
Nonfragile mixed H
∞
/l
2
−l
∞
state estimation for repeated scalar nonlinear systems 643
tion for MJSs with repeated scalar nonlinearities and
redundant channels. Threefold major contributions are
highlighted as following: (1.) A novel state estimation
technique, which integrates the H
∞
criterion with the
l
2
−l
∞
specification, is considered for the first time in
terms of MJSs with nonlinearity; (2.) by constructing
a mode-dependent Lyapunov function and combining
with some positive diagonal dominant mode-dependent
matrices, a sufficient condition for keeping the stochas-
tic stability of the error system (
¯
) is given out, which
confirms that the system is with a prescribed mixed
H
∞
/l
2
−l
∞
performance; (3.) in contrast to the previ-
ous methods in [11,20], this work addresses a critical
unified framework equipped with two parallel chan-
nels with hope to strive against the effect of randomly
occurring packet dropouts.
The remainder of this dissertation is structured
as follows. We give the system descriptions and the
address problem in Sect. 2. What’s more, the mixed
H
∞
/l
2
− l
∞
performance conditions are considered
first in the designing of state estimation for Markov
jump nonlinear systems. In Sect. 3, the parameter of
the state estimation has been figured out. Simultane-
ously, the designed state estimator guarantees that the
system is stochastically stable. Section 4 describes a
simulation performed to demonstrate the merits and the
validity of approaches. Ultimately, this paper draws the
conclusion in Sect. 5.
Notation The superscript “T ” signifies the trans-
pose and “∗” stands for symmetric entry. R
n
is the
n-dimensional Euclidean space. Z
+
stands for the
positive real number. I and 0 represent the iden-
tity matrix and zero matrix with suitable dimension,
respectively. E{·}denotes the expectation operator with
respect to some probability measure P ; l
2
[0, ∞) is the
space of square summable infinite vector sequences
over [0, ∞); for matrices A, B,diag{A, B} denote
A 0
0 B
; W ≥ 0(W > 0)⇔ W is positive semidefi-
nite (positive definite).
2 Problem formulation
Let us consider the MJSs with repeated scalar nonlin-
earities ()
x(k + 1) = A(δ
k
)x(k) + B(δ
k
)g(x(k))
+C(δ
k
)ω(k), (1)
z(k) = F(δ
k
)x(k), (2)
where x(k) ∈ R
a
is the system state vector, z(k) ∈ R
l
is the signal to be estimated, and ω(k) ∈ R
b
is the dis-
turbance input which belongs to l
2
[0, ∞). g(x(k))
[g
1
(x(k)) g
2
(x(k)) ··· g
a
(x(k)) ]
T
∈ R
a
denotes
the nonlinear function of the system. Let us sup-
pose that the nonlinear function g(•) is continu-
ous and bounded, conforming with the following
constraint
|g(u) + g(v)|≤|u + v|, ∀u,v∈ R. (3)
The matrices of system are expressed by A(δ
k
), B(δ
k
),
C(δ
k
), and F(δ
k
), which are known real constant pro-
vided with proper dimensions.
δ
k
(k ∈ Z
+
) refers to a discrete-time homogeneous
Markov chain, which dominates the random mode
switching and takes values based on a finite state s et
K ={1, 2,...,η}, with stationary probability:
π
ij
Pr{δ
k+1
= j | δ
k
= i}≥0, ∀i, j ∈ K, k ∈ Z
+
,
and
η
j=1
π
ij
= 1, i ∈ K.
Remark 1 Evidently, the nonlinear function g(•) is the
class of odd (when u is equal to v) as well as 1-Lipschitz
(when u is equal to −u). Hence, this goes to show that
it applies to some practical nonlinearities: (1) the semi-
linear function (i.e., the standard saturation sat(s) := s
if |s|≤1 and sat(s) := sgn(s) if |s|≤1); (2) the
hyperbolic tangent function tanh, which is extensively
employed to be activation function related to neural net-
works; (3) one of the most commonly used functions:
the sine function, etc.
Remark 2 When it comes to the communication net-
works, there are a mass of existing methodologies con-
figured by a single channel [23]. Virtually, when the
primary channel fails to deliver message, the worst
case of data loss will occur if merely considering sin-
gle channel. At this point, if one more channel is added
to resume the transfer, devoting to benefits of estima-
tion performance, which is far greater than necessary
in actual movement. As has been noted, the framework
presented in this paper allows that the data transmits
over any one of the channels, ensuring that the data
will still get through even if one of the channels is i n
123