As
X
k1
i¼kdðkþ1Þþ1
x
T
ðiÞQxðiÞ¼
X
k1
i¼kd
min
þ1
x
T
ðiÞQxðiÞ
þ
X
kd
min
i¼kdðkþ1Þþ1
x
T
ðiÞQxðiÞ
X
k1
i¼kdðkÞþ1
x
T
ðiÞQxðiÞ
þ
X
kd
min
i¼kd
max
þ1
x
T
ðiÞQxðiÞð7Þ
we have
DV
2
x
T
ðkÞQxðkÞx
T
ðk dðkÞÞQxðk dðkÞÞ
þ
X
kd
min
i¼kd
max
þ1
x
T
ðiÞQxðiÞð8Þ
Note that
DV
3
¼
X
d
min
þ1
j¼d
max
þ2
½x
T
ðkÞQxðkÞx
T
ðk þ j 1ÞQxð k þ j 1Þ
¼ðd
max
d
min
Þx
T
ðkÞQxðkÞ
X
kd
min
i¼kd
max
þ1
x
T
ðiÞQxðiÞ
ð9Þ
In addition
xðkÞxðk dðkÞÞ
h
ðkÞ¼0 ð10Þ
Therefore for any appropriately dimensioned matrices R,
S and T, we have the following equation
2½x
T
ðkÞR þ x
T
ðk dðkÞÞS þ
h
ðkÞ
T
T
½xðkÞxðk dðkÞÞ
h
ðkÞ ¼ 0 ð11Þ
It follows by adding (5), inequality (8), and (9) and (11) that
DV ¼ DV
1
þ DV
2
þ DV
3
x
T
ðkÞ½ðA þ A
1
Þ
T
PðA þ A
1
ÞPxðkÞ
2x
T
ðkÞðA þ A
1
Þ
T
PA
1
h
ðkÞ
þ
h
ðkÞ
T
A
T
1
PA
1
h
ðkÞþx
T
ðkÞQxðkÞ
x
T
ðk dðkÞÞQxðk dðkÞÞ
þ
X
kd
min
i¼kd
max
þ1
x
T
ðiÞQxðiÞþðd
max
d
min
Þx
T
ðkÞQxðkÞ
X
kd
min
i¼kd
max
þ1
x
T
ðiÞQxðiÞ
þ 2x
T
ðkÞRxðkÞþx
T
ðkÞ½ 2R þ 2S
T
xðk dðkÞÞ
þ x
T
ðkÞ½ 2 R þ 2T
T
h
ðkÞ
2x
T
ðk dðkÞÞ Sx ðk dðkÞÞ
þ x
T
ðk dðkÞÞ½2S 2T
T
h
ðkÞ
2
h
ðkÞ
T
T
h
ðkÞ
¼ x
T
ðkÞ½ðA þ A
1
Þ
T
PðA þ A
1
Þ
P þðd
max
d
min
þ 1ÞQ þ 2RxðkÞ
þ x
T
ðkÞ½2R þ 2S
T
xðk dðkÞÞ þ x
T
ðkÞ
½2ðA þ A
1
Þ
T
PA
1
2R þ 2T
T
h
ðkÞ
þ x
T
ðk dðkÞÞ½Q 2Sxðk dðkÞÞ
þ x
T
ðk dðkÞÞ½2S 2T
T
h
ðkÞ
þ
h
ðkÞ
T
½A
T
1
PA
1
2T
h
ðkÞ
¼
j
ðkÞ
T
Vj
ðkÞð12Þ
where we define
j
(k)
T
¼ [x
T
(k), x
T
(k 2 d(k)),
h
(k)
T
] and
V
¼
V
11
R þ S
T
ðA þ A
1
Þ
T
PA
1
R þ T
T
Q S S
T
S T
T
A
T
1
PA
1
T T
T
2
6
4
3
7
5
ð13Þ
where V
11
¼ (A þ A
1
)
T
P(A þ A
1
) 2 P þ (d
max
2 d
min
þ 1)
Q þ R þ R
T
.
From this, it follows that the inequality
V
, 0 guarantees
that DV , 0 for all non-zero
j
(k). Hence,
V
, 0 guarantees
that the unforced system given in (2) is asymptotically
stable for all time-varying delay d(k) satisfying d
min
d(k) d
max
. By Schur complement,
V
, 0 is equivalent
to LMI (3). This completes the proof of Theorem 1. A
Remark 1: Note that Theorem 1 only depends on the differ-
ence between the maximum and minimum delay bounds,
that is it only depends on the delay interval and not on the
actual delays themselves. Thus, Theorem 1 is not a delay-
dependent sufficient condition for asymptotic stability of
the system given in (2).
Theorem 1 presents a stability result that depends on the
difference between the maximum and minimum delay
bounds. Thus, for the constant delay case, as minimum
and maximum bounds in Assumption 1 are identical, d
min
¼
d
max
¼ d: Theorem 1 does not depend on the delay, which
gives the following.
Corollary 1: The unforced system given in (2) with constant
delay d(k) ¼ d is asymptotically stable if there exist n n
matrices P . 0, Q . 0, R, S and T satisfying the following
LMI
P þ Q þ R þ R
T
R þ S
T
Q S S
T
2
6
6
6
4
R þ T
T
ðA þ A
1
Þ
T
P
S T
T
0
T T
T
A
T
1
P
P
3
7
7
7
5
, 0 ð14Þ
IEE P roc.-Control Theory Appl., Vol. 153, No. 6, November 2006 691
Authorized licensed use limited to: Central South University. Downloaded on December 6, 2008 at 08:32 from IEEE Xplore. Restrictions apply.