A Delay-Dependent Stability Criterion For Systems by PooGyeon Park

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

876 IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 44, NO.

4, APRIL 1999

A Delay-Dependent Stability Criterion for Systems Lemma 1: Assume that a( ) 2 Rn and b( ) 2 Rn are given
with Uncertain Time-Invariant Delays for 2 . Then, for any positive definite matrix X 2 Rn 2n and
any matrix M 2 Rn 2n , the following holds:
PooGyeon Park
02 bT ( )a( ) d
T
Abstract—This paper provides a new stability criterion for systems with
 a( ) X XM a( )
time-invariant uncertain delays. Based on an improved upper bound for b( ) M T X (2; 2) b( ) d (2)
the inner product of two vectors, a new delay-dependent robust stability
criterion is derived, which is shown by an example less conservative than where (2; 2) denotes (M T X + I )X 01 (XM + I ).
existing stability criteria.
Based on this lemma, the following section will show our main
Index Terms—LMI’s, stability, uncertain delay. results.

I. INTRODUCTION II. MAIN RESULTS


Consider a system We now summarize the main results in the following theorem.
Theorem 1: Assume that an uncertain time-invariant delay lies in
x_ (t) = Ax(t) + Bx(t 0 h); h  0: (1) [0; h], i.e., h 2 [0; h]. Then if there exist P > 0, Q > 0, V > 0,
and W such that
There has been a large amount of literature during the last couple of
(1; 1) 0W T B AT B T V (1; 4)
decades dedicated to systems called Smith predictor theories [1]–[4],
in which h is known. For a system with unknown delays, however,
0BT W 0Q BT BT V 0 <0
V BA V BB 0V 0 (3)
the concepts of robust control theories have been recently introduced.
If an unknown delay term is constant but unlimited, i.e., h 2 [0; 1),
(1; 4)T 0 0 0V
researchers have provided several delay-independent stability criteria where
[5]–[7]. If an unknown delay term is constant and bounded, delay-
(1; 1) =1 (A + B )T P + P (A + B ) + W T B + B T W + Q
(1; 4) =1 h(W T + P )
dependent stability criteria in [8]–[12] improve stability margins
compared to delay-independent criteria.
The underlying idea to providing delay-dependent criteria has been
to use the following inequality: given a; b 2 Rn
then the system is asymptotically stable. In this case, the Lyapunov
function can be constructed as
02aT b  1 =1 X>
inf0faT Xa + bT X 01 bg V (x(t 0 ); 2 [0; h]) = V1 + V2 + V3 (4)

where aT indicates the transpose of a. In this case, the upper bound where
of (02aT b) 1 xT (t)Px(t)
V1 =
aT Xa + bT X 01 b 1 0 t x_ T ( )B T XBx_ ( ) d d
V2 =
0h t+
t
is always greater than or equal to zero. Therefore, if it holds that 1
V3 = xT ( )Qx( ) d :
(02aT b) < 0 t0h
Proof: We first note that V (x(t 0 ); 2 [0; h]) is radially
the upper bound 1 is not a good estimate.
unbounded with respect to x(t). Now consider its derivative
To improve the upper bound of (02aT b), we shall introduce
another free matrix M so that V_ = V_ 1 + V_ 2 + V_ 3 :
inf0f(a + Mb)T X (a + Mb) + bT X 01 bg:
02(a + Mb)T b  X> Since it holds that
t
Therefore, we finally have the relation x(t) 0 x(t 0 h)  x_ () d (5)
t0h
02aT b  2 =1 X>inf0; M f(a + Mb)T X (a + Mb) + bT X 01b (1) can be written as
t
+ 2bT Mbg x_ (t) = (A + B )x(t) 0 B x_ ( ) d (6)
t0h
where clearly 2  1 . Modifying this result yields the following
lemma.
and thus the derivative of V1 satisfies the relation
t
V_ 1 = 2xT (t)P (A + B )x(t) 0 2xT (t)PB x_ ( ) d :
t0h
Manuscript received March 14, 1997; revised July 28, 1997. This work was
supported by the Electrical Engineering and Science Research Institute and
monitored by Korea Electric Power Research Institute under Grant 97-081. Defining a(1) and b(1) in (2) as, for all 2 [t 0 h; t]
The author is with the Department of Electronic and Electrical Engineering,
Pohang University of Science and Technology, Pohang, Kyung-book, 790-784, a( ) =1 BAx( ) + BBx( 0 h)
Korea (e-mail: ppg@vision.postech.ac.kr).
Publisher Item Identifier S 0018-9286(99)01290-8.
1 P x(t)
b( ) =

0018–9286/99$10.00  1999 IEEE


IEEE TRANSACTIONS ON AUTOMATIC CONTROL, VOL. 44, NO. 4, APRIL 1999 877

TABLE I IV. CONCLUDING REMARKS


COMPARISONS AMONG VARIOUS DELAY-DEPENDENT CRITERIA By providing a new upper bound of a vector cross-product with
cross- as well as inner-products, rather than only with inner-products,
this paper obtained a new stability criterion for systems with time-
invariant uncertain delays. An example showed that this criterion
performs much better than several existing criteria.

REFERENCES

[1] O. J. Smith, “A controller to overcome dead time,” ISA J., vol. 6, pp.
28–33, Feb. 1959.
[2] Z. Q. Wang and S. Skogestad, “Robust control of time-delay systems
using the Smith predictor,” Int. J. Contr., vol. 57, no. 6, pp. 1405–1420,
and using Lemma 1 will supply 1993.

V_ 1  xT (t)f(A + B )T P + P (A + B )
[3] K. J. Aström, C. C. Hang, and B. C. Lin, “A new Smith predictor
for controlling a process with an integrator and long dead-time,” IEEE
T 01
+ hP (M X + I )X (XM + I )P gx(t) Trans. Automat. Contr., vol. 39, pp. 343–345, Feb. 1994.
t [4] M. R. Matauŝek and A. D. Micić, “A modified Smith predictor for
T
+ 2x (t)PM
T XB x_ ( ) d controlling a process with an integrator and long dead-time,” IEEE
t0h Trans. Automat. Contr., vol. 41, pp. 1199–1203, Aug. 1996.
t H
[5] J. H. Lee, S. W. Kim, and W. H. Kwon, “Memoryless 1 controllers
+ x_ T ( )B T XBx_ ( ) d : for state delayed systems,” IEEE Trans. Automat. Contr., vol. 39, pp.
t0h 159–162, Jan. 1994.
[6] J. Chen and H. A. Lathchman, “Frequency sweeping tests for stability
Since V_ 2 and V_ 3 yield the relation independent of delay,” IEEE Trans. Automat. Contr., vol. 40, pp.
1640–1645, Sept. 1995.
t
V_ 2 = hx_ T (t)B T XBx_ (t) 0 x_ T ( )B T XBx_ ( ) d [7] J. Chen, D. Xu, and B. Shafai, “On sufficient conditions for stability
t0h
independent of delay,” IEEE Trans. Automat. Contr., vol. 40, pp.
1675–1680, Sept. 1995.
V_ 3 = xT (t)Qx(t) 0 xT (t 0 h)Qx(t 0 h) [8] S. S. Wang, B. S. Chen, and T. P. Lin, “Robust stability of uncertain
1 1
choosing W = XMP and V = hX will yield
time-delay systems,” Int. J. Contr., vol. 46, pp. 963–976, 1987.
[9] S. I. Niculescu, C. E. de Souza, J. M. Dion, and L. Dugard, “Robust
stability and stabilization of uncertain linear systems with state delay:
V_ = V_ + V_ + V_
1 2 3 Single delay case (i),” in Proc. IFAC Symp. Robust Control Design, Rio
 xT (t)f(A + B)T P + P (A + B) de Janeiro, Brazil, Sept. 1994, pp. 469–474.
[10] J. H. Su, “Further results on the robust stability of linear systems with
2 T
+ h (W + P )V
01 (W + P )gx(t) a single time delay,” Syst. Contr. Lett., vol. 23, pp. 375–379, 1994.
t [11] S. I. Niculescu, A. T. Neto, J. M. Dion, and L. Dugard, “Delay-
+ 2x (t)W
T TB x_ ( ) d T
+ x_ (t)B
T V B x_ (t) dependent stability of linear systems with delayed state: An LMI
t0h approach,” in Proc. 34th Conf. Decision and Control, New Orleans, LA,
T
+ x (t)Qx(t) 0 xT (t 0 h)Qx(t 0 h) Dec. 1995, vol. 2, pp. 1495–1496.
[12] X. Li and C. E. de Souza, “Criteria for robust stability of uncertain linear
T
x(t) X11 X12 x(t) systems with time-varying state delays,” in IFAC: 13th World Congr.,
=
x(t 0 h) T X
X12 22 x(t 0 h)
San Francisco, CA, 1996, vol. 1, pp. 137–142.

where
1 (A + B )T P + P (A + B ) + h2 (W T + P )V 01 (W + P )
X11 =
T T T
+ A B V BA + Q + W B + B W
T
1 0W T B + AT B T V BB
X12 =
1 0Q + B T B T V BB:
X22 =
With some efforts, we can show that (3) guarantees the negativeness
of V_ whenever neither x(t) nor x(t 0 h) is zero, which immediately
implies the asymptotic stability of the system.

III. EXAMPLE
We shall illustrate the results by using an example. Consider

A= 02:0 0:0 01:0 0:0


0:0 00:9 ; B= 01:0 01:0 :
We summarize the maximal h according to each method in the
literature in Table I. Any delay-independent stability criterion fails
to verify that the system is asymptotically stable. When a delay term
is time-varying, the largest bound obtained in the literature is 0.7218
via Li and de Souza’s result [12], whereas when it is time-invariant,
one can find the larger bound 0.9999 via a result of Niculescu et al.
[11]. Based on our result, we obtain 4.3588.

You might also like