International Journal of Engineering Research& Science(IJOER)

ISSN: [2395-6992]

[Vol-2, Issue-2, February- 2016]

Controller Design for Nonlinear Networked Control Systems

The Key Lab for IOT and Information Fusion Technology of Zhejiang, the Institute of Information and Control,
Hangzhou Dianzi University, China
2,3,4,5
School of Engineering, Huzhou University, China

Abstract This paper investigates the state feedback control problem of nonlinear networked control systems in
presence of data packet dropout. The random data packet dropout is modeled as a Bernoulli distributed sequence
with a known conditional probability distribution. The closed-loop system model is obtained through the state
augmentation method. The sufficient conditions for the existence of a state feedback controller are derived by means
of Lyapunov stability theory such that the closed-loop control system is exponentially stable in the mean-square sense.
The maximum bound of the nonlinear item of the control system is also obtained. Then a linear matrix inequality
(LMI) based approach for designing such a controller which can be solved by Matlab LMI tool box is presented.
Finally, a numerical simulation example is used to demonstrate the effectiveness of the proposed method.
KeywordsData Packet Dropout, Exponentially Mean-square Stable, Linear Matrix Inequality, Nonlinear.

I.

INTRODUCTION

In networked control system (NCS), control loops are closed by real-time networks. Such kind of control systems bring a lot
of advantages, such as reduced system wiring, simple system maintenance and low cost. In the field of industrial control,
remote medical treatment, military and other fields, NCS has been widely used [1-2].However, the application of
communication networks in feedback control loops makes the NCS analysis and synthesis complex because data packets
through networks suffer not only transmission delays but also data packet drop out [3-4].Many scholars have investigated the
effect of data packet dropout on the performance of the control system.
In recent years, there are many results for the controller design of linear network control systems respect to the data packet
dropout. For the NCS with random data packet drop out, literature [5-6] modeled NCS as a kind of Markov jump system, and
the mode dependent controller is designed. In literature [7], if the data packet drop out accrues, the input of the actuator was
considered to be zero, and the optimal estimation of NCS was studied. For a class of singular controlled plant, literature [8]
modeled closed-loop system as a kind of asynchronous dynamical system, and the control problem was investigated.
However, there is not much work reported about nonlinear NCS. The state feedback control problem for nonlinear NCS is
still very limited.
In this paper, the state feedback control problem for a kind of nonlinear NCS is investigated. Assuming that the data packet
dropout obeys Bernoulli distribution, the closed-loop system is modeled as a class of control system which contains
stochastic variable. The sufficient condition and design method of the state feedback controller for a class of nonlinear NCS
with random data packet drop out are given. Simulation results show that the method proposed in this paper is effective.

II.

PROBLEM FORMULATION

A nonlinear controlled plant is given by:

x(k 1) Ax(k ) Bu(k ) f (k , x(k ))

(1)

Where x ( k ) R is system state vector and u ( k ) R is the system control input. A and B are all real constant matrices.
n

f ( k , x ( k )) is nonlinear disturbance, satisfying the following Lipschitz condition:

f T (k ,x(k ))f(k ,x(k )) 2 x T (k )H T Hx(k )

(2)

Where His real constant matrix, and 0is the bound of the nonlinear disturbance.
We consider the circumstance that data packet drop out occurs between sensor and controller. A stochastic variable is defined
Page | 36

International Journal of Engineering Research& Science(IJOER)

ISSN: [2395-6992]

[Vol-2, Issue-2, February- 2016]

as follows to describe the process of data packet drop out :

0 data packet dropout occurs

Prob{ (k ) 1} E{ (k )}: a , Prob{ (k ) 0} 1 E{ (k )}: 1 a , Var{ (k )} E{( (k ) a) 2 } a(1 a) : b 2 .

x(k 1) Ax(k ) (k ) Bu(k ) (1 (k )) Bu(k 1) f (k , x(k ))

(3)

u (k ) Kx(k )

(4)

Assume the state feedback control law is:

Substituting (4) into (3), we have the following closed-loop system:

x(k 1) ( A (k ) BK ) x(k ) (1 (k )) BKx(k 1) f (k , x(k ))

(5)

At sampling time k , if we augment the state-variable as (k )T x(k )T

x(k 1)T ,the closed-loop system (5) can be

(k 1) (k ) (k ) EF (k , (k ))

(6)

Where

A (k ) B (1 (k )) B
I
( k )
, E , F (k , (k )) f (k , G (k )) , G I 0 .
I
0

Before proceeding, we need the following definition.

0 and

0 1 such that {|| k ||2 } k {|| k ||2 } holds for all.

xT T1 x 0 such that xT T2 x 0 hold if there exist a scalar

0 such that T1 T2 0 .
III.

CONTROLLER DESIGN

Theorem 1 Take as given communication channel parameters Prob{ (k ) 0} 1 a ,The closed-loop NCS(6) is
exponentially mean-square stable if there exist positive definite matrices P,Q and nonnegative real scalar 0 such that the
following matrix inequality holds:

11 2 H T H

21

31

*
22
32

* 0
33 I

(7)

where

11 ( BK )T P( A aBK ) b 2 ( BK )T PBK Q P , 21 (1 a)( BK )T P( A aBK ) b 2 ( BK )T PBK , 31 IP( A aBK ) ,

Page | 37

International Journal of Engineering Research& Science(IJOER)

ISSN: [2395-6992]

[Vol-2, Issue-2, February- 2016]

Proof:
Define a Lyapunov function

P 0
V (k ) (k )T S (k ) (k )T
(k )T

0 Q
T
T
x(k ) Px(k ) x(k 1) Qx(k 1)
Where P and Q are positive definite matrices, we have:

E{Vk 1} E{Vk }
E{[( A (k ) BK ) x(k ) (1 (k ))BKx (k 1) f (k , x(k )]T P
[( A (k ) BK ) x(k ) (1 (k ))BKx (k 1) f (k , x(k )]
x(k )T Qx(k ) x(k )T Px (k ) x(k 1)T Qx(k 1)
E{[( A BK ) x(k ) ( (k ) ) BKx (k ) ( (k ) ) BKx (k 1) (1 ) BKx (k 1) f (k , x(k )]T P
[( A BK ) x(k ) ( (k ) ) BKx (k ) ( (k ) ) BKx (k 1) (1 ) BKx (k 1) f (k , x(k )]
x(k )T Qx(k ) x(k )T Px (k ) x(k 1)T Qx(k 1)
11
= (k )T 21
31
Where (k )T x(k )T

x(k 1)T

*
22
32

*
* (k )
33

(8)

f (k , x(k ))T .

From (2), we can get:

2 H T H

(k )T
0

* *

0 * (k ) 0
0 I

(9)

According to lemma 1, if there exists a scalar 0 such that

11 2 H T H

21

31

*
22
32

* 0
33 I

*
22
32

*
* 0
33

(10)

then the following inequality subject to (9) holds.

11

21
31
Since S 0 ,it is clear that
()

()

that +1
+1

. Because { Tk 1S k 1} { Tk S k} 0 , there exists0 < < max ()such

+ 1

min (s) {|| k ||2 } {kT Sk } (1

max (S )

()

.Consequently, we can get

) { kT S k }
Page | 38

International Journal of Engineering Research& Science(IJOER)

(1

max (S )

That is {|| k
where

) k {0T S0 } max (S )(1

max (S )

ISSN: [2395-6992]

[Vol-2, Issue-2, February- 2016]

) k {|| 0 ||2 }

||2 } k {|| k ||2 } ,

max ( S )

, 1
.
min ( S )
max ( S )

Theorem 2 Take as given communication channel parameters Prob{ (k ) 0} 1 a , the systems (6) is exponentially mean
square stable if there exist positive definite matrices M>0,N>0, matrix Y and a scalar 0 satisfying the following LMI

*
0

(11)

where

AM
diag ( N M , N , I ) , diag ( M , M , I ) , bBY
HM

(1 a) BY
bBY
0

I
0 and the controller is given by u (k ) YM 1 x(k ) .
0

Proof:
(7) Is equivalent to

11 2 H T H

21

31

*
22

32

* 0
33 I

(12)

Where

( BK )T P ( A aBK ) b2 ( BK )T PBK
(1 a)( BK )T P ( A aBK ) b2 ( BK )T PBK
,

Q P ,
11
21

IP ( A aBK ) ,
(1 a) IPBK
P ,
(1 a)2 ( BK )T PBK
,

b2 ( BK )T PBK
Q ,
32
31
32
22
P
Q
P , Q ,

And (12) can be written as

Q P 2 H T H
b2 ( BK )T PBK
*
*

2
T
2
T
b ( BK ) PBK
b ( BK ) PBK Q *

IP( A aBK )
0
I

( A aBK )T

T
(1 a)( BK ) P A aBK

(1 a)( BK ) I 0 .

By Schur complement, one can obtain

Q P 2 H T H
b2 ( BK )T PBK
*
*
*

2
T
2
T
b ( BK ) PBK
b ( BK ) PBK Q *
*

IP( A aBK )
0
I *

A aBK
(1 a) BK
I P 1

Similarly, using Schur again we can get

11 *

0
21 22

(13)

Page | 39

International Journal of Engineering Research& Science(IJOER)

ISSN: [2395-6992]

[Vol-2, Issue-2, February- 2016]

Where

11 diag (Q P, Q, I ) , 22 diag ( P, P, I ) , 21 bBK

*
bBK
0

*
* .
0

Performing a congruence transformation to (13) by diag{P 1 , P 1 , I , I , I , I , I } , and let M P 1 , N MQM , Y KM

2 , (11) can be obtained, which complete the proof.

>0,>0,>0,

. . (11)

and the solution is ,the maximum bound of nonlinear disturbance in (2)is =

Remark 1 the above convex optimization with LMI constraints can be solved by Matlab LMI toolbox objective function
minimization problem solver mincx.

IV.

SIMULATION EXAMPLE

Consider the following nonlinear controlled plant

0.1x1 (k )sin x1 (k )
0.2 x (k ) cos x (k ) .

2
2

Assume that the initial state of the system is x(0) 3 1 and the Probability of the data packet dropout is
T

Pr ob{ (k ) 0} 0.2 and the matrix H is a unit matrix. According to Theorem 2, and using Matlab LMI tool, we can get

40.0981 29.5003
Y 0.4610 0.2155 , M
. Hence, the controller gain matrix is
29.5003 63.9610
K YM 1 0.0212 -0.0131 .
By Matlab LMI toolbox objective function minimization problem solver mincx, the maximum bound of nonlinear
disturbance in (2) is obtained as max 0.2277 . The process of data packet drop out and the state trajectory of the closed-loop
system is shown by the Fig.1 and Fig.2, respectively.
1.5

0.5

-0.5

10

20

30

40

50

60

70

FIG.1 THE RANDOM DATA PACKET DROPOUT DISTRIBUTION

International Journal of Engineering Research& Science(IJOER)

ISSN: [2395-6992]

[Vol-2, Issue-2, February- 2016]

3.5
x1
x2

3
2.5
2
1.5
1
0.5
0
-0.5
-1

10

20

30

40

50

60

70

FIG.2 THE STATE TRAJECTORY OF THE CLOSED-LOOP SYSTEM

V.

CONCLUSION

In this paper, the sufficient condition and design method of the state feedback controller for a class of nonlinear NCS with
random data packet drop out are given. The controller gain matrix and the maximum bound of the nonlinear item of the
control system can be obtained by use of the relevant tool box of Matlab. Simulation results show that the method proposed
in this paper is effective.

VI.

ACKNOWLEDGEMENTS

This work was supported by the National Natural Science Foundation of China under grant nos.
61174029,61573136,61503136.

REFERENCES
[1] K. Y. You, L. H. Xie, Survey of recent progress in networked controlsystems, Acta AutomaticaSinica, Vol.39,No.2,pp.101-118, 2013.
[2] Y. Z. Liu, G. M. Li, H. Du, Robust control of NCS with delay and packet dropout ,Control and Decision,Vol.29, No.3, pp.517522,2014.
[3] D. Wang, J. Wang, and W. Wang, controller design of networked control systems with Markov packet dropouts, IEEE
Transactions on Systems, Man, and Cybernetics: Systems Vol.43, No.3,pp. 689-697. 2013
[4] R. Q. Lu, X. X. Zhou, F. Wu and A. K Xue., Quantized output feedback control for linear discrete-time systems, Journal of the
Franklin Institute,Vol.35,No.8,pp.2096-2108., 2013
[5] J. Wu , T. W. Chen , Design of networked control systems with packet dropouts,IEEE Trans on Automatic control,
Vol.52,No.7,pp.1314-1319, 2007
[6] D. X. Xie ,X. D. Han , et al., state feedback Control for networked control systems with data packet dropout,Systems Engineering
and Electronics,Vol.31,No.3,pp.629-633,2009.
[7] P. Seiler, R. Sengupta, An approach to networked control, IEEE Trans on Automatic control, Vol.50,No.3,pp.356-364 , 2005.
[8] Z. Z. Qiu, L. N. ZHAO,et al, controller design for singular networked control systems with data packet dropout, CAAI
Transactions on Intelligent S`ystems, Vol.10,No.1,pp.120-130, 2015
[9] V. A. Yakubovich, The S-procedure in nonlinear control theory,Vestnik Leningrad University Mathematics, Vol.4,pp.73-74.19

Page | 41