Non-fragile saturation control of nonlinear positive Markov jump systems 1497
space, and R
n×m
the set of n ×m matrices, respectively.
Let N
+
be the set of positive integers. For m ∈ N
+
,
denote by m
:= {1, 2,...,m}.Let(Φ, Ξ, P) be a
probability space, where Φ is the sample space, Ξ is
the algebra of subsets of the sample space, and P is
the probability measure. The inequality A 0 ( 0)
means that all elements of matrix A are nonnega-
tive (non-positive). Thus, A B (A B) means
that A − B 0 ( A − B 0). The 1-norm of a
vector x(t ) = (x
1
(t), x
2
(t),...,x
n
(t))
T
is defined
as x(t)
1
=
n
i=1
|x
i
(t)|. Given a f unction w(t) :
[0, ∞) → R
r
, its L
1
-norm is defined as w(t)
L
1
=
∞
0
w(t)
1
dt .Let1
m
= (1,...,1)
T
∈ R
m
and 1
(ı)
m
=
(0,...,0,
ı−1
1, 0,...,0
m−ı
)
T
. The symbol E{·} stands for
the mathematical expectation. The identity matrix in
R
n×n
is I
n
. A matrix A is said to be Metzler if all its
off-diagonal elements are nonnegative.
2 Problem statement
For a probability space (Φ, Ξ, P), we consider a class
of NMJSs:
⎧
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎩
˙x(t) = A
r(t)
f (x(t )) + A
d,r(t )
f (x(t − τ(t)))
+ B
r(t)
sat(u(t)) + E
r(t)
w(t),
y(t) = C
r(t)
g(x(t )) + F
r(t)
w(t),
x(θ) = ψ(θ), θ ∈[−
τ,0],
(1)
where x(t ) ∈ R
n
, u(t) ∈ R
m
, and y(t) ∈ R
s
are the
system state, the control input, and the output, respec-
tively. The external disturbance input w(t) ∈ R
r
sat-
isfies E{
∞
0
w(t)
1
dt} < w, where w>0isa
given bound. The function sat(·) represents the vector-
valued standard saturation term with sat(u(t)) =
(sat(u
1
(t)),...,sat(u
m
(t)))
T
, where sat(u
p
(t)) =
sign(u
p
(t)) min
p∈m
{1, |u
p
(t)|} for p ∈ m.Let{r(t), t ≥
0} be a homogeneous Markov chain in the finite set
S ={1, 2,...,N }, N ∈ N
+
. The transition probabili-
ties of {r(t), t ≥ 0} are given by
P{r(t + Δ) = j|r(t) = i}=
λ
ij
Δ + o(Δ), i = j,
1 + λ
ii
Δ + o(Δ), i = j,
where Δ ≥ 0, o(Δ)/Δ = 0asΔ → 0, λ
ij
≥ 0for
i = j is the transition rate from mode i at time t to
mode j at time t + Δ, and λ
ii
=−
N
j=1,i= j
λ
ij
for
all i, j ∈ S. The time-varying delay is τ(t) satisfying
0 ≤ τ
≤ τ(t) ≤ τ, ˙τ(t) ≤ d ≤ 1, (2)
where τ
, τ, and d are given positive constants. The
initial condition ψ(θ) is defined on interval [−
τ,0].
For r (t) = i ∈ S, the system matrices are denoted as
A
i
, A
di
, B
i
, E
i
, C
i
, and F
i
, which are constant matri-
ces with appropriate dimensions. For each mode i ∈ S,
it is assumed that A
i
is Metzler, A
di
0, B
i
0, E
i
0, C
i
0, and F
i
0. The following assumption is
imposed on the nonlinear terms f (x(t )) and g(x(t)).
Assumption 1 The nonlinear functions f (x(t)) and
g(x(t )) satisfy the following conditions:
h
1
x
2
i
(t) ≤ f
i
(x
i
(t))x
i
(t) ≤ h
2
x
2
i
(t), (3a)
h
3
x
2
i
(t) ≤ g
i
(x
i
(t))x
i
(t) ≤ h
4
x
2
i
(t), (3b)
where i ∈ n
, 0 < h
1
≤ h
2
, 0 < h
3
≤ h
4
, f
i
(0) = 0,
and g
i
(0) = 0.
Remark 1 The constraint conditions in (3) define a
sector-bounded nonlinearity. The constants h
1
, h
3
and
h
2
, h
4
are the weighted coefficients of lower and
upper bounds, respectively. Recently, a class of pos-
itive switched systems with sector-bounded nonlinear-
ities was considered in [27 ] and [28]. It is worth point-
ing out that the sector-bounded nonlinearity widely
exists in practical systems such as neural network sys-
tems, variable structure systems, and complex circuit
systems. Assume that x
i
(t)>0 ∀t, then f
i
(x
i
(t))
and g
i
(x
i
(t)) will cover the first and third orthants as
h
1
→ 0, h
2
→∞, and h
3
→ 0, h
4
→∞. This means
that the nonlinearities in (3) can describe a large class
of nonlinear functions.
Definition 1 ([27]) System (1) is said to be positive
if for each mode r (t) the initial condition ψ(θ) 0
implies that x(t) 0 and y(t) 0 hold for all inputs
u(t) 0 and w(t) 0 at all time.
For an MJS, it is called PMJS if all submodes of
MJS are positive.
Lemma 1 ([27]) Under Assumption 1,system(1) is
positive if and only if A
i
is a Metzler matrix, A
di
0, B
i
0, E
i
0, C
i
0, and F
i
0 for any i ∈ S.
Noting the assumption on the system matrices of sys-
tem (1), we have that system (1) is a PMJS by Lemma 1.
Lemma 2 ([12]) Amatrix A ∈ R
n×n
is Metzler if there
exists a positive constant β such that A + β I
n
0.
Definition 2 ([23]) System (1) is said to be stochasti-
cally stable with L
1
-gain performance if the following
statements hold:
123