Delay-dependent L2-L-infinity model reduction for polytopic systems with time-varying delay

Similar documents
Research Article Convex Polyhedron Method to Stability of Continuous Systems with Two Additive Time-Varying Delay Components

LOW ORDER H CONTROLLER DESIGN: AN LMI APPROACH

Research Article Delay-Dependent H Filtering for Singular Time-Delay Systems

Delay-dependent stability and stabilization of neutral time-delay systems

ROBUST STABILITY TEST FOR UNCERTAIN DISCRETE-TIME SYSTEMS: A DESCRIPTOR SYSTEM APPROACH

On Design of Reduced-Order H Filters for Discrete-Time Systems from Incomplete Measurements

Fixed-Order Robust H Filter Design for Markovian Jump Systems With Uncertain Switching Probabilities

Positive observers for positive interval linear discrete-time delay systems. Title. Li, P; Lam, J; Shu, Z

Research Article An Equivalent LMI Representation of Bounded Real Lemma for Continuous-Time Systems

An LMI Approach to Robust Controller Designs of Takagi-Sugeno fuzzy Systems with Parametric Uncertainties

Delay-Dependent Stability Criteria for Linear Systems with Multiple Time Delays

Static Output Feedback Stabilisation with H Performance for a Class of Plants

Research Article Stabilization Analysis and Synthesis of Discrete-Time Descriptor Markov Jump Systems with Partially Unknown Transition Probabilities

State Estimation with Finite Signal-to-Noise Models

Circuits, Systems, And Signal Processing, 2012, v. 31 n. 1, p The original publication is available at

A new robust delay-dependent stability criterion for a class of uncertain systems with delay

Delay-dependent Stability Analysis for Markovian Jump Systems with Interval Time-varying-delays

Delay-Dependent Exponential Stability of Linear Systems with Fast Time-Varying Delay

Gramians based model reduction for hybrid switched systems

Convex Optimization Approach to Dynamic Output Feedback Control for Delay Differential Systems of Neutral Type 1,2

On Bounded Real Matrix Inequality Dilation

On Delay-Dependent Robust H Control of Uncertain Continuous- and Discrete-Time Linear Systems with Lumped Delays

LMI based output-feedback controllers: γ-optimal versus linear quadratic.

Deakin Research Online

Stability of linear time-varying systems through quadratically parameter-dependent Lyapunov functions

LINEAR QUADRATIC OPTIMAL CONTROL BASED ON DYNAMIC COMPENSATION. Received October 2010; revised March 2011

7.1 Linear Systems Stability Consider the Continuous-Time (CT) Linear Time-Invariant (LTI) system

Delay-dependent robust stabilisation of discrete-time systems with time-varying delay

A Singular Value Decomposition Based Closed Loop Stability Preserving Controller Reduction Method

ON THE ROBUST STABILITY OF NEUTRAL SYSTEMS WITH TIME-VARYING DELAYS

June Engineering Department, Stanford University. System Analysis and Synthesis. Linear Matrix Inequalities. Stephen Boyd (E.

Marcus Pantoja da Silva 1 and Celso Pascoli Bottura 2. Abstract: Nonlinear systems with time-varying uncertainties

Research Article On Exponential Stability Conditions of Descriptor Systems with Time-Varying Delay

STABILITY ANALYSIS FOR SYSTEMS WITH LARGE DELAY PERIOD: A SWITCHING METHOD. Received March 2011; revised July 2011

Research Article Delay-Range-Dependent Stability Criteria for Takagi-Sugeno Fuzzy Systems with Fast Time-Varying Delays

Results on stability of linear systems with time varying delay

New Lyapunov Krasovskii functionals for stability of linear retarded and neutral type systems

Appendix A Solving Linear Matrix Inequality (LMI) Problems

Correspondence should be addressed to Chien-Yu Lu,

IMPULSIVE CONTROL OF DISCRETE-TIME NETWORKED SYSTEMS WITH COMMUNICATION DELAYS. Shumei Mu, Tianguang Chu, and Long Wang

Robust Variance Constrained Filter Design for Systems with Non-Gaussian Noises

Controller synthesis for positive systems under l 1-induced performance

Eects of small delays on stability of singularly perturbed systems

IEEE Transactions on Automatic Control, 2013, v. 58 n. 6, p

STABILIZATION FOR A CLASS OF UNCERTAIN MULTI-TIME DELAYS SYSTEM USING SLIDING MODE CONTROLLER. Received April 2010; revised August 2010

THE phenomena of time delays are often encountered in

Delay and Its Time-derivative Dependent Robust Stability of Uncertain Neutral Systems with Saturating Actuators

On Dwell Time Minimization for Switched Delay Systems: Free-Weighting Matrices Method

Delay-Dependent H 1 Control of Uncertain Discrete Delay Systems

ON POLE PLACEMENT IN LMI REGION FOR DESCRIPTOR LINEAR SYSTEMS. Received January 2011; revised May 2011

2nd Symposium on System, Structure and Control, Oaxaca, 2004

A DELAY-DEPENDENT APPROACH TO DESIGN STATE ESTIMATOR FOR DISCRETE STOCHASTIC RECURRENT NEURAL NETWORK WITH INTERVAL TIME-VARYING DELAYS

Multiple-mode switched observer-based unknown input estimation for a class of switched systems

Stability Analysis of Linear Systems with Time-varying State and Measurement Delays

H fault detection for a class of T-S fuzzy model-based nonlinear networked control systems

Improved delay-dependent globally asymptotic stability of delayed uncertain recurrent neural networks with Markovian jumping parameters

Filter Design for Linear Time Delay Systems

Linear Matrix Inequality (LMI)

Stability Analysis of Continuous-Time Switched Systems With a Random Switching Signal. Title. Xiong, J; Lam, J; Shu, Z; Mao, X

Reduced-order filter for stochastic bilinear systems with multiplicative noise

On Computing the Worst-case Performance of Lur'e Systems with Uncertain Time-invariant Delays

Discrete-Time H Gaussian Filter

COMPUTATION OF ROBUST H CONTROLLERS FOR TIME-DELAY SYSTEMS USING GENETIC ALGORITHMS

SYNTHESIS OF ROBUST DISCRETE-TIME SYSTEMS BASED ON COMPARISON WITH STOCHASTIC MODEL 1. P. V. Pakshin, S. G. Soloviev

Technical Notes and Correspondence

José C. Geromel. Australian National University Canberra, December 7-8, 2017

An LMI Optimization Approach for Structured Linear Controllers

L 2 -induced Gains of Switched Systems and Classes of Switching Signals

Xu, S; Lam, J; Zou, Y; Lin, Z; Paszke, W. IEEE Transactions on Signal Processing, 2005, v. 53 n. 5, p

Robust Input-Output Energy Decoupling for Uncertain Singular Systems

Switching H 2/H Control of Singular Perturbation Systems

To appear in IEEE Trans. on Automatic Control Revised 12/31/97. Output Feedback

Optimization based robust control

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and

H Filter/Controller Design for Discrete-time Takagi-Sugeno Fuzzy Systems with Time Delays

IN many practical systems, there is such a kind of systems

An LQ R weight selection approach to the discrete generalized H 2 control problem

Robust absolute stability criteria for a class of uncertain Lur e systems of neutral type

Robust Anti-Windup Controller Synthesis: A Mixed H 2 /H Setting

State feedback gain scheduling for linear systems with time-varying parameters

Model reduction for linear systems by balancing

ROBUST CONTROLLER DESIGN: POLYNOMIALLY PARAMETER DEPENDENT LYAPUNOV FUNCTION APPROACH

Stability Analysis and H Synthesis for Linear Systems With Time-Varying Delays

Input/output delay approach to robust sampled-data H control

A New Strategy to the Multi-Objective Control of Linear Systems

Design of State Observer for a Class of Non linear Singular Systems Described by Takagi-Sugeno Model

OBSERVER DESIGN WITH GUARANTEED BOUND FOR LPV SYSTEMS. Jamal Daafouz Gilles Millerioux Lionel Rosier

Robust Output Feedback Controller Design via Genetic Algorithms and LMIs: The Mixed H 2 /H Problem

Ieee Transactions On Automatic Control, 1998, v. 43 n. 8, p

PI OBSERVER DESIGN FOR DISCRETE-TIME DECOUPLED MULTIPLE MODELS. Rodolfo Orjuela, Benoît Marx, José Ragot and Didier Maquin

Static Output Feedback Controller for Nonlinear Interconnected Systems: Fuzzy Logic Approach

ROBUST STATE FEEDBACK CONTROL OF UNCERTAIN POLYNOMIAL DISCRETE-TIME SYSTEMS: AN INTEGRAL ACTION APPROACH

On linear quadratic optimal control of linear time-varying singular systems

ROBUST PASSIVE OBSERVER-BASED CONTROL FOR A CLASS OF SINGULAR SYSTEMS

ROBUST QUANTIZED H CONTROL FOR NETWORK CONTROL SYSTEMS WITH MARKOVIAN JUMPS AND TIME DELAYS. Received December 2012; revised April 2013

Research Article Delay-Dependent Exponential Stability for Discrete-Time BAM Neural Networks with Time-Varying Delays

Stability of neutral delay-diœerential systems with nonlinear perturbations

STABILITY ANALYSIS FOR DISCRETE T-S FUZZY SYSTEMS

H State Feedback Control of Discrete-time Markov Jump Linear Systems through Linear Matrix Inequalities

H Synchronization of Chaotic Systems via Delayed Feedback Control

Transcription:

itle Delay-dependent L-L-infinity model reduction for polytopic systems with time-varying delay Authors) Wang, Q; Lam, J Citation he 8 IEEE International Conference on Automation and Logistics ICAL 8), Qingdao, China, 1-3 September 8. In Conference Proceedings, 8, p. 5-55 Issued Date 8 URL http://hdl.handle.net/17/6 Rights IEEE International Conference on Automation and Logistics. Copyright IEEE.; 8 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.; his work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4. International License.

Proceedings of the IEEE International Conference on Automation and Logistics Qingdao, China September 8 Delay-dependent L -L Model Reduction for Polytopic Systems with ime-varying Delay Qing Wang School of information Science and echnology Sun Yat-sen University Guangzhou, 5175, China wangq5@mail.sysu.edu.cn James Lam Department of Mechanical Engineering he University of Hong Kong Pokfulam Road, Hong Kong james.lam@hku.hku.hk Abstract his paper considers the L -L model reduction problems for polytopic system with time-varying delay. In terms of the solution of linear matrix inequalities LMIs) and inverse constraints, sufficient conditions are presented to construct the reduced order models such that the L -L gain of the error system between the full order model and the reduced order one is less than a given scalar. Index erms model reduction problems, polytopic systems with time-varying delay, L -L norm I. INRODUCION he L -L gain of a system Σ is defined as Σ = sup y L u L 1 with zero initial state, where yt) is the output of system Σ and ut) is the input of system Σ with u L = 1/ u t) dt) and u L = sup t u t), u t) = u t) u t) 1 and 16. he L -L gain of error system, which is also referred as the energy-to-peak gain in 14 is a natural criterion for model reduction. he L -L model reduction problem for continuous-time and discretetime systems without delay is studied in 11. Necessary and sufficient conditions are given in terms of LMIs and a rank constraint, which can be solved by the alternating projection algorithm 4, Chap.13. he model reduction problem has been extensively studied in the literature, as can be seen from 13 and the references listed therein. In many physical, industrial and engineering systems, control systems cannot be described accurately without the introduction of delay element. Delays can lead to poor performance and instability of control systems. Considerable research has been carried out on systems with time-varying delay in recent years. he criteria for asymptotic stability of such systems can be classified as delay-independent 18 or delay-dependent 3, 6, 9, 1, which is less conservative than delay-independent stability criteria in general. For his work was supported by CRCG 611159157, Sun Yat-sen University Young eachers Foundation under grant 3171911 and Young eachers Foundation of School of information Science and echnology under grant 3577. polytopic systems with constant delay, the robust stability and stabilization are studied in 17. he extension to timevarying delay is investigated in 1 with stability criteria less conservative through the introduction of slack variables in the LMIs. he polytopic system has been extensively studied in the literature, as can be seen from 8, 19 and the references listed therein. o the authors knowledge, the L -L model reduction problems for polytopic systems with time-varying delay are a challenging topic worthwhile to tackle. II. L -L MODEL REDUCION A polytopic system Σ with time-varying delay is given by Σ:ẋ t) = At)x t)+a h t)xt ht)) + Bt)u t) y t) = Ct)x t)+c h t)xt ht)) x t) =, t h, where x t) R n is the state, u t) R m is the control input which belongs to L, ) and y t) R p is the controlled output. At), A h t), Bt), Ct) and C h t) are appropriately dimensioned continuous functions of time t, and satisfy the real convex polytopic model A t) A h t) B t)) = α i t)a i A hi B i ), C t) C h t)) = α i t)c i C hi ) q α i t), α it) =1, where α i t), i =1,,...,q, are time-varying functions of t, C i,c hi,b i,a i and A hi, i =1,,...,q,are constant matrices with appropriate dimensions. ht) is the time-varying delay satisfying <ht) h <, ḣt) μ<1 with h > and μ>. System Σ, which is to be reduced, is assumed to be quadratically stable. Definition.1: If there exists an ˆnth-order quadratically stable system ˆΣ ˆΣ :.ˆx t) = Ât)ˆx t)+âht)ˆxt ht)) + ˆBt)u t) ŷ t) = Ĉt)ˆx t)+ĉht)xt ht)) ˆx t) =, t h, 978-1-444-53-7/8/$. 8 IEEE 5

where ˆx t) Rˆn, ŷ t) R p, and ˆn <n,such that Σ ˆΣ = sup y ŷ L <γ u L 1 then we say the L -L model reduction problem for system Σ is solvable. he L -L gain of retarded system Σ R Σ R : ẋ t) = Ax t)+a h xt ht)) + But) y t) = Cxt)+C h xt ht)) can be characterized by the following algebraic conditions, where A, A h,b, C and C h are contant matrices. Lemma.: 15 If there exist matrices P >, Q>, X>, Z>and Y, and a scalar <α<1 such that A P + PA+ hx + Y +Y + Q PA h Y PB τa Z A h P Y 1 μ)q τa h Z B P I m τb Z τza τza h τzb τz αp C 1 α)p Ch C C h γ I p X Y Y Z < < then system Σ R is asymptotically stable and Σ R < γ, where τ = h 1 u. In the case of ht) =h, if there exists matrices P>, Q>, and a scalar <α<1 such that A P + PA+ hx + Y + PA h Y PB Y + Q A h P Y 1 μ)q < B P I m αp C 1 α)p Ch < C C h γ I p he following theorem gives the solution of L -L model reduction problem for system Σ. heorem.3: With ñ = n+ˆn, if there exist matrices P>, P >, X>, X >, Q>, Q >, M>, M >, N>, Ñ>, Z>, Z > and Y, and a scalar <α<1 such that for i =1,,...,q, X11 X 1 X1 < 1) X Y11 Y 1 Y1 < ) Y αi 1P I1 Ci 1 α)i 1 P I1 Chi < 3) γ I p X Y Z N Y M 4) 5) P P = Iñ, Q Q = Iñ, X X = I 6) Z Z = Iñ, MM = Iñ, ÑN = Iñ 7) where denotes the symmetric terms in a symmetric matrix and Λ i11 Λ i1 B i Λ i14 Λ i15 X 11 = Λ Λ i1 I m Bi Λ i44 Λ i45 Λ i55 I 1 P I1 P I1 P I1 P X 1 = Iñ X = I P I P I P I P h 1 X Q M Ñ Y 11 = A i I 1P I1 + I 1 P I ) 1 A i hx + Y +I 1 +Y I + Q 1 Y 1 = I 1 P I1 A hi I 1 Y I1 τa i I 1Z 1 μ)i1 QI1 Y = τa hi I 1Z τz and Λ i11 = I 1 P I 1 A i + A i I 1 P I 1, Λ i1 = A hi I 1 8) Λ i14 = I 1 P I 1 A i, Λ i15 = A i I 1 P I 9) Λ = 1 μ)q, Λ i44 = τ 1 I 1 ZI 1 1) Λ i55 = τ 1 I ZI 11) Λ i45 = τ 1 I 1 ZI A i I 1 P I 1) then there exists a quadratically stable system ˆΣ such that the L -L model reduction problem is sovable with Σ ˆΣ <γ. In this case, a desired reduced system corresponding to a feasible solution P, P, X, X, Q, Q, M, M, N, Ñ, Z, Z, Y, α) to 1) 7) is given by ˆB t) Â t) Âh t) ) = Ĉ t) Ĉh t) ) = α i t)ḡi1 13) α i t)ḡi 14) 51

where Ḡ i1 = ˆBi  i  hi = U 1 Ω i1 1 V i1 Λ i1 Λ1 V i1 Λ 1 +U 1 i1 W 1 ) 1 i1 L i1 Λ1 V i1 Λ 1 15) Ḡ i = Ĉ i Ĉ hi = U 1 Ω i V i Λ Λ V i Λ +U 1 i W 1 ) 1 i L i Λ V i Λ 16) and for i = 1,,...,q, L i1 and L i are any matrices satisfying L i1 < 1 and L i < 1, and U i1 > and U i > such that V i1 = Ω1 U 1 Ω i1 1 Φ i1 > V i = Ω U 1 Ω i Φ > W i1 = U i1 Ω 1 W i = U i Ω Φ i1 = Φ i = Ω 1 = Λ 1 = Λ = ) V i1 V i1 Λ 1 Ω 1 Λ1 V i1 Λ 1 Λ1 V i1 ) V V i Λ Ω Λ V i Λ Λ V i Ā i P + P hx Āi+ P + Y + Āhi Y P B i τā i Z Y + Q 1 μ)q τā hi Z I m τ B i Z τz 17) αp C i 1 α)p C hi 18) γ I p P Iˆn, Ω = 19) I p τz Iˆn I m Iˆn ) Iˆn Iˆn 1) Iˆn Ā i = B i = Ai Bi Ahi, Ā hi = ), Ci = C i 3) C hi = C hi 4) I 1 = I n, I = Iˆn 5) Proof. We first consider the error system Σ ˆΣ given by x t) = à t) x t)+ãh t) xt ht)) + B t) u t) ỹ t) = C t) x t)+ C h t) xt ht)) where x t) = x t) ˆx t), ỹ t) =y t) ŷ t), = B t) à t) Ãh t) = Bi + F Ḡi1 N α i t) Ā i + F Ḡi1 H Āhi + F Ḡi1 M 6) C t) Ch t) α i t) Ci + Ḡi J Chi + Ḡi K 7) with Ḡi1 and Ḡi, i =1,,...,q, defined in 15) and 16), B i, Ā i, Ā hi, Ci and C hi given in ) and 3), and F = M = J =, Iˆn H =, Iˆn Iˆn Iˆn N = I m 8) 9), K = Iˆn 3) From Lemma., if there exist matrices Ḡ i1 and Ḡi satisfying 4) and Āi + F Ḡi1 H ) P +P ) Ā i + F Ḡi1 H P ) Ā hi + F Ḡi1 M + hx + Y + Y Y + Q 1 μ)q P ) Bi + F Ḡi1 N τ Ā i + F Ḡi1 H ) Z τ Ā hi + F Ḡi1 M ) Z I m τ Bi + F Ḡi1 N ) Z τz < 31) αp Ci + Ḡi J ) 1 α)p Chi + Ḡi K ) γ I p < 3) 5

It follows from 31) and 3) that < Ã t) P + P Ã t) + hx + Y +Y + Q ) F Ḡi1 N τã t) Z τãh t) Z I m τ B t) Z τz P Bi + P Ãh t) Y 1 μ)q αp C t) 1 α)p Ch t) < γ I p hen we have Σ ˆΣ <γ.it is easy to show that matrix inequalities 31) and 3) can be rewritten as Φ i1 + Ω 1 Ḡ i1 Λ1 + Ω1 ) Ḡ i1 Λ1 < 33) Φ i + Ω Ḡ i Λ + Ω ) Ḡ i Λ < 34) Ω 1 Φ i1 Ω 1 <, equals to where Λ i11 Λ i1 B i 1 μ)q I m Λ i14 Λ i15 I 1 P I1 P Λ i1 Bi Λ i44 Λ i45 Λ i55 I P I P h 1 X Q +M 1 Y N 1 +M 1 Y N 1 ) < 37) M 1 = P I 1 Iñ P I N 1 = P I 1 P I where Φ ij, Ω j, and Λ j,,,...,q, j =1,, are defined in 17) 1). LMIs 33) and 34) have solutions Ḡi1 and Ḡi, if and only if Ω j Ω Φ 1 Ω i1 1 <, Λ 1 Φi1 Λ 1 < 35) Ω Φ Ω i <, Λ Φi Λ < 36) and Λ j,j=1,, can be chosen as follows: Ω 1 = Λ 1 = Ω = I 1 P 1 Iñ I m τ 1 I 1 Z 1 I P 1 τ 1 I Z 1 I 1 I 1 Iñ Iñ Iñ Λ = I 1 I 1 I p where I 1 and I are defined in 5). hen, by Schur complement and the conditions in 6) and 7), we obtain that and Λ i11, Λ i1, Λ i14, Λ i15, Λ i44 and Λ i45,,,...,q, are defined in 8) 1). Since for any matrices M, N,Y and M>and N>, MY N +MY N ) MMM + N NN, holds if YM 1 Y N. LMI 37) is true if 1) and 5) hold. Λ 1 Φi1 Λ 1 <, is equivalent to ). For the inequalities in 36), Ω Φ αp Ω i = 1 α)p < and Λ Φ i Λ < is equivalent to 3). From LMIs 1) 5), if there exist matrices P, P, X, X, Q, Q, M, M, N, Ñ, Z, Z and Y satisfying 35) and 36), there exist matrices Ḡi1 and Ḡi such that LMIs 33) and 34) hold. herefore, we have Σ ˆΣ <γ.all the parameters of the reduced order models satisfying 33) and 34) can be constructed by the parametrization method of Gahinet and Apkarian 7. his completes the proof. In the case of ht) =h, a delay-independent sufficient condition is given for the L -L model reduction problem from Lemma. and the proof of heorem.3. Corollary.4: If there exist matrices P>, P >, Q> and Q >, and a scalar <α<1such that for i = 53

1,,...,q, I 1 P I 1 A i + A i I 1 P I 1 A hi I 1 B i I 1 P Q I m < Q A i I 1P I1 + I 1 P I1 A i + I 1 QI1 I 1 P I1 A hi < I 1 QI1 αi 1 P I1 Ci 1 α)i 1 P I1 Chi < γ I p P P = Iñ, Q Q = Iñ there exists a quadratically stable system ˆΣ with ht) = h solving the L -L model reduction problem with Σ ˆΣ <γ. Remark.5: If P, P = P 1 and Q have the following special form ˆX X P = 1 X1, P 1 Ŷ 1 Y = 1 X Y1 Y Ẑ X Q = 1 X1 38) X Corollary.4 can be expressed by LMIs with rank constraint: if there exist matrices ˆX >, Ŷ > and Ẑ>, and a scalar <α<1 such that for i =1,,...,q, A i Ŷ + ŶA i+ Ẑ + Ŷ ˆX ŶA hi ŶB i Ẑ + Ŷ ˆX < I m A i ˆX + ˆXAi + Ẑ ˆXA hi < Ẑ α ˆX Ci 1 α) ˆX Chi < γ I p rankŷ ˆX) ˆn there exists a quadratically stable system ˆΣ with ht) = h solving the L -L model reduction problem with Σ ˆΣ <γ. If ht) =,C h =and A h =,q=1, system Σ reduces to a continuous time-invariant system without delay Σ C : ẋ t) = A 1 x t)+b 1 u t) y t) = C 1 x t) Corollary.6: If there exist matrices P > and P > such that I 1 P I 1 A 1 + A 1 I 1 P I 1 + B 1 B1 < 39) A 1 I 1 P I1 + I 1 P I1 A 1 < 4) I 1 P I1 + γ C1 C 1 < 41) P P = I n+ˆn 4) there exists an asymptotically stable continuous system ˆΣ C ˆΣ C :. ˆx t) = Â1ˆx t)+ ˆB 1 u t) ŷ t) = Ĉ1ˆx t) that solves the L -L model reduction problem with Σ C ˆΣ C < γ. If P > has special form 38), conditions 39) 4) are equivalent to the existence of matrices ˆX > and Ŷ > such that A 1 Ŷ + ŶA 1 + B 1 B1 <,A ˆX 1 + ˆXA 1 <, ˆX + γ C1 C 1 < and rankŷ ˆX) ˆn, which recovers the result in 11. In this particular case, the model reduction condition is both necessary and sufficient. III. MODEL REDUCION ALGORIHM In order to use the cone complementarity linearization CCL) algorithm 5 to solve the L -L model reduction problem with fixed < α < 1, we define a convex set C := {X X satisfies LMIs 1) 5)} and a nonconvex set = {X X satisfies the inverse constraints in 6-7)} where X = P>, P >, X>, X >, Q>, Q >, Z >, Z >, M >, M >, N>, Ñ>, Y) he solvability of the L -L model reduction problem can be translated into the following nonconvex feasibility problem: Find X C subject to X 43) through the sufficient condition in heorem.3. hen, by the CCL approach, the above nonconvex problem 43) has a solution if and only if the minimization problem min X C Cin { tracep P + X X + Q Q+ Z Z + M M + NÑ) where P I X I,, I P I X Q I Z I C in =,, I Q I Z M I N I, I M I Ñ } 44) 54

achieves a minimum, 1ñ, that is, an optimal solution of problem 44) satisfying tracep P ) = tracex X) =traceq Q) =tracez Z) = tracem M) =tracenñ) =ñ Otherwise 43) is infeasible. herefore, in order to solve the L -L model reduction problem for systems with polytopic uncertainties and time-varying delay, we transform it to a global solution of the minimization problem 44). he CCL algorithm solves problem like 44) effectively although it is still a nonconvex optimization problem. Interested readers may refer to 5 for the details of the CCL algorithm. An algorithm is presented to solve the L -L model reduction problem based on the above analysis. L -L CCL Model Reduction Algorithm: Step 1 Given system Σ, the order of the reduced order model ˆn, prescribed model reduction error γ>and δ > is a sufficiently small prescribed a scalar to control the convergence accuracy. Step Set k =and for a fixed α, choose any initial guess X = P, P,X, X,Q, Q,Z, Z, M, M,N, Ñ, Y ) C Step 3 Define f k P, P,X, X,Q, Q, Z, Z,M, M,N,Ñ) = tracep k P + Pk P + X k X + X k X + Q k Q + Qk Q + Z k Z + Zk Z +M k M + Mk M + N k Ñ + ÑkN) Solve the following convex minimization problem: { } fk P, P,X, X,Q, Q, min X C Cin Z, Z,M, M,N,Ñ) and denote the minimizer Xk and compute the minimum value fk = f kxk ). Step 4 If fk 1ñ <δ, then construct a reduced order model based on 13) and 14); otherwise set k = k +1, and assign Xk 1 = P k, Pk,X k, Xk,Q k, Q k,z k, Z k,m k, M k,n k, Ñk,Y k ) and go to step 3. It can be seen that Step is a simple LMI feasibility problem, and Step 3 is a convex programming with LMI constraints. From the explanation in, 5, {fk } is decreasing and bounded below by 1ñ. Once it converges, then 43) is feasible, which implies that the L -L model reduction problem is solvable for given γ>. IV. CONCLUSION his paper presents a delay-dependent sufficient condition for the L -L model reduction problems of polytopic systems with time-varying delay. An explicit formula for the construction of reduced order models has been given in terms of a set of the solution of LMIs and inverse constraints. An effective algorithm has been exploited to solve the LMI problems with inverse constraints. REFERENCES 1 M. Corless, G. Zhu, and R. E. Skelton. Improved robustness bounds using covariance matrices. In Proceedings of the 8th Conference on Decision and Control, pages 667 67, ampa, Florids, 1989. M. C. de Oliveira and J. C. Geromel. Numerical comparison of output feedback design methods. In Proceedings of the American Control Conference, pages 7 76, Albuquerque, New Mexico, 1997. 3 C. E. de Souza and X. Li. Delay-dependent robust H control of uncertain linear state-delayed systems. Automatica, 357):1313 131, 1999. 4 L. El Ghaoui and S.-I. Niculescu. Advances in Linear Matrix Inequality Methods in Control. Society for Industrial and Applied Mathematics, Philadelphia,. 5 L. El Ghaoui, F. Oustry, and M. AitRami. A cone complementarity linearization algorithm for static output-feedback and related problems. IEEE ransactions on Automatic Control, 48):1171 1176, 1997. 6 E. Fridman and U. Shaked. A descriptor system approach to H control of linear time-delay systems. IEEE ransactions on Automatic Control, 47):53 7,. 7 P. Gahinet and P. Apkarian. A linear matrix inequality approach to H control. International Journal of Robust and Nonlinear Control, 4:41 448, 1994. 8 H. Gao, J. Lam, and C. Wang. Mixed H /H filtering for continuoustime polytopic systems: a parameter-dependent approach. Circuits Systems Signal Process, 46):689 7, 5. 9 H. Gao and C. Wang. Delay-dependent robust H and L L filtering for a class of uncertain nonlinear time-delay systems. IEEE ransactions on Automatic Control, 489):1661 1666, 3. 1 H. Gao, C. Wang, and L. Zhao. Comments on: An LMI-based approach for robust stabilization of uncertain stochastic systems with time-varying delays IEEE ransactions on Automatic Control, 4811):73 74, 3. 11 K. M. Grigoriadis. L and L -L model reduction via linear matrix inequalities. International Journal of Control, 683):485 498, 1997. 1 Y. He, M. Wu, J.-H. She, and G.-P. Liu. Parameter-dependent lyapunov functional for stability of time-delay systems with polytopic-type uncertainties. IEEE ransactions on Automatic Control, 495):88 83, 4. 13 G. Obinata and B. D. O. Anderson. Model Reduction for Control System Design. Springer-Verlag, London, 1. 14 R. E. Skelton,. Iwasaki, and K. M. Grigoriadis. A Unified Algebraic Approach to Linear Control Design. aylor & Francis Ltd., London, 1998. 15 Q. Wang, J. Lam, S. Xu, and H. Gao. Delay-dependent and delay-independent energy-to-peak model approximation for systems with time-varying delay. International Journal of Systems Science, 368):445 46, 5. 16 D. A. Wilson. Convolution and Hankel operator norms for linear systems. IEEE ransactions on Automatic Control, 341):94 97, 1989. 17 Y. Xia and Y. Jia. Robust control of state delayed systems with polytopic type uncertainties via parameter-dependent Lyapunov functionals. Systems & Control Letters, 53):183 193, 3. 18 S. Xu, J. Lam, S. Huang, and C. Yang. H model reduction for linear time-delay systems: continuous-time case. International Journal of Control, 7411):16 174, 1. 19 S. Zhou, J. Lam, and S. Xu. Robust H control for discrete-time polytopic uncertain systems with linear fractional vertices. Journal of Control heory and Applications, 1):75 81, 4. 55