ _

0 1
g
4l=3 aml cos
2
88

0
_ _
sin x
1

x
2
_ _

0
a cos 88

4l=3 aml cos

2
88

_ _
u
By utilizing the concept of parallel distributed compensation
(PDC), the following two rules are designed for the controller
where cx sin x
1
; x
2

T
:
R
1
: If x
1
is about 0 then u
1
K
1
cx.
R
2
: If x
1
is about p=2 then u
2
K
2
cx.
3.6. Comparison
By chossing Q 100 0; 0 0 and R 1 and by using
linear quadratic regulator design method, the gain values of
TSFCLC are obtained as K
1
196:5089 47:1922; K
2

3575:4 1168:6 and the gain values of TSFCNC are
obtained as K
1
196:5089 47:1922; K
2
5616:1
1464:7. The performance index is evaluated by
J
timestep
t
_
t
0
x
T
tQxt u
T
tRut (14)
Table 1 shows the performance index of the TSFCLC and
TSFCNC for a simulation of the pendulum for 3 s, with
integration timestep 0:01, starting with initial angle,
x
1
85

. Fig. 7 shows the plot of u for both TSFCLC and

TSFCNC. The results shows better performance for the new
TSFCNC.
4. Dynamic positioning of ships a real-life example
The model of dynamic positioning of ships (DPS) [6
8,10,11,14] exhibits nonlinear interaction in three degrees of
freedom (surge, sway, and yaw). The DPS control problem is
usually solved under the assumption that the kinematic
equations can be linearized about a constant yaw angle such
that linear theory and gain scheduling techniques can be
applied. Later, it was solved by globally exponentially stable
(GES) nonlinear control law in [7] and solved using Takagi
Sugeno fuzzy controller with linear consequence in [5].
The state space model of the normalized DPS system [5,7] is
as given below:

X AX Bu (15)
where
X
x
1
x
2
x
3
x
4
x
5
x
6
_

_
_

_
B
0 0 0
0 0 0
0 0 0
h
311
h
312
h
313
h
321
h
322
h
323
h
331
h
332
h
333
_

_
_

_
u
u
1
u
2
u
3
_
_
_
_
(16)
where x x
1
x
2
x
3
x
4
x
5
x
6

T
x y fu v r
T
. x
1
x; x
2
y
is the earth-xed position, x
3
f is the yaw angle, x
4
u is
the surge velocity, x
5
v is the sway velocity, and x
6
r is
the yaw velocity. The control thrust in the surge, sway and
yaw modes are respectively, t
1
; t
2
, and t
3
where u
u
1
u
2
u
3

T
t
1
t
2
t
3

T
.
In this section, TakagiSugeno fuzzy controller with
nonlinear consequence (TSFCNC) is used to design fuzzy
controller for the dynamic positioning of ships and is compared
with TakagiSugeno fuzzy controller with linear consequence
(TSFCLC).
R. Rajesh, M.R. Kaimal / Applied Soft Computing 7 (2007) 772782 777
Fig. 7. Performance comparision of pendulum angle u for TSFCLC and
TSFCNC.
Table 1
Performence of TSFCLC and TSFCNC
Performance index TSFCLC TSFCNC
J 2.9634e+005 2.9455e+005
AX
0 0 0 cos x
3
x
4
sin x
3
x
5
0
0 0 0 sin x
3
x
4
cos x
3
x
5
0
0 0 0 0 0 1
h
111
x
1
h
112
x
2
h
113
x
3
h
211
x
4
h
212
x
5
h
213
x
6
h
121
x
1
h
122
x
2
h
123
x
3
h
221
x
4
h
222
x
5
h
223
x
6
h
131
x
1
h
132
x
2
h
133
x
3
h
231
x
4
h
232
x
5
h
233
x
6
_

_
_

_
(17)
4.1. TSFCLC for DPS
The TakagiSugeno fuzzy controller with linear conse-
quence with three rules obtained by approximating AX at yaw
angle 0; p=2 and p=2 are as given below:
R
1
: IF x
3
is about 0 THEN x A
1
x B
1
u
R
2
: IF x
3
is about p=2 THEN x A
2
x B
2
u
R
3
: IF x
3
is about p=2 THEN x A
3
x B
3
u
where a sin 2

; b cos 88

and
A
1

0 0 0 1 a 0
0 0 0 a 1 0
0 0 0 0 0 1
h
111
h
112
h
113
h
211
h
212
h
213
h
121
h
122
h
123
h
221
h
222
h
223
h
131
h
132
h
133
h
231
h
232
h
233
_

_
_

_
and
A
2

0 0 0 b 1 0
0 0 0 1 b 0
0 0 0 0 0 1
h
111
h
112
h
113
h
211
h
212
h
213
h
121
h
122
h
123
h
221
h
222
h
223
h
131
h
132
h
133
h
231
h
232
h
233
_

_
_

_
(18)
A
3

0 0 0 b 1 0
0 0 0 1 b 0
0 0 0 0 0 1
h
111
h
112
h
113
h
211
h
212
h
213
h
121
h
122
h
123
h
221
h
222
h
223
h
131
h
132
h
133
h
231
h
232
h
233
_

_
_

_
and
B
1
B
2
B
3

0 0 0
0 0 0
0 0 0
h
311
h
312
h
313
h
321
h
322
h
323
h
331
h
332
h
333
_

_
_

_
(19)
Parallel distributed compensation is used to design fuzzy
controller rules and are as given below:
R
1
: IF x
3
is about 0 THEN u F
1
x
R
2
: IF x
3
is about p=2 THEN u F
2
x
R
3
: IF x
3
is about p=2 THEN u F
3
x
4.2. TSFCNC for DPS
The Eq. (17) representing AX in the state space model (15)
can also be written as
which gives importance to the variables x
5
and x
4
by providing it
in the 3rd column of 1st and 2nd, respectively, and leads to non-
linear consequent. It is clear that the approximation of the above
AX requires approximations in x
3
; x
4
and x
5
. By assuming the
range of yawangle (x
3
), the surgevelocity(x
4
) and the swayvelo-
city (x
5
), respectively, as p=2; p=2; 10; 10 and 6; 6, the
12 rules of TakagiSugeno fuzzy controller with nonlinear con-
sequence are obtainedat different regions andare as given below:
R
i
: IF x
3
is about j
1;i
and IF x
4
is about j
2;i
and IF x
5
is about
j
3;i
THEN x A
i
c B
i
u i 1; 2; . . . ; 12
where c x
1
; x
2
; sin x
3
; x
4
; x
5
; x
6

T
; j
1;1
j
1;2
j
1;3

j
1;4
p=2; j
1;5
j
1;6
j
1;7
j
1;8
0; j
1;9
j
1;10
j
1;11
j
1;12
p=2; j
2;1
j
2;2
j
2;5
j
2;6
j
2;9
j
2;10
10;
j
2;3
j
2;4
j
2;7
j
2;8
j
2;11
j
2;12
10; j
3;1
j
3;3

j
3;5
j
3;7
j
3;9
j
3;11
6j
3;2
j
3;4
j
3;6
j
3;8
j
3;10
j
3;12
6, and
A
i

0 0 M
1;i
M
2;i
M
3;i
0
0 0 M
4;i
M
5;i
M
6;i
0
0 0 0 0 0 1
h
111
h
112
d
i
h
113
h
211
h
212
h
213
h
121
h
122
d
i
h
123
h
221
h
222
h
223
h
131
h
132
d
i
h
133
h
231
h
232
h
233
_

_
_

_
and
B
i

0 0 0
0 0 0
0 0 0
h
311
h
312
h
313
h
321
h
322
h
323
h
331
h
332
h
333
_

_
_

_
(21)
R. Rajesh, M.R. Kaimal / Applied Soft Computing 7 (2007) 772782 778
AX
0 0
1
2
x
5
cos x
3

1
2
sin x
3
0
0 0
1
2
x
4
1
2
sin x
3
cos x
3
0
0 0 0 0 0 1
h
111
h
112
h
113
x
3
sin x
3

h
211
h
212
h
213
h
121
h
122
h
123
x
3
sin x
3

h
221
h
222
h
223
h
131
h
132
h
133
x
3
sin x
3

h
231
h
232
h
233
_

_
_

_
x
1
x
2
sin x
3

x
4
x
5
x
6
_

_
_

_
(20)
and d
i
p=2 for i 14; 912 and d
i
1 for i 58 and
M
1;1
M
1;3
M
1;5
M
1;7
M
1;9
M
1;11

6
2
M
1;2
M
1;4
M
1;6
M
1;8
M
1;10
M
1;12

6
2
M
2;1
M
2;2
M
2;3
M
2;4
M
2;9
M
2;10
M
2;11
M
2;12
b
M
2;5
M
2;6
M
2;7
M
2;8
1
M
3;1
M
3;2
M
3;3
M
3;4

1
2
M
3;5
M
3;6
M
3;7
M
3;8

a
2
M
3;9
M
3;10
M
3;11
M
3;12

1
2
M
4;1
M
4;2
M
4;5
M
4;6
M
4;9
M
4;10

10
2
M
4;3
M
4;4
M
4;7
M
4;8
M
4;11
M
4;12

10
2
M
5;1
M
5;2
M
5;3
M
5;4

1
2
M
5;5
M
5;6
M
5;7
M
5;8

a
2
M
5;9
M
5;10
M
5;11
M
5;12

1
2
M
6;1
M
6;2
M
6;3
M
6;4
M
6;9
M
6;10
M
6;11
M
6;12
b
M
6;5
M
6;6
M
6;7
M
6;8
1
Parallel distributed compensation is used to design fuzzy
controller rules and are as given below:
R
i
: IF x
3
is about j
1;i
and IF x
4
is about j
2;i
and IF x
5
is about
j
3;i
THEN u K
i
c i 1; 2; . . . ; 12.
where c x
1
; x
2
; sin x
3
; x
4
; x
5
; x
6

T
.
4.3. Comparison
For the following values of Q and R, linear quadratic
regulator techniques are applied to nd out the gain values
R. Rajesh, M.R. Kaimal / Applied Soft Computing 7 (2007) 772782 779
Table 2
Gain values of TSFCLC
Gain Value
F
1 0:9613 0:0348 0:0001 1:3614 0:0002 0:0001
0:0346 0:9665 0:1144 0:0000 1:8917 0:3115
0:0041 0:1143 0:9934 0:0002 0:4727 0:5725
_
_
_
_
F
2 0:0014 0:9996 0:0043 1:3890 0:0042 0:0016
0:9937 0:0110 0:1124 0:0011 1:9188 0:3168
0:1125 0:0001 0:9937 0:0059 0:4721 0:5725
_
_
_
_
F
3 0:0014 0:9996 0:0043 1:3890 0:0042 0:0016
0:9937 0:0110 0:1124 0:0011 1:9188 0:3168
0:1125 0:0001 0:9937 0:0059 0:4721 0:5725
_

_
_

_
Fig. 8. Performance comparison of x
3
for TSFCLC and TSFCNC.
Fig. 9. Performance comparison of x
4
for TSFCLC and TSFCNC.
Fig. 10. Performance comparison of x
5
for TSFCLC and TSFCNC.
of F
i
s in TSFCLC and K
i
s in TSFCNC and are given in
Tables 2 and 3, respectively:
Q
1 0 0 0 0 0
0 1 0 0 0 0
0 0 1 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 0 0 0
_

_
_

_
and R
1 0 0
0 1 0
0 0 1
_
_
_
_
The performance index J
1
and J
2
given by the following
equations are shown in Table 4. It shows that TSFCNC per-
forms better than TSFCLC:
J
1

t0:01:0:01:2
x
1
t x
2
t x
3
t x
4
t x
5
t x
6
t
(22)
R. Rajesh, M.R. Kaimal / Applied Soft Computing 7 (2007) 772782 780
Table 3
Gain values of TSFCNC
Gain Value
K
1 0:5723 0:4171 0:1642 0:3910 0:7157 0:1786
0:7362 0:1749 0:7306 0:3188 0:8662 0:0060
0:4183 0:8866 3:3640 0:2950 0:0094 0:9011
_

_
_

_
K
2 0:6258 0:4692 0:1573 0:5234 0:7228 0:1249
0:4106 0:4491 0:6245 0:4113 0:5848 0:0425
0:6254 0:7759 3:4878 0:1514 0:3387 0:9773
_

_
_

_
K
3 0:6358 0:4658 0:1733 0:5218 0:7200 0:1797
0:6103 0:1138 0:7175 0:3208 0:6521 0:0501
0:4170 0:8810 3:3548 0:2969 0:0162 0:8959
_

_
_

_
K
4 0:5597 0:4176 0:1494 0:3918 0:7177 0:1239
0:5257 0:5047 0:6091 0:4084 0:7924 0:0024
0:6737 0:7375 3:4764 0:1507 0:3959 0:9891
_

_
_

_
K
5 0:5723 0:4171 0:1642 0:3910 0:7157 0:1786
0:7362 0:1749 0:7306 0:3188 0:8662 0:0060
0:4183 0:8866 3:3640 0:2950 0:0094 0:9011
_

_
_

_
K
6 0:6258 0:4692 0:1573 0:5234 0:7228 0:1249
0:4106 0:4491 0:6245 0:4113 0:5848 0:0425
0:6254 0:7759 3:4878 0:1514 0:3387 0:9773
_

_
_

_
K
7 0:6358 0:4658 0:1733 0:5218 0:7200 0:1797
0:6103 0:1138 0:7175 0:3208 0:6521 0:0501
0:4170 0:8810 3:3548 0:2969 0:0162 0:8959
_

_
_

_
K
8 0:5597 0:4176 0:1494 0:3918 0:7177 0:1239
0:5257 0:5047 0:6091 0:4084 0:7924 0:0024
0:6737 0:7375 3:4764 0:1507 0:3959 0:9891
_

_
_

_
K
9 0:5723 0:4171 0:1642 0:3910 0:7157 0:1786
0:7362 0:1749 0:7306 0:3188 0:8662 0:0060
0:4183 0:8866 3:3640 0:2950 0:0094 0:9011
_

_
_

_
K
10 0:6258 0:4692 0:1573 0:5234 0:7228 0:1249
0:4106 0:4491 0:6245 0:4113 0:5848 0:0425
0:6254 0:7759 3:4878 0:1514 0:3387 0:9773
_

_
_

_
K
11 0:6358 0:4658 0:1733 0:5218 0:7200 0:1797
0:6103 0:1138 0:7175 0:3208 0:6521 0:0501
0:4170 0:8810 3:3548 0:2969 0:0162 0:8959
_

_
_

_
K
12 0:5597 0:4176 0:1494 0:3918 0:7177 0:1239
0:5257 0:5047 0:6091 0:4084 0:7924 0:0024
0:6737 0:7375 3:4764 0:1507 0:3959 0:9891
_

_
_

_
J
2

t0:01:0:01:2
u
1
t u
2
t u
3
t (23)
The dotted lines in Figs. 811 show the performance of
TSFCLCand the solid lines showthe performance of TSFCNC.
The gures show that TSFCNL performs better than TSFCLC.
The plot of x
1
versus y
1
, shown in Fig. 12, shows the path of the
ship from the earth xed position 10; 10 to the position (0,
0). Since the dotted line representing TSFCLC and solid line
representing TSFCNL concides, it seems to be a straight solid
line. But at the nishing point, after zooming it is clear that
TSFCNC performs better than TSFCLC.
5. Conclusion
This paper presents TS fuzzy model with nonlinear
consequence. It is proved that it can approximate a particular
class of nonlinear functions, nonlinear dynamic systems and
nonlinear control systems. Moreover, TSFCNC can approx-
imate state feedback controllers by using parallel distributed
compensation method. A benchmarking control problem,
namely, inverted pendulum is simulated and the results show
that TSFCNC performs better than TSFCLC. A real-life
example of dynamic positioning of ship is simulated and the
results also show that TSFCNC performs better than
TSFCLC.
Acknowledgements
The authors are very thankful to Indian Space Research
Organization for nancial support through RESPOND project.
The authors are also thankful to the reviewers for their valuable
help in improving the quality of the paper.
References
[1] J.J. Buckley, Universal fuzzy controllers, Automatica 28 (1992) 1245
1248.
[2] O. Castillo, P. Melin, Special issue on soft computing for control of non-
linear dynamical systems, Appl. Soft Comput. 3 (2003) 303304.
[3] O. Castillo, P. Melin, Intelligent adaptive model-based control of robotic
dynamic systems with a hybrid fuzzy-neural approach, Appl. Soft Com-
put. 3 (2003) 363378.
[4] J.L. Castro, Fuzzy logic controllers are universal approximators, IEEE
Trans. Syst. Man Cybern. 25 (1998) 629635.
[5] W.-J. Chang, G.-J. Chen, Y.-L. Yeh, Fuzzy control of dynamic positioning
of ships, J. Mar. Sci. Technol. 10 (2002) 4753.
[6] T.I. Fossen, Guidance and Control of Ocean Vehicles, Wiley, New York,
1994.
[7] T.I. Fossen, A. Grovlen, Nonlinear output feedback control of dynamically
positioned ships using vectorial observer backstepping, IEEE Trans.
Control Syst. Technol. 6 (1998) 121128.
[8] J.M. Godhavn, T.I. Fossen, S.P. Berge, Nonlinear and adaptive back-
stepping designs for tracking control os ships, Int. J. Adapt. Control Signal
Process. 12 (1998) 649670.
[9] M. Golob, Decomposed fuzzy proportional-integral-derivative control-
lers, Appl. Soft Comput. 1 (2001) 201214.
[10] M.-H. Kim, Nonlinear control and robust observer design for marine
vehicles, Ph.D. thesis, Virgina Polytechnic Institute and State University,
2000.
[11] A. Loria, T.I. Fossen, E. Panteley, A separation principle for dynamic
positioning of ships: theoretical and experimental results, IEEE Trans.
Control Syst. Technol. 8 (2000) 332343.
[12] E.H. Mamdani, Application of fuzzy algorithms for control of simple
dynamic plant, Proc. Inst. Electr. Eng. 121 (1974) 15851588.
[13] P. Melin, O. Castillo, Adaptive intelligent control of aircraft systems with a
hybrid approach combining neural networks, fuzzy logic and fractal
theory, Appl. Soft Comput. 3 (2003) 353362.
[14] K.Y. Pettersen, H. Nijmeijer, Global practical stabilization and tracking
for underactuated ship a combined averaging and backstepping
approach, in: Proceedings of the IFAC Conference on System Structure
and Control, Nantes, France, 1998, pp. 5964.
[15] T. Takagi, M. Sugeno, Fuzzy identication of systems and its applications
to modeling and control, IEEE Trans. Syst. Man Cybern. 15 (1985) 116
132.
R. Rajesh, M.R. Kaimal / Applied Soft Computing 7 (2007) 772782 781
Fig. 11. Performance comparison of x
6
for TSFCLC and TSFCNC.
Fig. 12. Performance comparison of x
1
y
1
for TSFCLC and TSFCNC.
Table 4
Performance of TSFCLC and TSFCNC
Performance index TSFCLC TSFCNC
J
1
200.6146 200.4305
J
2
189.0391 54.5321
[16] H.O. Wang, J. Li, K. Tanaka, TS Fuzzy model with linear rule con-
sequence and PDC controllers: a universal framework for nonlinear
control systems, Int. J. Fuzzy Syst. (2003) 106113.
[17] H.O. Wang, K. Tanaka, M. Grifn, Parallel distributed compensation of
nonlinear systems by TakagiSugeno Fuzzy Model, in: Proc. Fuzz-IEEE/
IFES95 (1995) 531538.
[18] Y. Yang, C. Zhou, J. Ren, Model reference adaptive robust fuzzy control
for ship steering autopilot with uncertain nonlinear systems, Appl. Soft
Comput. 3 (2003) 305316.
[19] L.A. Zadeh, Fuzzy Sets, Inform. Control 8 (1965) 338353.
R. Rajesh, M.R. Kaimal / Applied Soft Computing 7 (2007) 772782 782