Sie sind auf Seite 1von 7

152

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 1, JANUARY 2010

case when p = 1 (BIMBO method), under assumptions of Gaussianity. In particular, we have proved that this approach leads to consistent estimates of the system parameters, regardless of the variance of the measurement noise at the quantizers input or of a dithering signal. Moreover, we have established that the corresponding criterion is locally convex, which makes it amenable to simple optimization tools. The extension of this approach to less simple systems (IIR or nonlinear) or to a framework of online identication is also being studied.

Robust Stability and Stabilization of Fractional-Order Interval Systems with the Fractional Case Order : The
Jun-Guo Lu and Yang-Quan Chen, Senior Member, IEEE

REFERENCES
[1] L. Y. Wang et al., System identication using binary sensors, IEEE Trans. Autom. Control, vol. 48, no. 11, pp. 18921907, Nov. 2003. [2] L. Ljung, System IdenticationTheory for the User. Englewood Cliffs, NJ: Prentice Hall, 1999. [3] E. Walter and L. Pronzato, Identication of Parametric Models From Experimental Data. New York: Springer, 1997. [4] M. Negreiros et al., Ultimate low cost analog BIST, in Proc. Design Autom. Conf., 2003, pp. 570573. [5] J. Juillard and E. Colinet, Initialization of the BIMBO self-test method using binary inputs and outputs, in Proc. 46th IEEE Conf. Decision Control, 2008, pp. 161166. [6] T. Wigren, Adaptive ltering using quantized output measurements, IEEE Trans. Signal Processing, vol. 46, no. 12, pp. 34233426, Dec. 1998. [7] L. Y. Wang et al., Joint identication of plant rational models and noise distribution functions using binary-valued observations, Automatica, pp. 535547, 2006. [8] Y. Zhao et al., Identication of Wiener systems with binary-valued output observations, Automatica, pp. 17521765, 2007. [9] E. Rafaljowicz, Linear systems identication from random threshold binary data, IEEE Trans. Signal Processing, vol. 44, no. 8, pp. 20642070, Aug. 1996. [10] E. Rafaljowicz, System identication from cheap, qualitative output observations, IEEE Trans. Autom. Control, vol. 41, no. 9, pp. 13811385, Sep. 1996. [11] E. Bai and J. Reyland, Towards identication of Wiener systems with the least amount of a priori information on the nonlinearity, Automatica, pp. 910919, 2008.

AbstractThis technical note rstly presents a sufcient and necessary condition for the robust asymptotical stability of fractional-order interval 1. And then a sufsystems with the fractional order satisfying 0 cient condition for the robust asymptotical stabilization of such fractionalorder interval systems is derived. All the results are obtained in terms of linear matrix inequalities. Finally, two illustrative examples are given to show that our results are effective for checking the robust stability and designing the robust stabilizing controller for fractional-order interval systems. Index TermsFractional-order (FO) system, interval system, linear matrix inequality (LMI), stability, stabilization.

I. INTRODUCTION Recently, fractional-order control systems have attracted increasing interest [1][3]. On the one hand, this is mainly due to the fact that many real-world physical systems are well characterized by fractional-order state equations [1], i.e., equations involving the so-called fractional derivatives and integrals. On the other hand, with the success in the synthesis of real noninteger differentiators and the emergence of a new electrical circuit element called fractance [4], [5], fractional-order controllers [6][8] have been designed and applied to control a variety of dynamical processes, including integer-order and fractional-order systems, so as to enhance the robustness and performance of the control systems. Stability is fundamental to all control systems, certainly including fractional-order control systems [9][19]. Recently, stability and stabilization problems of fractional-order linear time-invariant (FO-LTI) interval systems have been investigated in [13][18]. For example, for FO-LTI interval systems described in the transfer function form, the stability issue was discussed rst in [13] and then further in [14]. Note that, in [13], [14], the results were based on an experimentally veried Kharitonov-like procedure and only for single-input single-output FO-LTI systems. For FO-LTI interval systems described in the statespace form, the robust stability problem was tackled in [15], where the matrix perturbation theory was used to nd the ranges of the corresponding interval eigenvalues. As commented in [15], the result is rather conservative. To reduce the conservatism, in [16], a new robust

Manuscript received May 11, 2009; revised August 09, 2009. First published November 24, 2009; current version published January 13, 2010. This work was supported in part by the National Natural Science Foundation of China under Grants 60404005, 60744002, 60775062 and 60974002, the Shanghai Natural Science Foundation under Grant 09ZR1414200, the National High Technology Research and Development Program of China under Grant 2006AA040203, and the Program for New Century Excellent Talents in University under Grant NCET-07-0538. Recommended by Associate Editor K. Morris. J.-G. Lu is with the Department of Automation, Shanghai Jiaotong University, Shanghai 200240, China (e-mail: jglu@sjtu.edu.cn). Y.-Q. Chen is with the Center for Self-Organizing and Intelligent Systems (CSOIS), Department of Electrical and Computer Engineering, College of Engineering, Utah State University, Logan, UT 84322-4160 USA (e-mail: yqchen@ece.usu.edu). Color versions of one or more of the gures in this technical note are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TAC.2009.2033738 0018-9286/$26.00 2009 IEEE

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 1, JANUARY 2010

153

stability checking method was proposed for FO-LTI interval systems by utilizing the Lyapunov inequality to nding the maximum eigenvalue of a Hermitian matrix. However the results presented in [15] and [16] only provide sufcient conditions. [17] extended the results of [16] to establish a necessary and sufcient condition for the robust stability of FO-LTI interval systems with fractional order , 1  < 2 by using the complex Lyapunov inequality. [18] presented the necessary and sufcient conditions for the robust stability and stabilization of FO-LTI interval systems with fractional order , 1  < 2 by using the linear matrix inequality. Note that all the above-mentioned results about the robust stability and stabilization for uncertain FO-LTI interval systems described in the state-space form only consider the fractional order belonging to 1  < 2. The robust stability and stabilization problems for the fractional order belonging to 0 < < 1 is still open to our best knowledge. With the above motivation, the robust stability and stabilization problems of uncertain FO-LTI interval systems with the fractional order belonging to 0 < < 1 are investigated in this technical note. A sufcient and necessary robust stability condition is rstly presented. And then a sufcient robust stabilization condition is derived. All the results are obtained in terms of linear matrix inequalities. Finally, two illustrative examples are used to demonstrate the effectiveness of our proposed results. Notations and Facts: We denote by M T the transpose of M , by 3 ~ the conjugate of the scalar M the transpose conjugate of M , by z number z , by Re(z ) its real part and by Im(z ) its imaginary part. I means the unit matrix with appropriate dimensions, In is the identity matrix of order n. diag (a1 ; a2 ; 1 1 1 ; an ) denotes the block-diagonal matrix. Matrices, if not explicitly stated, are assumed to have appropriate dimensions.
is the Kronecker product of two matrices and (A
B )(C
D) = (AC )
(BD). i denotes the imaginary unit. SymfX g denotes the expression X T + X . Fact 1: A complex Hermitian matrix H satises H < 0 if and only if

of classical integer-order derivatives with clear physical interpretations [1]:


(t) 1 D f (t) = d f = 0( 0 m) dt t
0

f (m) ( ) (t 0  ) +10m

(4)

where m is an integer satisfying m 0 1 <  m. The robust stability and stabilization problem for FO-LTI interval systems with 1  < 2 have been investigated in [15][18]. This technical note focuses on the robust stability and stabilization problem for the case of 0 < < 1. To deal with the uncertainty interval, we introduce the following notations:

); 1A = 1 (A  0 A) = f ij gn2n A0 = 1 (A + A

B0

2 1  ); = (B + B 2

1B

2 1  0 B ) = f ij gn2p : = (B 2

(5) (6)

It can be seen that all elements of 1A and 1B are nonnegative, so we can dene
n 1 1 1 en 1 1 1 DA = [ 11 e1 n 1 p n1 en 1 1 1 p1 n nn en ]n2n n p p n n EA = [ 11 e1 1 1 1 1n en 1 1 1 p n1 en 1 1 1 p nn en ]T 1 n n 2n n n DB = 11 e1 1 1 1 1p e1 1 1 1 n 1 1 1 en n1 en np n p 1 1 1 ep 1 1 1 11 e1 1p p p 1 1 1 ep n1 e1 np p n2(np)

(7) (8)

(9)

EB =

T
(np)2p

(10)

Re(H ) Im(H ) 0Im(H ) Re(H )

< 0:

n 2 n and ep 2 p denote the column vectors with the k th where ek k element being 1 and all the others being 0. Also, denote

II. PROBLEM FORMULATION AND PRELIMINARIES Consider the following FO-LTI interval system:

HA = diag(11 ; . . . ; 1n ; . . . ; n1 ; . . . nn ) 2 Rn 2n ; jij j  1; i; j = 1; . . . ; ng (11) (np)2(np) HB = diag(11 ; . . . ; 1p ; . . . ; n1 ; . . . np ) 2 R ; jij j  1; i = 1; . . . ; n; j = 1; . . . ; pg : (12)
(1) Here, for AI and BI , we have the following lemma: Lemma 1: [18] Let

d x(t) = Ax(t) + Bu(t) dt

where is the fractional commensurate order, x(t) 2 n and u(t) 2 p denote the state and control vector, respectively, the system matrices A and B are interval uncertain satisfying

AJ = fA = A0 + DA FA EA jFA 2 HA g BJ = fB = B0 + DB FB EB jFB 2 HB g:

(13)

] A 2 AI =[A; A ij ; 1  i; j  n = [aij ] : aij  aij  a ] B 2 BI =[B; B bij ; 1  i  n; 1  j  p = [bij ] : bij  bij  

(2) (3)

Then AI = AJ , BI = BJ . To proof the main results in the next section, we need the following lemmas. Lemma 2: [21] For any matrices X and Y with appropriate dimensions, we have

 = [ aij ]n2n satisfy aij  a where A = [aij ]n2n , A ij for all 1   = [ bij for all i; j  n, and B = [bij ]n2p , B bij ]n2p satisfy bij   1  i  n, 1  j  p. The following Caputo denition is adopted for fractional derivatives of order of function f (t), because the Laplace transform of the Caputo derivative allows utilization of initial values

X T Y + Y T X  "X T X + (1=")Y T Y; for any " > 0:


Lemma 3: [20] Let X; Y; F be real matrices of suitable dimensions. Then, for any x 2 n

max (xT XF Y x)2 : F T F

I

= (xT XX T x)(xT Y T Y x):

154

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 1, JANUARY 2010

Lemma 4: [20] Let X; Y; Z be n 2 n symmetric matrices such as X 0, Y < 0 and Z  0. Furthermore, assume that (xT Y x)2 0 T 4(x Xx)(xT Zx) > 0 for all nonzero x 2 n . Then, there exists a constant  > 0 such that 2 X + Y + Z < 0. From [19], we have the following Lemma: Lemma 5: [19] Let A 2 n2n , 0 < < 1 and  = (1 0 )(=2). The fractional-order system (d x(t)=dt ) = Ax(t) is asymptotically stable if and only if there exist two positive denite Hermitian matrices n2n and Q = Q3 2 n2n such that e Q A3 + Q01 = Q3 1 2 1 2 2 0  3   0 A3 Q1 + e0 AQ1 + e Q 2 A + e AQ2 < 0, or e Q1 A + e e Q2 A + e A3 Q2 < 0.
i i i i i i i i

which can be rewritten as

Sym AP11 sin  + AP12 cos  2 2  + Sym AP21 sin 0 AP22 cos  2 2  + (P11 AT 0 AP11 ) cos 2  T T + AP12 0 P12 A sin 2  + (AP21 0 P21 AT ) cos 2 T AT sin  < 0: + AP22 0 P22 2
i i i i

(19)

III. MAIN RESULTS In this section, a sufcient and necessary condition is rstly derived for robust asymptotical stability of the uncertain FO-LTI interval system (1). To proceed, we need the following theorem. 1. The fractional-order Theorem 1: Let 2 n2n and 0 system ( ( ) ) = ( ) is asymptotically stable if and only if there exist two real symmetric positive denite matrices k1 2 n2n , = 1, 2, and two skew-symmetric matrices k2 2 n2n , = 1, 2, such that

d x t =dt

Ax t

< <

According to Fact 1, (19) is equivalent to (14). This ends the proof. In the following, a sufcient and necessary condition is derived for robust asymptotical stability of the uncertain FO-LTI interval system (1) in terms of LMIs. Theorem 2: The uncertain FO-LTI interval system (1) with input ( ) = 0 and 0 1 is robustly asymptotically stable if and only if there exist two real symmetric positive denite matrices k1 2 n2n , = 1, 2, two skew-symmetric matrices k2 2 n2n , = 1, 2, and 0( = 1 2), such that scalar constants ij

ut k

< <

" > i; j

P k

Sym f2ij
(APij )g < 0 i=1 j =1 P11 P12 P21 P22 0P12 P11 > 0; 0P22 P21 > 0
where

(14) (15) where


2

P11 P12 0P12 P11 > 0;

11 =
12 =
2

11 T
12

0P22

P21

12
22

<0 P22 > 0 P21

(20) (21)

0 212 = 0 221 = 0 0 2 222 = : (16) 0 2 0 cos  2 Proof: Dene Pk1 = Re(Qk ), Pk2 = Im(Qk ), k = 1, 2. Since Pk1 0 PkT2 = Pk1 + Pk2 , e = cos  + sin  and e0 = cos  0 sin  , it follows from Lemma 5 that (d x(t)=dt ) = Ax(t) with 0 < < 1 is asymptotically stable if and only if there exist two real symmetric positive denite matrices Pk1 2 n2n , k = 1, 2, and two skew-symmetric matrices Pk2 2 n2n , k = 1, 2, such that
211 = sin  2 cos  2 cos  2 sin  2 sin  2 cos  2 cos  2 sin 
i
i

cos  2 sin  2 sin  2 cos  2 cos  2 sin  2 sin 

I2
(EA P11 )T I2
(EA P12 )T I2
(EA P21 )T I2
(EA P22 )T
22 = 0 diag ("11 ; "12 ;"21 ;"22 )
I2n and 2ij (i;j = 1; 2) satisfy (16).
2 2

i=1 j =1

T Sym f2ij
(A0 Pij )g + "ij I2
DA DA

Proof: (Sufciency) Suppose that (20), (21) hold. It follows from Lemma 1 that

i=1 j =1

Symf2ij
(APij )g =
+
2 2

i=1 j =1

Symf2ij
(A0 Pij )g
(22)

i=1 j =1

Sym f2ij
(DAFA EA Pij )g :

(15), respectively. Note that, (18) is equivalent to

P11 + P12 > 0; P21 + P22 > 0 (17) T AT (cos  + sin ) P11 0 P12 + (cos  0 sin )A(P11 + P12 ) T AT + (cos  0 sin ) P21 0 P22 + (cos  + sin )A(P21 + P22 ) < 0 (18) where  = (1 0 )(=2). According to Fact 1, (17) is equivalent to
i i i i i i i i i i

Note that (11) and (13) implies


T  I: FA FA

(23)

It follows from (23) that


T I2
FA )(I2
FA )T = (I2
FA ) I2
FA T  I: = I2
FA FA (24) T = I (i;j = 1; 2), it follows from (24) and Lemma Note that 2ij 2ij 2 2 that for any real scalars "ij > 0 (i;j = 1; 2) Sym f2ij
(DA FA EA P11 )g = Sym f(2ij
DA )(I2
FA )(I2
EA Pij )g  "ij (2ij
DA )(I2
FA )(I2
FA )T (2ij
DA )T 01(I2
EA Pij )T (I2
EA Pij ) + "ij

SymfAP11 cos  + AP12 sin  + AP21 cos  0 AP22 sin g T AT + AP + (P11 AT 0 AP11 ) sin  + 0P12 12 cos  T AT + AP + (0P21 AT + AP21 ) sin  + 0P22 22 cos  <0
i i i i

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 1, JANUARY 2010

155

 " I2
D D + "01 (I2
E 2 (I2
E P11 ):
ij A T A ij A

P11

Consequently, given any  (25)

2 R2

and 

6= 0, one has
A A

Substituting (25) into (22), one has


2 2

( 811  )2 > 4 max


T

T

Sym
i=1 j =1

f2
(AP )g
ij ij

~ (I2
E P11 ) (211
D )F ~ F ~ I F
A T A A A A A

;
2

(32)

Dene
Sym

i=1 j =1

2
(A0 P ) + "
ij ij

ij

I2

D
P11

D :

T A

5=
(26)

T

~ (I2
E (211
D )F

P11

~ F ~ I )  jF
T A A T ; FA FA

i=1 j =1

01 I2 "ij

(
E

P11

) (I2
E
T

Noting that 0y

max

T

2 5 for any y 2 5, we have ~ (I2


E P11 ) (211
D )F
A A A

~ ~ I )

2

Taking (26) into account and using the Schur complement of (20), one obtains
2 2

= max

T

Sym
i=1 j =1

f2
(AP )g < 0:
ij ij

(27)

~ (I2
E (211
D )F ~ I : ~ F F
A A T A A A A

P11

(33)

By using Lemma 3, one obtains from (32) and (33) that

It follows from the above inequality (27) and Theorem 1 that the uncertain FO-LTI interval system (1) with input u(t) = 0 and 0 < < 1 is asymptotically stable. Necessity: Suppose that the FO-LTI interval system (1) with u(t) = 0 is asymptotically stable. It follows from Theorem 1 n that there exist 2n , k = 1, 2, two real symmetric positive denite matrices Pk1 2 n2n , k = 1, 2, such that and two skew-symmetric matrices Pk2 2
2 2

( 811  )2 > 4
T

T

(211
D )(211
D )  2  (I2
E P11 ) (I2
E
T T A T

P11  :

(34)

Now, applying Lemma 4 to (34), it follows that there exists a real scalar "11 > , such that

2 "11

Sym
i=1 j =1

f2
(AP )g < 0
ij ij

(28)

(211
D )(211
D ) + "11 811 + (I2
E P11 ) (I2
E
A A T A T

P11 < :

) 0

Therefore

where 2ij (i; j = 1; 2) satisfy (16). Note that (11) and (13) implies
F FA
T A

 I:

(29)

~A Dene F
T A A

= I2
F

. It follows from (29) that


T A

~ F ~ = (I2
F ) (I2
F ) F = I2
F (I2
F ) =
A T A A

I2

T A

FA

 I:

(30)

= 811 + "11 (211


D )(211
D ) 01(I2
E P11 ) (I2
E P11 ) + "11 ~ (I2
E P11 ) = T00 0 Sym (211
D )F + "11 (211
D )(21
D ) 01(I2
E P11 ) (I2
E P11 ) < 0: + "11 (35) ~ (I2
E P12 )g. Inequality Let 812 = T11 0 Symf(212
D )F (35) implies that for all 1 ; 2 2 R ,  = [1 2 ] 6= 0
T11
A A T A T A A A A A A T A T A A A n T A T T

Using Lemma 1 with some calculations, one can obtain from (28) that
T00

Sym
i=1 j =1

f2
(A0 P )g
ij ij

 T 12  <

~ (I2
E 02 (212
D )F
T A A

P12

:

Hence
A A

Sym
i=1 j =1

~ (I2
E (2
D )F
ij A

Pij

< :

(31)

 T 12  <

02 max

T

~A (I2
EA P11 )g. Inequality Let 811 = T00 0 Symf(211
DA )F T T T 2 ] 6= 0 (31) implies that for all 1 ; 2 2 Rn ,  = [1

T

~ (I2
E P12 ) ; (212
D )F ~  I  0: ~ F F
A A A T A A

Consequently, given any 

2 R2

and 

6= 0, one has
A A

~ (I2
E 811  < 02 (211
D )F
T A A

P11

:

( 812  )2 > 4 max


T

Hence
 T 11  <

02 max

T

~ (I2
E P11 ) ; (211
D )F ~  I  0: ~ F F
A A A T A A

~ (I2
E (212
D )F ~ F ~ I 2 F ~ (I2
E = 4 max  (211
D )F ~ I : ~ F F
T
A T A A T A A T A A

P12

) )

;

P11

(36)

156

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 1, JANUARY 2010

Therefore
T22 = 822 + "22 (222
22

~A (I2
EA P22 ) 0 Sym (222
DA )F T + "22 (222
DA )(222
DA ) 01 T + "22 (I2
EA P22 ) (I2
EA P22 ) 2 2 Sym f2ij
(A0 Pij )g = = T21

01 T + " (I2
EA P22 ) (I2
EA P22 )

DA )(222
DA )T

i=1 j =1
2

i=1 j =1
2 2

"ij (2ij "ij (I2

DA )(2ij
DA )T
EA Pij )T (I2
EA Pij )
(38)

01

i=1 j =1

< 0:
Fig. 1. Time response of the randomly selected system in Example-1.

T = I (i; j = 1; 2), it follows from (38) that Note that 2ij 2ij 2
2 2

T22 =

By using Lemma 3 and (36), one obtains


T 2 T ( 812  ) > 4  (212

i=1 j =1
+

Sym 2ij
2 2

(A0 Pij ) + "ij

I2

T
DA DA

DA )(212
DA )T  2  T (I2
EA P12 )T (I2
EA P12 )

(37)

i=1 j =1

"ij (I2

01

EA Pij )T (I2
EA Pij ) < 0:

(39)

Now, applying Lemma 4 to (37), it follows that there exists a real scalar "12 > 0, such that
"12 (212
2

DA )(212
DA )T + "12 812 + T (I2
EA P12 ) (I2
EA P12 ) < 0:
DA )(212
0 1 T + "12 (I2
EA P12 ) (I2
EA P12 ) ~A (I2
EA P12 ) = T11 0 Sym (212
DA )F T + "12 (212
DA )(212
DA ) 01 T + "11 (I2
EA P12 ) (I2
EA P12 ) < 0:
T DA )

By the Schur complement [22], (20) can be easily obtained from (39). This ends the proof. Next, a stabilization result is established. Theorem 3: The uncertain FO-LTI interval system (1) with input u(t) = Kx(t) and 0 < < 1 is asymptotically stabilizable if there are a matrix X 2 Rp2n , a symmetric positive-denite matrix Q 2 Rn2n , and four real scalars i > 0 and i > 0, (i = 1; 2), such that
0= 011 021 012 022

Therefore
T12 = 812 + "12 (212

<0

(40)

where
2

011 =

i=1
+

Sym 2i1
2

(A0 Q + B0 X )g
2

~A (I2
EA P21 )g, it Similarly, Let 821 = T12 0 Symf(221
DA )F follows that there exists a real scalar "21 > 0, such that

"21 (221

DA )(221
DA )T +
"21 821 + (I2

EA P21 )T (I2
EA P21 ) < 0:

(EA Q)T T I2
(EB X ) 022 = 0 diag ( 1 ; 2 ; 1 ; 2 )
I2n :
012 =

I2

(EA Q)T T I2
(EB X )

i=1

i I2

T
DA DA
I2

i=1

i I2

T
DB DB

Therefore
T21 = 821 + "21 (221

DA )(221
DA )T 01 T + "21 (I2
EA P21 ) (I2
EA P21 ) ~A (I2
EA P21 ) = T12 0 Sym (221
DA )F T + "21 (221
DA )(221
DA ) 0 1 T + "21 (I2
EA P21 ) (I2
EA P21 ) < 0:

Moreover, a stabilizing state-feedback gain matrix is given by K = XQ01 . Proof: The uncertain FO-LTI interval system (1) with input u(t) = Kx(t) is asymptotically stable for some matrix K if the fractional-order system d x(t) = (A + BK )x(t) (41) dt is asymptotically stable. It follows from Theorem 1 that this is equivalent to that there exist two real symmetric positive denite matrices Pk1 2 n2n , k = 1, 2, and two skew-symmetric matrices Pk2 2 n2n , k = 1, 2, such that
2 2

~A (I2
EA P22 )g, it follows Let 822 = T21 0 Symf(222
DA )F that there exists a real scalar "22 > 0, such that

"22 (222

DA )(222
DA )T +
"22 822 + (I2

EA P22 )T (I2
EA P22 ) < 0:

i=1 j =1

Sym 2ij

(Ac Pij )g < 0:

(42)

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 1, JANUARY 2010

157

Ac = A + BK and 2ij (i; j = 1; 2) satisfy (16). By setting P11 = P21 ; = Q, P12 = P22 = 0 in (42), one can conclude that if
where

Sym f211
(Ac Q)g + Sym f221
(Ac Q)g < 0

(43)

the uncertain FO-LTI interval system (1) is asymptotically stabilizable. Similar to the proof of Theorem 2, (43) is equivalent to that there exist a symmetric positive-denite matrix Q 2 Rn2n and four real scalars i > 0 and i > 0, (i = 1, 2) such that

9=

2
i=1

Sym f2i1
(A0c Q)g +
2

2
i=1

T i I2
DA DA

i=1

i01 (I2
EA Q)T (I2
EA Q)
T + i I2
DB DB

+ i01(I2
EB KQ)T (I2
EB KQ) i=1 <0
where A0c = A0 + B0 K . Substitute A0c set X = KQ, we have
2
i=1

Fig. 2. Time response of the selected system in Example-2 with

( ) = 0.

(44)

= A0 + B0 K into (44) and


2
i=1

7=

Sym f2i1
(A0 Q + B0 X )g +
2

i I2
DA D

T A

A=

01:3696 0:5954 1:0961 01:0948 03:3785 0:9047 00:4898 01:2956 02:7861

2 AI :

i=1

1 T T 0 i (I2
EA Q) (I2
EA Q)+ i (I2
DB DB )

is shown in Fig. 1, which shows that it is asymptotically stable and its states converge to zero. B. Example 2 (45) Consider the robust stabilization of the following uncertain FO-LTI interval system:

+ i01 (I2
EB X )T (I2
EB X ) i=1 < 0:

Inequality (45) is equivalent to (40) by the Schur complement [22]. This ends the proof. Remark 1: Based on Lemma 5, LMI methods have been used for the robust stability and stabilization of FO-LTI interval systems with fractional order , 0 < < 1, which corrects our previous opinion in [18] that LMI methods could not be used for such case 0 < < 1. IV. NUMERICAL EXAMPLES A. Example 1 Consider the robust stability of the following uncertain FO-LTI interval system:

d x(t) = Ax(t) + Bu(t) dt ]; B 2 BI = [B; B  ], with where = 0:5, and A 2 AI = [A; A 00:3 0:1 0:1 1 A = 01 03 1 ; B = 0:5 00:6 01:5 02:5 0:5 0:3 0:6 1:2 2 = 1 :  = 00:8 02:4 1:2 ; B A 00:4 01:2 02:0 1

(47)

d x(t) = Ax(t) dt ] with where = 0:9, and A 2 AI = [A; A 01:95 0:35 0:7 A = 01:3 03:9 0:7 00:65 01:95 03:25 01:05 0:65 1:3  = 00:7 02:1 1:3 : A 00:35 01:05 01:75

(46)

Using the Matlab LMI toolbox, we nd that the linear matrix inequalities (20), (21) in Theorem 2 are feasible, which concludes that the fractional order interval system (46) is robust asymptotically stable. For example, the time response of the randomly selected system with

Using the Matlab LMI toolbox, for the fractional order interval system (47) with control u(t) = 0, we nd that the linear matrix inequalities (20), (21) in Theorem 2 are not feasible, which concludes that the fractional order interval system (47) with control u(t) = 0 isnt robust asymptotically stable. For example, the time response of the frac 2 AI = [A; A ] and control tional-order system (47) with A = A u(t) = 0 is shown in Fig. 2, which shows that it is not asymptotically stable and its states are not convergent. But, using the Matlab LMI toolbox, we nd that the linear matrix inequality (40) in Theorem 3 is feasible, which concludes that the fractional order interval system (47) is asymptotically stabilizable. From Theorem 3, the stabilizing state-feedback gain matrix is obtained as K = [016:3511 0 4:6507 0 2:8007]. The time response of the randomly selected system with

A + BK =

025:4640 06:0315 02:4123 09:6304 05:4912 01:5186 013:5531 03:6948 03:5583

2 AI

+ BI K

158

IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 55, NO. 1, JANUARY 2010

Fig. 3. Time response of the randomly selected system in Example-2 with ( ) = ( ).

is shown in Fig. 3, which shows that it is asymptotically stable and its states converge to zero. V. CONCLUSION This technical note has presented a sufcient and necessary condition for the robust asymptotical stability of fractional-order interval systems with the fractional order belonging to 0 < < 1, and a sufcient condition for the robust asymptotical stabilization of such fractional-order interval systems. All the results are expressed in terms of LMIs and are very convenient to be used in practice. Finally, two illustrative examples have shown the effectiveness of our results. ACKNOWLEDGMENT The authors would like to thank the Associate Editor and the reviewers for their constructive comments and suggestions.

[11] C. Bonnet and J. R. Partington, Analysis of fractional delay systems of retarded and neutral type, Automatica, vol. 38, no. 7, pp. 11331138, 2002. [12] C. Hwang and Y. C. Cheng, A numerical algorithm for stability testing of fractional delay systems, Automatica, vol. 42, pp. 825831, 2006. [13] I. Petr s, Y. Q. Chen, and B. M. Vinagre, Robust Stability Test for Interval Fractional Order Linear Systems, V. D. Blondel and A. Megretski, Eds. Princeton, NJ: Princeton Univ. Press, Jul. 2004, vol. 208-210, ch. 6.5. [14] I. Petr s, Y. Q. Chen, B. M. Vinagre, and I. Podlubny, Stability of linear time invariant systems with interval fractional orders and interval coefcients, in Proc. Int. Conf. Comput. Cybern. (ICCC04), Viena, Austria, August 30September 1 2005, pp. 14. [15] Y. Q. Chen, H. S. Ahn, and I. Podlubny, Robust stability check of fractional order linear time invariant systems with interval uncertainties, Signal Processing, vol. 86, pp. 26112618, 2006. [16] H. S. Ahn, Y. Q. Chen, and I. Podlubny, Robust stability test of a class of linear time-invariant interval fractional-order system using Lyapunov inequality, Appl. Math. Comput., vol. 187, no. 1, pp. 2734, 2007. [17] H. S. Ahn and Y. Q. Chen, Necessary and sufcient stability condition of fractional-order interval linear systems, Automatica, vol. 44, no. 11, pp. 29852988, 2008. [18] J. G. Lu and G. R. Chen, Robust stability and stabilization of fractional-order interval systems: An LMI approach, IEEE Trans. Autom. Control, vol. 54, no. 6, pp. 12941299, Jun. 2009. [19] J. Sabatier, M. Moze, and C. Farges, On stability of fractional order systems, in Proc. 3rd IFAC Workshop Fractional Differentiation Appl., Ankara, Turkey, Nov. 0507, 2008, [CD ROM]. [20] I. R. Petersen, A stabilization algorithm for a class of uncertain linear systems, Syst. Control Lett., vol. 8, pp. 351357, 1987. [21] P. P. Khargonekar, I. R. Petersen, and K. Zhou, Robust stabilization control of uncertain linear systems: Quadratic stabilizability and theory, IEEE Trans. Autom. Control, vol. 35, no. 3, pp. 356361, Mar. 1990. [22] S. Boyd, L. E. Ghaoui, E. Feron, and V. Balakrishnan, Linear Matrix Inequalities in System and Control Theory. Philadelphia, PA: SIAM, 1994.

REFERENCES
[1] I. Podlubny, Fractional Differential Equations. New York: Academic Press, 1999. [2] J. A. T. Machado, Special issue on fractional calculus and applications, Nonlin. Dynam., vol. 29, pp. 1385, Mar. 2002. [3] M. D. Ortigueira and J. A. T. Machado, Special issue on fractional signal processing and applications, Signal Processing, vol. 83, no. 11, pp. 22852480, Nov. 2003. [4] M. Nakagava and K. Sorimachi, Basic characteristics of a fractance device, IEICE Trans. Fund., vol. E75-A, no. 12, pp. 18141818, 1992. [5] S. Westerlund, Capacitor theory, IEEE Trans. Dielectr. Electron. Insul., vol. 1, no. 5, pp. 826839, Oct. 1994. [6] I. Podlubny, Fractional-order systems and -controllers, IEEE Trans. Autom. Control, vol. 44, no. 1, pp. 208214, Jan. 1999. [7] H. Raynaud and A. Zerganoh, State-space representation for fractional order controllers, Automatica, vol. 36, pp. 10171021, 2000. [8] A. Oustaloup, B. Mathieu, and P. Lanusse, The CRONE control of resonant plants: Application to a exible transmission, Eur. J. Control, vol. 1, no. 2, pp. 113121, 1995. [9] S. B. Skaar, A. N. Michel, and R. K. Miller, Stability of viscoelastic control systems, IEEE Trans. Autom. Control, vol. AC-33, no. 4, pp. 348357, Apr. 1988. [10] D. Matignon, Stability result on fractional differential equations with applications to control processing, in Proc. IMACS, IEEE-SMC, Lille, France, Jul. 1996, pp. 963968.

Das könnte Ihnen auch gefallen