MPE Mathematical Problems in Engineering 1563-5147 1024-123X Hindawi Publishing Corporation 10.1155/2014/749574 749574 Research Article H Stochastic Control of a Class of Networked Control Systems with Time Delays and Packet Dropouts Wang Jufeng 1 http://orcid.org/0000-0003-1792-1714 Liu Chunfeng 2 Li Kai 3 Bakhoum Ezzat G. 1 Department of Mathematics China Jiliang University Hangzhou 310018 China cjlu.edu.cn 2 School of Management Hangzhou Dianzi University Hangzhou 310018 China hdu.edu.cn 3 School of Management Hefei University of Technology Hefei 230009 China hfut.edu.cn 2014 172014 2014 26 01 2014 05 06 2014 1 7 2014 2014 Copyright © 2014 Jufeng Wang et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

This paper studies the H-infinity stochastic control problem for a class of networked control systems (NCSs) with time delays and packet dropouts. The state feedback closed-loop NCS is modeled as a Markovian jump linear system. Through using a Lyapunov function, a sufficient condition is obtained, under which the system is stochastically exponential stability with a desired H-infinity disturbance attenuation level. The designed H-infinity controller is obtained by solving a set of linear matrix inequalities with some inversion constraints. An numerical example is presented to demonstrate the effectiveness of the proposed method.

1. Introduction

In the past few years, the networked control systems (NCSs) whose control loops are connected via communication networks have received increasing attention due to their advantages, such as reduced cost, low weight, easier installation, and maintenance. Time delay and packet dropout are the two major causes of instability of system and deterioration of system performance. Therefore, the time delay and packet dropout problems have been investigated in the existing literature. In , time delays are time-varying in intervals. In [2, 3], the bounds were imposed on the maximum number of successive dropouts. In , the sufficient condition that establishes the quantitative relation between the packet-dropout rate and the stability of the NCS with a constant delay is obtained.

Considering the disturbance attenuation problem, there has been much research effort on H controller design. In , the controller dynamics is continuous, but in many NCSs, the system is controlled by a discrete-time controller with sample and hold devices. In , a discrete-time controller is designed; however, it should be pointed out that the packet dropout or the delay problem is studied separately.

In [11, 12], H control of a class of systems with random packet dropout is investigated. It is noticed that the plant is a discrete-time system and the delay is a multiple of the sampling time; therefore, the result of the papers cannot be applied to the NCSs when the plant is a continuous-time system and the delay is smaller than the sampling period. In , the plant studied is a continuous-time system; the delay takes values in a finite set at a fixed rate. In fact, the time delays and packet dropouts may be random and modeled as Markov chains in most cases. Unfortunately, they do not take into account the time delay and packet dropout with Markovian characterization in .

In this paper, we investigate the H stochastic control of a class of NCSs with time delays and packet dropouts. The random time delay and packet dropout are described by a Markov chain. By using a Lyapunov function, we obtain the system with exponential stability with a desired H disturbance attenuation level. The designed H stochastic controller is obtained by an iterative linear matrix inequality approach. To demonstrate the effectiveness of the method, an illustrative example is presented.

2. Model for Networked Control System

The structure of the NCS is shown in Figure 1. Consider a continuous-time linear system described by (1)x˙=Apx(t)+Bpu(t)+Epw(t),z(t)=Cx(t), where x(t)Rn is the state, u(t)Rm is the plant input, w(t)Rq is the disturbance input, and z(t)Rl is the plant output. Ap, Bp, Ep, and C are constant matrices of appropriate dimensions.

The structure of NCS.

The following assumptions are made for the considered NCS throughout the paper .

The controller is event-driven; both the sensor and the actuator are time-driven. The sampling period of the sensor is T. The actuator has a receiving buffer which contains the most recently updated packet from the controller. The actuator reads the buffer periodically at a smaller period than T, say T0=T/N for some integer N large enough. The sensor and the actuator are time synchronized. Upon reading a new value, the actuator with a zero-order-hold device will update the output value. The network-induced delay τ(k) satisfies 0τ(k)<T.

Based on the above assumptions, the discrete-time state feedback H controller can be expressed as follows: (2)u(k)=Kx^(k), where (3)x^(k)={x(k),ifx(k)issuccessfullytransmitted,x^(k-1),ifx(k)islostduringtransmission, where x(k) is the value of x(t) at the sampling time kT.

Consider (4)z(k)=Cx(k), where z(k) is the value of the z(t) at the sampling time kT.

During each sampling period, several different cases may arise, which leads to the following discrete-time switched system model : (5)x~(k+1)=Aσ(k)x~(k)+E~w(k),z(k)=C~x~(k), where (6)x~(k+1)=[x(k+1)x^(k)],E~=[E0],C~=[C0],(7)Aσ(k)=[A+A1σ(k)KA0σ(k)KI0],llforσ(k)=0,1,2,N-1(8)Aσ(k)=[ABK0I],forσ(k)=N(9)A=exp{ApT},A0σ(k)=T-σ(k)T0Texp{Apτ}Bpdτ,B=0Texp{Apτ}Bpdτ,A1σ(k)=0T-σ(k)T0exp{Apτ}Bpdτ,E=0Texp{Apτ}Epdτ. The σ(k) is called a switching signal. Note that σ(k)=i, i=0,1,,N-1, implies τ(k)=iT0, while σ(k)=N implies packet dropout.

σ ( k ) is modeled as Markov chain that takes values in {0,1,,N-1,N}. The transition probability matrices of σ(k) are Π=[πij]. That means that σ(k) jump from mode i to mode j, from mode with probabilities πij: (10)πij=Pr(σ(k+1)=jσ(k)=i), where πij0 and j=0Nπij=1.

3. <inline-formula><mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" id="M53"><mml:mrow><mml:msub><mml:mrow><mml:mi>H</mml:mi></mml:mrow><mml:mrow><mml:mi>∞</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:math></inline-formula> Disturbance Attenuation Analysis Definition 1.

System (5) is said to be stochastically and exponentially stable, if there exist constants C>0 and 0<λ<1, such that E(x~(k)2)CλkE(x~(0)2) for w(t)0.

Definition 2.

System (5) is said to be stochastically and exponentially stable with an H disturbance attenuation level γ, if system (5) is stochastically and exponentially stable and for the zero initial condition, k=0E{zT(k)z(k)}γ2k=0E{wT(k)w(k)}.

Lemma 3 (see [<xref ref-type="bibr" rid="B14">14</xref>]).

Define Vσ(k)(k)=x~T(k)Pσ(k)x~(k), where Pσ(k) is a positive definite matrix; then there exist constant scalars β1,β2>0 such that (11)β1x~(k)2Vσ(k)(k)β2x~(k)2,  σ(k)=0,1,2,,N.

Theorem 4.

For given positive scalars πij(i,j=0,1,2,,N), λ, and γ, if there exist matrices Pi>0, Qi>0, such that (12)Ωi=[Ci11*Ci21Ci22]<0,i=0,1,,N-1,(13)Ω¯=[C¯11*C¯21C¯22]<0, where (14)Ci11=[j=0NπijQj-λPi+CTC000-λQi000-γ2I],Ci21=[πi0(A+A1iK)πi0A0iKπi0Eπi1(A+A1iK)πi1A0iKπi1EπiN(A+A1iK)πiNA0iKπiNE],Ci22=diag{-P0-1-P1-1-PN-1},C¯11=[-λPN+CTC0j=0NπNjPjE*j=0NπNjQj-λQN0**j=0NπNjETPjE-γ2I],C¯21=[πN0AπN0BK0πN1AπN1BK0πNNAπNNBK0],C¯22=diag{-P0-1-P1-1-PN-1} then system (5) is stochastically and exponentially stable with an H disturbance attenuation level γ.

Proof.

Let the Lyapunov function (15)Vi(k)=xT(k)Pix(k)+x^T(k-1)Qix^(k-1)=x~T(k)P~ix~(k) correspond to the subsystem as follows: (16)x~(k+1)=Aix~(k)+E~w(k),z(k)=C~x~(k), where (17)P~i=[Pi00Qi].

When σ(k)=i    (i=0,1,2,,N-1), we obtain (18)E(Vσ(k+1)(k+1)-λVσ(k)(k)+zT(k)z(k)-γ2wT(k)w(k))=E(Vσ(k+1)(k+1)σ(k)=i)-λVi(k)+x~T(k)C~TC~x~(k)-γ2wT(k)w(k)=j=0Nπij(x~T(k)AiT+wT(k)E~T)P~j(Aix~(k)+E~w(k))-λx~T(k)P~ix~(k)+xT(k)C~TC~x(k)-γ2wT(k)w(k)=x~T(k)(j=0NπijAiTP~jAi)x~(k)+x~T(k)×(j=0NπijAiTP~jE~)w(k)+wT(k)(j=0NπijE~TP~jAi)x~(k)+wT(k)(j=0NπijE~TP~jE~)w(k)-λx~T(k)P~ix~(k)+xT(k)C~TC~x(k)-γ2wT(k)w(k)=[x~T(k)wT(k)]Θi[x~(k)w(k)], where (19)Θi=[j=0NπijAiTP~jAi-λP~i+C~TC~j=0NπijAiTP~jE~j=0NπijE~TP~jAij=0NπijE~TP~jE~-γ2I].

From (6) and (7), it can be obtained that (3)Θi=[Ai11Ai12Ai13*Ai22Ai23**Ai33].Ai11=j=0Nπij[(AT+KTA1iT)Pj(A+A1iK)+Qj]-λPi+CTC,Ai12=j=0Nπij(AT+KTA1iT)PjA0iK,Ai13=j=0Nπij(A+A1iK)TPjE,Ai22=j=0NπijKTA0iTPjA0iK-λQi,Ai23=j=0Nπij(A0iK)TPjE,Ai33=j=0NπijETPjE-γ2I.

Θ i < 0 can be rewritten as follows: (21)Φi+[πi0(AT+KTA1iT)πi0KTA0iTπi0ET]×P0[πi0(A+A1iK)πi0A0iKπi0E]<0, where (22)Φi=[Bi11Bi12Bi13*Bi22Bi23**Bi33].Bi11=j=1Nπij(AT+KTA1iT)Pj(A+A1iK)+j=0NπijQj-λPi+CTC,Bi12=j=1Nπij(AT+KTA1iT)PjA0iK,Bi13=j=1Nπij(A+A1iK)TPjE,Bi22=j=1NπijKTA0iTPjA0iK-λQi,Bi23=j=1Nπij(A0iK)TPjE,Bi33=j=1NπijETPjE-γ2I.

From the Schur complement, we have that (21) is equivalent to (23)Ψi=[Bi11Bi12Bi13πi0(AT+KTA10T)*Bi22Bi23πi0KTA0iT**Bi33πi0ET***-P0-1]<0.

Similarly, we can see that Ψi<0 is equivalent to (24)Ωi=[Ci11*Ci21Ci22]<0.

It can be seen that if (12) holds, Θi<0 is true, which means (25)E(Vσ(k+1)(k+1)-λVσ(k)(k)+zT(k)z(k)-γ2wT(k)w(k))<0.

When σ(k)=N, (26)E(Vσ(k+1)(k+1)-λVσ(k)(k)+zT(k)z(k)-γ2wT(k)w(k))=[x~T(k)wT(k)]Θ¯[x~(k)w(k)], where (27)Θ¯=[j=0NπNjANTP~jAN-λP~N+C~TC~j=0NπNjANTP~jE~j=0NπNjE~TP~jANj=0NπNjE~TP~jE~-γ2I].

From (6) and (8), it can be seen that (28)Θ¯=[A¯11A¯12A¯13*A¯22A¯23**A¯33],A¯11=j=0NπNjATPjA-λPN+CTC,A¯12=j=0NπNjATPjBK,A¯13=j=0NπNjPjE,      A¯22=j=0NπNjKTBTPjBK+j=0NπNjQj-λQN,A¯23=0,A¯33=j=0NπNjETPjE-γ2I.

Θ ¯ < 0 can be rewritten as follows: (29)Φ¯+[πN0ATπN0KTBT0]P0[πN0AπN0BK0]<0, where (30)Φ¯=[B¯¯11B¯12B¯13*B¯22B¯23**B¯33],B¯11=j=1NπNjATPjA-λPN+CTC,B¯12=j=1NπNjATPjBK,B~13=j=0NπNjPjE,B¯22=j=1NπNjKTBTPjBK+j=0NπNjQj-λQN,B¯23=0,B¯33=j=0NπNjETPjE-γ2I.

From the Schur complement, we have that (29) is equivalent to (31)Ψ¯=[B¯11B¯12B¯13πN0AT*B¯22B23πN0KTBT**B¯330***-P0-1]<0.

Similarly, it is easy to see that Ψ¯<0 is equivalent to (32)Ω¯=[C¯11*C¯21C¯22]<0.

It can be seen that if (13) holds, Θ¯<0 is true, which means (33)E(Vσ(k+1)(k+1)-λVN(k)+zT(k)z(k)-γ2wT(k)w(k))<0.

It follows from (25) and (33) that (34)E(Vσ(k+1)(k+1)-λVσ(k)(k)+zT(k)z(k)-γ2wT(k)w(k))<0 which means (35)E(Vσ(k+1)(k+1))<E(Vσ(k)(k))-E(zT(k)z(k))+γ2E(wT(k)w(k)),E(Vσ()(k+1))<E(Vσ(0)(0))-k=0E(zT(k)z(k))+γ2k=0E(wT(k)w(k)),k=0E(zT(k)z(k))γ2k=0E(wT(k)w(k)).

Next, we prove the stochastically and exponentially stable system (5). The perturbation w(t) is assumed to be zero.

When σ(k)=i,  (i=0,1,2,,N-1), we obtain (36)E(Vσ(k+1)(k+1)-λVσ(k)(k))=E(Vσ(k+1)(k+1)σ(k)=i)-λVi(k)=x~T(k)(j=0NπijAiTP~jAi-λP~i)x~.

From (21), (23), and (24), it can be seen that Θ<0 is equivalent to Ω<0.

Then, it can be seen from (19) that if (12) holds, we have (37)j=1NπijAiTP~jAi-λP~i+C~TC~<0 and then (38)j=1NπijAiTP~jAi-λP~i<0 which means (39)E(Vσ(k+1)(k+1)-λVσ(k)(k))<0.

When σ(k)=N, (40)E(Vσ(k+1)(k+1)-λVN(k))=x~T(k)(j=0NπNjANTP~jAN-λP~N)x~(k).

From (27), (29), (31), and (32), it can be seen that Θ¯<0 is equivalent to Ω¯<0.

Then, it can be seen from (27) that if (13) holds, we have (41)j=0NπNjANTP~jAN-λP~N+C~TC~<0 and then (42)j=0NπNjANTP~jAN-λP~N<0 which means (43)E(Vσ(k+1)(k+1)-λVN(k))<0.

From (36) and (43), we have (44)E(Vσ(k+1)(k+1))<E(λVσ(k)(k)),E(Vσ(k)(k))<λkE(Vσ(0)(0)).

From Lemma 3, we get (45)E(x~(k)2)CλkE(x~(0)2). Then, the result is established.

The conditions in Theorem 4 are a set of LMIs with some inversion constraints. K can be solved by an iterative LMI approach which is called the cone complementarity linearization algorithm [15, 16].

4. Numerical Example

Consider the following system . Suppose γ=0.91, λ=0.9760. The transition probability matrices of σ(k) are taken as follow: (46)[0.10.800.10.20.700.10.40.500.10.60.300.1] which means (47)π01=0.1,π02=0.8,π03=0,π04=0.1,π11=0.3,π12=0.7,π13=0,π14=0,π21=0.3,π22=0.7,π23=0,π24=0,π31=0.6,π32=0.3,π33=0,π34=0.1. Using Theorem 4 and the cone complementarity linearization algorithm, we obtain (48)K=[-1.4252-5.7880]. Figure 2 is the possible realizations of the mode σ(k). Under this mode sequence, the corresponding state trajectories of the closed-loop system are shown in Figure 3. It is shown that the closed-loop system is stochastically and exponentially stable.

Random mode σ(k).

State trajectories of the NCS.

5. Conclusions

In this paper, by modeling the random delays and packet dropouts as a Markov chain, a new Markovian jump system model is presented to describe the networked control system with disturbance attenuation. The criteria for the system are stochastically and exponentially stable with an H disturbance attenuation level which is derived by an iterative LMI approach.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.

Acknowledgments

This work is supported by Zhejiang Provincial Natural Science Foundation of China (Grant no. LY14G020014), Zhejiang Provincial Key Research Base of Humanistic and Social Sciences in Hangzhou Dianzi University (no. ZD01-201402), the Research Center of Information Technology & Economic and Social Development, and the National Natural Science Foundation of China (no. 71101040). Professor Chunfeng Liu and Kai Li give the authors some useful comments and suggestions on the language and the structure of the paper. The work is supported by Professor Chunfeng Liu and Kai Li.

Wang G. Zhang Q. Sreeram V. Robust delay-range-dependent stabilization for Markovian jump systems with mode-dependent time delays and nonlinearities Optimal Control Applications & Methods 2010 31 3 249 264 10.1002/oca.901 MR2662715 ZBL1204.93104 Hu L.-S. Bai T. Shi P. Wu Z. Sampled-data control of networked linear control systems Automatica 2007 43 5 903 911 10.1016/j.automatica.2006.11.015 MR2306738 ZBL1117.93044 Cloosterman M. B. G. Hetel L. van de Wouw N. Heemels W. P. M. H. Daafouz J. Nijmeijer H. Controller synthesis for networked control systems Automatica 2010 46 10 1584 1594 10.1016/j.automatica.2010.06.017 MR2877314 ZBL1204.93044 Zhang W.-A. Yu L. Modelling and control of networked control systems with both network-induced delay and packet-dropout Automatica 2008 44 12 3206 3210 10.1016/j.automatica.2008.09.001 MR2531428 ZBL1153.93321 Wang G. Zhang Q. Sreeram V. H control for discrete-time singularly perturbed systems with two Markov processes Journal of the Franklin Institute 2010 347 5 836 847 10.1016/j.jfranklin.2010.03.007 MR2645394 ZBL1286.93194 Li M. Zhou W. Lu H. Fang J. Robust H control for a generic linear rational expectations model of economy Applied Mathematics and Computation 2010 216 7 2145 2154 10.1016/j.amc.2010.03.049 MR2647082 ZBL1193.93092 Lien C.-H. Yu K.-W. Lin Y.-F. Chung Y.-J. Chung L.-Y. Robust reliable H control for uncertain nonlinear systems via LMI approach Applied Mathematics and Computation 2008 198 1 453 462 10.1016/j.amc.2007.08.085 MR2403965 ZBL1141.93322 Peng C. Yue D. Tian Y.-C. Delay distribution based robust H control of networked control systems with uncertainties Asian Journal of Control 2010 12 1 46 57 MR2676770 Jia T. Niu Y. Wang X. H control for networked systems with data packet dropout International Journal of Control, Automation and Systems 2010 8 2 198 203 2-s2.0-77951154864 10.1007/s12555-010-0203-y Li J. G. Yuan J. Q. Lu J. G. Observer-based H control for networked nonlinear systems with random packet losses ISA Transactions 2010 49 1 39 46 2-s2.0-72249113841 10.1016/j.isatra.2009.08.001 Li X. Sun S. Robust H control for networked systems with random packet dropouts and time delays Procedia Engineering 2012 29 4192 4197 2-s2.0-84863345198 10.1016/j.proeng.2012.01.642 Wen S. Zeng Z. Huang T. Bao G. Observer-based H control of a class of mixed delay systems with random data losses and stochastic nonlinearities ISA Transactions 2013 52 2 207 214 2-s2.0-84875270638 10.1016/j.isatra.2012.10.001 Wang J. Yang H. H control of a class of networked control systems with time delay and packet dropout Applied Mathematics and Computation 2011 217 18 7469 7477 2-s2.0-79953224435 10.1016/j.amc.2011.02.047 Zhai G. Hu B. Yasuda K. Michel A. N. Qualitative analysis of discrete-time switched systems Proceedings of the American Control Conference May 2002 Anchorage, Alaska, USA 1880 1885 2-s2.0-0036349370 10.1109/ACC.2002.1023907 El Ghaoui L. Oustry F. AitRami M. A cone complementarity linearization algorithm for static output-feedback and related problems IEEE Transactions on Automatic Control 1997 42 8 1171 1176 10.1109/9.618250 MR1469081 ZBL0887.93017 Zhang L. Huang B. Lam J. H model reduction of Markovian jump linear systems Systems & Control Letters 2003 50 2 103 118 10.1016/S0167-6911(03)00133-6 MR2012300 ZBL1157.93519