New LMI-Based Conditions for Quadratic Stabilization of LPV Systems

(1)

Volume 2008, Article ID 563845,12pages doi:10.1155/2008/563845

Research Article

New LMI-Based Conditions for Quadratic Stabilization of LPV Systems

Wei Xie

College of Automation Science and Technology, South China University of Technology, Guangzhou 510641, China

Correspondence should be addressed to Wei Xie,[email protected]

Received 16 January 2008; Revised 5 August 2008; Accepted 24 November 2008 Recommended by Alexander Domoshnitsky

This paper is concerned with quadratic stabilization problem of linear parameter varyingLPV systems, where arbitrary time-varying dependent parameters are belonging to a polytope.

It provides improved linear matrix inequality- LMI- based conditions to compute a gain- scheduling state-feedback gain that makes closed-loop system quadratically stable. The proposed conditions, based on the philosophy of P ´olya’s theorem, are written as a sequence of progressively less and less conservative LMI. More importantly, by adding an additional decision variable, at each step, these new conditions provide less conservative or at least the same results than previous methods in the literature.

Copyrightq2008 Wei Xie. 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.

1. Introduction

Linear parameter varying LPV systems are formalized as a certain type of nonlinear systems, and a control strategy has been developed for these systems based on classical gain- scheduled adaptive methodology1,2. There are many examples of dependent physical parameters including inertia, stiffness, or viscosity coefficients in mechanical systems, aero dynamical coefficients in flight control, resistor and capacitor values in electrical circuits and so forth. Therefore, it is often desirable to obtain guarantees of stability and performance against dependent parameter when analyzing these control systems. In the past decade, main papers and special publications concerning LPV controller design problem have appeared in 3–10.

Quadratic stability has been widely used to assess closed-loop stability and performance. This approach allows us to describe several problems of stability analysis and synthesis as LMI optimization problems, which can be solved in polynomial time by interior point algorithms3,11–15. It is known that the results based on quadratic stability are frequently conservative in the context of analysis and synthesis for LPV systems when

(2)

compared to the results from conditions based on parameter-dependent Lyapunov functions see, e.g.,3,6–9,16–18. But quadratic stability remains attractive due to its low numerical complexity, being largely employed as a first step in the investigation of stability performance and control design of LPV systems.

As to quadratic stabilization studies of LPV systems, even though stability analysis problem is NP-hard in general, a number of more or less conservative analysis methods are presented to assess quadratic stability 3–8, where a fixed quadratic Lyapunov function is found to prove stability of LPV systems. More recently, a kind of necessary and sufficient LMI-based condition has been proposed to compute a quadratically stabilizing state feedback controller for continuous-time linear systems with arbitrary time-varying parameters belonging to a polytope 19. These conditions are based on an extension of P ólya’s theorem20and are written as a sequence of progressively less and less conservative LMI. However, at each step, the LMI-based conditions are still sufficient, and have some conservatism. It leads to higher computational times.

The main contribution of this paper is to provide new necessary and suﬃcient LMI- based conditions to compute a quadratically stabilizing gain-scheduling state feedback for LPV systems. The proposed conditions are based on the systematic construction of homogeneous polynomial solution for parameter-dependent LMI too. At each step, a set of LMIs provides suﬃcient conditions for the existence of such a gain-scheduling state feedback.

Necessity is asymptotically attained through a relaxation based on the philosophy of P ´olya’s theorem. More importantly, by adding an additional decision variable, at each step, these new conditions can provide less conservative or at least the same results than the most recent existing methods in the literature. Consequently, the feasible solutions can be obtained in much lower steps.

2. Preliminary

Consider an LPV systemP∂tdescribed by state space equations as

xt ˙ A

∂t

xt B

∂t

ut. 2.1

Here, state-space matrices have compatible dimensions of time-varying parameters∂t ∂1t ∂2t · · · ∂ntT ∈Rⁿ.

Moreover, we have the following assumptions.

1The state-space matricesA∂tandB∂tare continuous and bounded functions and depend aﬃnely on∂t.

2The real parameters ∂t, that can be known in advance or online measurement values, exist in LPV plant and vary in a polytopeΘas

∂t∈Θ:Co

ω1, ω2, . . . , ωN

_r

i1

αitωi:αit≥0,

r i1

αit 1, r 2ⁿ

. 2.2

(3)

With above assumptions, the LPV plant is called polytopic, when it ranges in a matrix polytope, LPV systemP∂tcan be expressed as

A∂ ^r

i1

αitA ωi

, B∂ ^r

i1

αitBωi withαi≥0,

r i1

αi1. 2.3

The aim of this paper is to establish new LMI-based conditions of a gain-scheduling state feedback that quadratically stabilizes the class of system2.1. The control law is given with a state feedback as

ut −K∂xt, K∂ ^r

j1

αitK ωj

. 2.4

Substituting2.4into2.1, the closed-loop system can be written as

xt ˙ A_cl∂xt, ∂∈Θ, 2.5

whereAcl∂ A∂−B∂K∂.

According to quadratic stability theory3, the closed-loop system2.5is said to be quadratically stable if and only if there exists a symmetric positive definite matrixP ∈R^n×n such that

P A_cl∂ A^T_cl∂P <0, ∂∈Θ. 2.6

The concept of quadratic stability has been widely used for stability evaluation, control, and filter design for continuous and discrete, time-varying and time-invariant systems. The next lemma presents convex LMI conditions of infinite dimension that are necessary and suﬃcient to assure the existence of such a state-feedback gain.

Lemma 2.1. LPV system 2.1 is quadratically stabilizable if and only if there exist a symmetric positive definite matrixQ∈R^n×nand a parameter-dependent matrixN∂∈R^m×nsuch that

A∂Q QA^T∂−B∂N∂−N^T∂B^T∂<0_n, ∂t∈Θ. 2.7

In this case, the state-feedback gain is given by

K∂ N∂Q⁻¹. 2.8

(4)

Proof. Using the change of variablesN∂ K∂Q,2.7can be rewritten as A∂−B∂K∂

Q Q

A∂−B∂K∂_T

<0, 2.9

which, pre- and post-multiplied byQ⁻¹, yields2.6withP Q⁻¹. Conversely, pre- and post- multiplying2.6byP⁻¹and makingQP⁻¹give the equivalence condition

A∂−B∂K∂

Q Q

A∂−B∂K∂_T

<0, 2.10

which yields2.7by makingK∂ N∂Q⁻¹. To simplify notation, we define

G_ijQA^T_i A_iQ−N_j^TB^T_i −B_iN_j, i, j1,2, . . . , r, K_iN_iQ⁻¹, i1,2, . . . , r with A_iAωi, N_jNωj.

2.11

3. Main result

In this section, by adding an additional decision variable, a useful lemma is introduced below.

Lemma 3.1. LPV system 2.1 is quadratically stabilizable if and only if there exist a symmetric positive definite matrixW ∈R^n×nand parameter-dependent matricesN∂∈R^m×n,Y∂ Y^T∂∈ R^n×nsuch that one of the following equivalent conditions holds

i ϕ∂ A∂Q QA^T∂−B∂N∂−N^T∂B^T∂< Y∂≤0n, 3.1 ii ϕd∂

α1 α2 · · · αr

_d

A∂Q QA^T∂−B∂N∂−N^T∂B^T∂

<

α₁ α₂ · · · α_r_d

Y∂≤0_n, ∀d∈Z .

3.2

Proof. Conditioniis obtained directly through the use of a quadratic Lyapunov function associated to the closed-loop system 2.5. For any fixed ∂t ∈ Θand for all d ∈ Z , the equivalence betweeni andiiis immediate since∂t ∈ Θimplies_r

i1αi^d 1 for all d∈Z .

Remark 3.2. When the parameter-dependent matrixY∂is assumed to be zero, the condition iis reduced to the most recent existing conditions19. The existing conditions are written as a sequence of progressively less and less conservative LMI. With the increase of this positive integerd, necessity is asymptotically attained. However, at each step, the LMI-based conditions are still suﬃcient and have some conservatism. It leads to higher computational times. Here, an additional decision variableY∂is introduced to decrease the conservatism

(5)

at each step. It provides more design freedom to get a feasible solution. In the following, we only give the LMI-based conditions withd1, d2.

Theorem 3.3 d 1. LPV system 2.1 is quadratically stabilizable via the gain-scheduling controller2.4if there exist matricesQ > 0,Ni, i 1,2, . . . , r andYijl Y_lji^T, i, j, l 1,2, . . . , r, satisfying

Gii< Yiii, i1,2, . . . , r,

Gii Gij Gji< Yiij Yiji Y_iij^T, i1,2, . . . , r, j /i, j1,2, . . . , r, Gij Gil Gji Gjl Gli Glj < Yijl Yilj Yjil Y_ijl^T Y_ilj^T Y_jil^T,

i1,2, . . . , r−2, ji 1, . . . , r−1, jj 1, . . . , r,

⎡

⎢⎢

⎢⎣

Y_1i1 Y_1i2 · · · Y_1ir Y2i1 Y2i2 · · · Y2ir

... ... . .. ...

Yri1 Yri2 · · · Yrir

⎤

⎥⎥

⎥⎦≤0, i1,2, . . . , r.

3.3

Moreover, in this case, local state-feedback gains areKj NjQ⁻¹, j 1,2, . . . , r.

Remark 3.4. The details concerning these LMI-based results above can be referred to 21, Theorem 5 for Takagi-Sugeno fuzzy systems. New proposed LMI-based conditions are presented according to conditioniiofLemma 3.1in the case ofd2. Meanwhile, a simple proof is also given.

Theorem 3.5 d 2. LPV system 2.1 is quadratically stabilizable via the gain-scheduling controller2.4, if there exist matricesQ > 0;Ni, i1,2, . . . , r;YijmnY_njmi^T , i1,2, . . . , r, j 1,2, . . . , r, m1,2, . . . , r, n1,2, . . . , rsatisfying

G_ii< Y_iiii, i1,2, . . . , r, i /j, 3.4 2G_ii G_ij G_ji< Y_iiij Y_iiij^T Y_iiji Y_ijii, i1,2, . . . , r, j1,2, . . . , r, i /j, 3.5 G_ii G_ij G_ji< Y_iijj Y_iijj^T Y_jiij, i1,2, . . . , r, j1,2, . . . , r, i /j, 3.6

2G_ii 2G_ij 2G_im 2G_ji 2G_jm G_mi G_mj

< Y_iijm Y_ijim Y_ijmi Y_miji Y_mjii Y_miij Y_iijm^T Y_ijim^T Yimji Y_miji^T Y_mjii^T Y_miij^T , i1,2, . . . , r−3, j1,2, . . . , r−2, m1,2, . . . , r−1,

3.7

(6)

2

Gij Gim Gin Gji Gjm Gjn Gmi Gmj Gmn Gni Gnj Gnm

<

⎛

⎜⎜

⎝

Y_ijmn Y_ijnm Y_imjn Y_imnj Y_injm Y_inmj Yjinm Yjimn Yjmin Yjmni Yjnmi Yjnim

Y_ijmn^T Y_ijnm^T Y_imjn^T Y_imnj^T Y_injm^T Y_inmj^T Y_jinm^T Y_jimn^T Y_jmin^T Y_jmni^T Y_jnmi^T Y_jnim^T

⎞

⎟⎟

⎠ ,

i1,2, . . . , r−3, j1,2, . . . , r−2, m1,2, . . . , r−1, n1,2, . . . , r,

3.8

⎡

⎢⎢

⎣

Y_1ij1 Y_1ij2 · · · Y_1ijr Y2ij1 Y2ij2 · · · Y2ijr

... ... . .. ...

Yrij1 Yrij2 · · · Yrijr

⎤

⎥⎥

⎦

≤0, i1,2, . . . , r, j1,2, . . . , r. 3.9

In this case, if the conditions above are feasible, local state feedback gains are Kj NjQ⁻¹, j 1,2, . . . , r.

Proof. Consider a candidate of quadratic functionVxt x^TtP⁻¹xt. The equilibrium of 2.5is quadratically stable if

V˙ xt

x^Tt _r

i1 r j1

αiαj

QA^T_i AiQ−N^T_jB^T_i −BiNj

xt<0 ∀xt/0. 3.10

From inequality3.10above, the equilibrium of2.5is quadratically stable if

r i1

r j1

αiαj

QA^T_i AiQ−N_j^TB_i^T−BiNj

^r

i1 r j1

α_iα_jG_ij _r

i1

α_i ₂ _r

i1 r j1

α_iα_jG_ij

α₁ α₂ · · · α_r₂ ^r

i1 r j1

α_iα_jG_ij ^r

i1

α⁴_iG_ii

r i,j1

i /j

α³_iα_j

2G_ii G_ij G_ji ^r

i,j1 i /j

α²_iα²_j

G_ii G_ij G_ji

r−2 i1

r−1 ji 1

r mj 1

α²_iαjαm

2Gii 2Gij 2Gim 2Gji 2Gjm Gmi Gmj

r−3 i1

r−2 ji 1

r−1 mj 1

r nm 1

αiαjαmαn∗2

G_ij G_im G_in G_ji G_jm G_jn Gmi G_mj G_mn G_ni G_nj G_nm

Δ

(7)

∇<

r i1

α⁴_iYiiii r

i,j1 i /j

α³_iαj

Yiiij Yiiji Yijii Yjiii

^r

i,j1 i /j

α²_iα²_j

Yiijj Yjiji Yjiij

r−2 i1

r−1 ji 1

r mj 1

α²_iαjαm

⎛

⎝Y_iijm Y_iimj Y_imij Y_ijim Y_ijmi Y_imji Ymiji Y_jimi Y_mjii Y_jmii Y_miij Y_jiim

⎞

⎠

r−3 i1

r−2 ji 1

r−1 mj 1

r nm 1

α_iα_jα_mα_n

⎛

⎜⎜

⎝

Y_ijmn Y_ijnm Y_imjn Y_imnj Y_injm Y_inmj Yjinm Yjimn Yjmin Yjmni Yjnmi Yjnim

Y_ijmn^T Y_ijnm^T Y_imjn^T Y_imnj^T Y_injm^T Y_inmj^T Y_jinm^T Y_jimn^T Y_jmin^T Y_jmni^T Y_jnmi^T Y_jnim^T

⎞

⎟⎟

⎠

α1

⎛

⎜⎜

⎜⎝ α₁

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦

T⎡

⎢⎢

⎣

Y₁₁₁₁ Y₁₁₁₂ · · · Y_111r Y₂₁₁₁ Y₂₁₁₂ · · · Y_211r ... ... . .. ... Y_r111 Y_r112 · · · Y_ri11r

⎤

⎥⎥

⎦

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦ α2

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦

T⎡

⎢⎢

⎣

Y₁₁₂₁ Y₁₁₂₂ · · · Y_112r Y₂₁₂₁ Y₂₁₂₂ · · · Y_212r ... ... . .. ...

Y_r₁₂₁ Y_r122 · · · Y_r12r

⎤

⎥⎥

⎦

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦

· · · αr

⎡

⎢⎢

⎣ α1I α2I ... αrI

⎤

⎥⎥

⎦

T⎡

⎢⎢

⎣

Y11r1 Y11r2 · · · Y11rr

Y21r1 Y21r2 · · · Y21rr

... ... . .. ...

Yr1r1 Yr1r2 · · · Yr1rr

⎤

⎥⎥

⎦

⎡

⎢⎢

⎣ α1I α2I ... αrI

⎤

⎥⎥

⎦

⎞

⎟⎟

⎟⎠

α2

⎛

⎜⎜

⎝ α₁

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦

T⎡

⎢⎢

⎣

Y₁₂₁₁ Y₁₂₁₂ · · · Y_121r Y₂₂₁₁ Y₂₂₁₂ · · · Y_221r ... ... . .. ...

Y_r211 Y_r212 · · · Y_r21r

⎤

⎥⎥

⎦

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦ ∂2

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦

T⎡

⎢⎢

⎣

Y₁₂₂₁ Y₁₂₂₂ · · · Y_122r Y₂₂₂₁ Y₂₂₂₂ · · · Y_222r ... ... . .. ...

Y_r221 Y_r₂₂₂ · · · Y_r22r

⎤

⎥⎥

⎦

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦

· · · α_r

⎡

⎢⎢

⎣ α₁I α2I ... αrI

⎤

⎥⎥

⎦

T⎡

⎢⎢

⎣

Y_12r1 Y_12r2 · · · Y_12rr Y22r1 Y22r2 · · · Y22rr

... ... . .. ...

Yr2r1 Yr2r2 · · · Yr2rr

⎤

⎥⎥

⎦

⎡

⎢⎢

⎤

⎥⎥

⎦

⎞

⎟⎟

⎠

(8)

· · · αr

⎛

⎜⎜

⎝ α₁

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦

T⎡

⎢⎢

⎣

Y_1r11 Y_1r12 · · · Y_1r1r Y_2r11 Y_2r12 · · · Y_2r1r ... ... . .. ...

Y_rr11 Y_rr12 · · · Y_rr1r

⎤

⎥⎥

⎦

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦ α2

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦

T⎡

⎢⎢

⎣

Y_1r21 Y_1r22 · · · Y_1r2r Y_2r21 Y_2r22 · · · Y_2r2r ... ... . .. ...

Y_rr21 Y_rr22 · · · Y_rr2r

⎤

⎥⎥

⎦

⎡

⎢⎢

⎣ α₁I α₂I ... α_rI

⎤

⎥⎥

⎦

· · · α_r

⎡

⎢⎢

⎤

⎥⎥

⎦

T⎡

⎢⎢

⎣

Y_1rr1 Y_1rr2 · · · Y_1rrr Y2rr1 Y2rr2 · · · Y2rrr

... ... . .. ...

Yrrr1 Yrrr2 · · · Yrrrr

⎤

⎥⎥

⎦

⎡

⎢⎢

⎤

⎥⎥

⎦

⎞

⎟⎟

⎠

^r

i1 r j1

αiαj

⎛

⎜⎜

⎝

⎡

⎢⎢

⎤

⎥⎥

⎦

T⎡

⎢⎢

⎣

Y_1ij1 Y_1ij2 · · · Y_1ijr Y2ij1 Y2ij2 · · · Y2ijr

... ... . .. ...

Yrij1 Yrij2 · · · Yrijr

⎤

⎥⎥

⎦

⎡

⎢⎢

⎣ α₁I α2I ... αFrI

⎤

⎥⎥

⎦

⎞

⎟⎟

⎠ .

3.11 Thus, if3.9holds∇<0. In other words, the LPV system2.1is quadratically stabilizable via the gain-scheduled controller2.4.

Remark 3.6. The relationship between Theorems 3.3 and 3.5 is discussed here. One can find that in the case of j i, the conditions suggested herein reduce to the conditions of Theorem 3.3. That is,Theorem 3.3 is a special case of Theorem 3.5here. Consequently, with the increase of this positive integer d, the LMI-based conditions will provide more additional slack matrix variables which bring us more design freedom. Although the numerical complexity is increased much, a sequence of LMI-based conditions which are less and less conservative can be obtained. InSection 4, two simple numerical examples will be illustrated to compare the proposed conditions with the most recent existing conditions, where the additional decision variable is not added.

4. Numerical example

To illustrate the eﬀectiveness of the proposals, two simple numerical examples are given here.

All of LMIs-based conditions are solved by Matlab LMI toolbox22.

Example 4.1. Consider state-space expressions of two vertexes of an LPV plant as follows:

A1

1.59 −7.29 0.01 0

, B1 1

0

, A2

0.02 −4.64 0.35 0.21

, B2 8

0

, A3

−9 −4.33

0 0.05

, B3 3

−1

.

4.1

(9)

The problem is considered to seek an LPV state feedback controller as2.4such that closed- loop system is quadratically stable. The case ofd 2 is considered with and without the additional decision variable, respectively.

iWithout the additional decision variable19, after 21 iterations, these LMI-based conditions are not feasible. Therefore, an LPV state feedback controller cannot be found. Since the existing methods19provide necessary and suﬃcient conditions for such a state-feedback, there could exist a feasible solution in the case ofd >2.

iiWith the additional decision variable, according to 3.4–3.9, after 19 iterations, these LMI-based conditions can be solvable with

Q

1.037 −0.122

−0.122 0.029

, N1

−3.466 0.351

, N2

−1.285 −0.1333

, N3

0.8826 0.2923 .

4.2

Therefore, the vertex matrices of state feedback are given as K₁N₁⁻¹Q

−3.789 −3.788 , K₂N₂⁻¹Q

−3.451 −18.758 , K₃N₃⁻¹Q

3.929 26.099 .

4.3

Example 4.2. To illustrate the proposed approach, consider the problem of balancing an inverted pendulum on a cart. The equations of motion for the pendulum are as follows23:

˙ x₁x₂,

˙

x₂ gsin x1

−am/x₂² sin 2x1

/2−acos x1

u 4l/3−amlcos²

x1

, 4.4

where x₁ denotes the anglein radians of the pendulum from the vertical, and x₂ is the angular velocity.g 9.8 m/s² is the gravity constant,mis the mass of the pendulum, M is the mass of the cart, 2l is the length of the pendulum, andu is the force applied to the cartin Newtons:a 1/m M.We choosem 2.0 kg,M 8.0 kg, and 2l 1.0 m. We first represent the nonlinear system above by LPV model. Notice that whenx1 ±π/2, the system is uncontrollable. Hence, we approximate the system with state-space expressions of the vertex as follows:

A₁

⎡

⎣ 0 1 g 4l/3−aml 0

⎤

⎦, B₁

⎡

⎣ 0

− a

4l/3−aml

⎤

⎦,

A2

⎡

⎢⎣

0 1

2g π

4l/3−amlβ² 0

⎤

⎥⎦, B2

⎡

⎢⎣

0

− aβ

4l/3−amlβ²

⎤

⎥⎦,

4.5

whereβcos88^◦.

(10)

5 4

3 2

1 0

Times

−0.4

−0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

States

Figure 1: Trajectory of the states of this plant with initial valuesx0 −0.25 0.15^T.

According to the approach proposed here, several cases will be considered with the increasing of the scalard.

In the case ofd 0, after 39 iterations, due to the conservatism of the conditions, we cannot find a feasible solution to this state feedback.

In the case ofd≥1, we can find feasible solutions to this state feedback.

Whend1, after 5 iterations, we can obtain

W

0.0302 −0.185

−0.185 1.569

,

Z1

−1.833 39.95

, Z2

17.776 −80.410 .

4.6

Then, the state feedback is obtained as

K∂ ^r

j1

αjKj with K1Z1W⁻¹

346.9 66.42

, K2Z2W⁻¹

996.13 66.40 .

4.7

Here, we chooseα₁t δ, α₂t 1−δ,in whichδ :1.5701−x₁t/3.141. It is easy to check that theα_itare convex coordinates, since they satisfy 0≤α_it≤1, ₂

i1α_it 1.

The trajectory of the states of this plant can be drawn for the initial valuesx0 −0.25 0.15T

as shown inFigure 1.

From these numerical examples above, one can see that by adding an additional decision variable, at each step, these new conditions can provide less conservative or at least the same results than the most recent existing methods in the literature. Consequently, the feasible solutions can be obtained in much lower steps.

(11)

5. Conclusion

A sequence of new LMI-based conditions has been proposed for quadratic stabilization of LPV systems. One can find that with the increase of this positive integer d, a sequence of LMI-based conditions which are less and less conservative will be obtained. Here, we only present the conditions in the case of d 2. By adding an additional decision variable, at each step, these new conditions relaxed the conservatism of the previous existing works. As a result, the feasible solutions can be obtained in much lower steps.

Acknowledgments

This work is supported by the National Natural Science Foundation of ChinaGrant no.

60704022, 90816028, U0735003and the Guangdong Natural Science FoundationGrant no.

07006470.

References

1 J. S. Shamma and M. Athans, “Analysis of gain scheduled control for nonlinear plants,” IEEE Transactions on Automatic Control, vol. 35, no. 8, pp. 898–907, 1990.

2 J. S. Shamma and M. Athans, “Guaranteed properties of gain scheduled control for linear parameter- varying plants,” Automatica, vol. 27, no. 3, pp. 559–564, 1991.

3 G. Becker, A. Packard, D. Philbrick, and G. Blas, “Control of parametrically-dependent linear systems:

a single quadratic Lyapunov approach,” in Proceedings of the American Control Conference, pp. 2795–

2799, San Francisco, Calif, USA, June 1993.

4 P. Gahinet, P. Apkarian, and M. Chilali, “Aﬃne parameter-dependent Lyapunov functions and real parametric uncertainty,” IEEE Transactions on Automatic Control, vol. 41, no. 3, pp. 436–442, 1996.

5 P. Apkarian, P. Gahinet, and G. Becker, “Self-scheduledH^∞ control of linear parameter-varying systems: a design example,” Automatica, vol. 31, no. 9, pp. 1251–1261, 1995.

6 W. Xie and T. Eisaka, “Design of LPV control systems based on Youla parameterization,” IEE Proceedings: Control Theory and Applications, vol. 151, no. 4, pp. 465–472, 2004.

7 W. J. Rugh and J. S. Shamma, “Research on gain scheduling,” Automatica, vol. 36, no. 10, pp. 1401–

1425, 2000.

8 F. Wu, “A generalized LPV system analysis and control synthesis framework,” International Journal of Control, vol. 74, no. 7, pp. 745–759, 2001.

9 W. Xie, “H2 gain scheduled state feedback for LPV system with new LMI formulation,” IEE Proceedings: Control Theory and Applications, vol. 152, no. 6, pp. 693–697, 2005.

10 W. Xie and T. Eisaka, “Two-degree-of-freedom controller design for linear parameter-varying systems,” Asian Journal of Control, vol. 10, no. 1, pp. 115–120, 2008.

11 J. Bernussou, P. L. D. Peres, and J. C. Geromel, “A linear programming oriented procedure for quadratic stabilization of uncertain systems,” Systems & Control Letters, vol. 13, no. 1, pp. 65–72, 1989.

12 S. Boyd and Q. Yang, “Structured and simultaneous Lyapunov functions for system stability problems,” International Journal of Control, vol. 49, no. 6, pp. 2215–2240, 1989.

13 P. P. Khargonekar, I. R. Petersen, and K. Zhou, “Robust stabilization of uncertain linear systems:

quadratic stabilizability andH^∞control theory,” IEEE Transactions on Automatic Control, vol. 35, no.

3, pp. 356–361, 1990.

14 A. Packard and J. Doyle, “Quadratic stability with real and complex perturbations,” IEEE Transactions on Automatic Control, vol. 35, no. 2, pp. 198–201, 1990.

15 S. Boyd, L. El Ghaoui, E. Feron, and V. Balakrishnan, Linear Matrix Inequalities in System and Control Theory, vol. 15 of SIAM Studies in Applied Mathematics, SIAM, Philadelphia, Pa, USA, 1994.

16 E. Feron, P. Apkarian, and P. Gahinet, “Analysis and synthesis of robust control systems via parameter-dependent Lyapunov functions,” IEEE Transactions on Automatic Control, vol. 41, no. 7, pp.

1041–1046, 1996.

17 P. Gahinet, P. Apkarian, and M. Chilali, “Aﬃne parameter-dependent Lyapunov functions and real parametric uncertainty,” IEEE Transactions on Automatic Control, vol. 41, no. 3, pp. 436–442, 1996.

(12)

18 W. Xie, “Quadratic stabilization of LPV system by an LTI controller based on ILMI algorithm,”

Mathematical Problems in Engineering, vol. 2007, Article ID 28262, 9 pages, 2007.

19 R. C. L. F. Oliveira and P. L. D. Peres, “Stability of polytopes of matrices via aﬃne parameter- dependent Lyapunov functions: asymptotically exact LMI conditions,” Linear Algebra and Its Applications, vol. 405, pp. 209–228, 2005.

20 G. H. Hardy, J. E. Littlewood, and G. P ´olya, Inequalities, Cambridge University Press, Cambridge, UK, 2nd edition, 1952.

21 C.-H. Fang, Y.-S. Liu, S.-W. Kau, L. Hong, and C.-H. Lee, “A new LMI-based approach to relaxed quadratic stabilization of T-S fuzzy control systems,” IEEE Transactions on Fuzzy Systems, vol. 14, no.

3, pp. 386–397, 2006.

22 P. Gahinet, A. Nemirovskii, A. J. Laub, and M. Chilali, “LMI Control Toolbox User’s Guide,” The MathWorks, Natick, Mass, USA, 1995.

23 W. Xie, “An equivalent LMI representation of bounded real lemma for continuous-time systems,”

Journal of Inequalities and Applications, vol. 2008, Article ID 672905, 8 pages, 2008.