Sliding-Mode Synchronization for Nonidentical
Markovian Jump Neural Networks with Leakage
Delay and Partially Unknown Transition Probabilities
Jing Lv
School of Electronics
and Information
Nantong University
Nantong 226019, China
Xiaomei Zhang
School of Electronics
and Information
Nantong University
Nantong 226019, China
The corresponding author.
Email: zhang.xm@ntu.edu.cn
Lei Yan
School of Electronics
and Information
Nantong University
Nantong 226019, China
Yueping Zhu
School of Sciences
Nantong University
Nantong 226019, China
Abstract—This paper deals with an integral sliding mode
control design for synchronization of two nonidentical Markovian
jump neural networks with leakage delay, discrete time-varying
delays and partially unknown transition probabilities. Based on a
mode-dependent Lyapunov-Krasovskii functional combined with
Finsler’s lemma, delay-dependent conditions guaranteeing the
mean-square exponential stability of the synchronization error
dynamics in sliding mode are derived in terms of linear matrix
inequalities. Then, a sliding mode synchronization controller is
designed such that the synchronization error system’s trajectories
converge to predefined sliding surfaces in a finite time and remain
there for all subsequent times. Finally, a numerical example is
provided to illustrate the effectiveness of the proposed approach.
I. INTRODUCTION
Drive-response synchronization of chaotic neural networks
has attracted considerable attention due to its theoretical im-
portance and practical applications, such as secure communi-
cation, information processing, biological neural networks [1]-
[5]. In the literature, however, synchronization in chaotic neural
networks mainly focus on the synchronization of two chaotic
identical neural networks by using various control methods
such as adaptive control [6], impulsive control [7], sampled-
data control [8], linear feedback control ( including static
feedback control [9] and dynamic feedback control [10]), time-
delay feedback control [11], periodically intermittent control
[12], etc..
By contrast, synchronization of two non-identical chaotic
systems has not been widely studied [13]-[15]. However, it
is of great significance to study the synchronization prob-
lem of nonidentical chaotic systems because of its practical
applications, for example, the adaptive synchronization of
two different chaotic systems has been used to develop an
asymmetric image cryptosystem [16].
As is stated in [17], a neural network sometimes switch
from one mode to another and can be modeled as a Markovian
jumping system. There is a large amount of theory in the
literature on stability or synchronization of Markovian jumping
neural networks with completely known transition probabilities
[18]-[21] or partly unknown transition probabilities [22]-[24].
In [21], a sliding mode approach has been applied to master-
slave synchronization of identical time-delay systems with
Markovian jumping parameters and nonlinear uncertainties.
In this paper, we will be concerned with drive-response
synchronization of Markovian jump neural networks with
leakage delay and partially unknown transition probabilities
using integral sliding mode control. We will propose a mode-
dependent Lyapunov-Krasovski functional and employ both
Jensen’s inequality and Finsler’s lemma to derive some delay-
dependent stability criteria for synchronization error dynamics
in sliding mode. The effectiveness of the synchronization
criteria will be illustrated by a numerical example.
II. P
ROBLEM FORMULATION AND PRELIMINARIES
Let (Ω, ℱ, 𝒫) be a complete probability space with the
space of elementary events Ω, a natural filtration {ℱ
𝑡
}
𝑡≥0
and the probability measure 𝒫, and let {𝑟(𝑡),𝑡 ≥ 0} be a
right-continuous homogeneous Markov chain on the proba-
bility space (Ω, ℱ,𝑃) taking values in a finite state space
𝒮 = {1, 2, ⋅⋅⋅ ,𝑟} with generator Π=(𝜋
𝑖𝑗
)
𝑟×𝑟
given by
𝑃 {𝑟(𝑡 + △𝑡)=𝑗∣𝑟(𝑡)=𝑖} =
𝜋
𝑖𝑗
△𝑡 + 𝑜(△𝑡),𝑖∕= 𝑗,
1+𝜋
𝑖𝑖
△𝑡 + 𝑜(△𝑡),𝑖= 𝑗
where △𝑡>0 and lim
△𝑡→0
𝑜(△𝑡)
△𝑡
=0,𝜋
𝑖𝑗
≥ 0(𝑖, 𝑗 ∈𝒮,𝑗 ∕=
𝑖) is the transition rate from mode 𝑖 at time 𝑡 to mode 𝑗 at
time 𝑡 + △𝑡, and for each 𝑖 ∈𝒮, 𝜋
𝑖𝑖
= −
𝑗∕=𝑖
𝜋
𝑖𝑗
.
Consider the following Markovian jump neural networks
with leakage delay and discrete time-varying delays:
˙𝑥(𝑡)=−𝐶
1
(𝑟(𝑡))𝑥(𝑡 − 𝜎)+𝐴
1
(𝑟(𝑡))𝑓(𝑥(𝑡))
+𝐵
1
(𝑟(𝑡))𝑓(𝑥(𝑡 − 𝜏(𝑡))) + 𝐽
1
,
𝑥(𝑡)=𝜙(𝑡),𝑡 ∈ [−𝑑, 0], (1)
where 𝑥(𝑡)=[𝑥
1
(𝑡),𝑥
2
(𝑡),...,𝑥
𝑛
(𝑡)]
𝑇
∈ ℝ
𝑛
is the state
vector associated with the 𝑛 neurons. 𝜎>0 is the leakage
delay. The diagonal matrix 𝐶
1
(𝑟(𝑡)) has positive entries.
247
We41.3
2014 13th International Conference on Control, Automation, Robotics & Vision
Marina Ba
Sands, Sin
apore, 10-12th December 2014 (ICARCV 2014)
978-1-4799-5199-4/14/$31.00 ©2014 IEEE