Extended robust global exponential stability for uncertain switched memristor-based neural networks with time-varying delays
Extended robust global exponential stability for uncertain switched memristor-based neural networks with time-varying delays
复制标题
DOI：
10.1016/j.amc.2017.12.032
复制
发表时间：
2018-05
期刊：
影响因子：
--
通讯作者：
Xiaoqing Li;Kun She;S. Zhong;Kaibo Shi;W. Kang;Jun Cheng;Yongbin Yu
中科院分区：
文献类型：
--
作者：
Xiaoqing Li;Kun She;S. Zhong;Kaibo Shi;W. Kang;Jun Cheng;Yongbin Yu
研究了具有时变时滞和开关参数的不确定忆阻神经网络在不稳定子系统下的全局指数稳定性问题。与现有文献不同的是，本文将离散时滞不确定切换神经网络建模为具有不确定时变参数的切换神经网络。基于多重Lyapunov-Krasovskii泛函（MLF）方法、平均驻留时间（ADT）技术和依赖于模式的平均驻留时间（MDADT）方法，推导了基于LMI的稳定性准则，用于设计切换信号，保证所考虑的不确定切换神经网络的指数稳定性.通过研究各子系统的模态依赖性，将其分为稳定子系统和不稳定子系统。与现有的只考虑所有子系统稳定性的神经网络模型相比，同时考虑稳定子系统和不稳定子系统的神经网络模型具有更广泛的通用性和适用性，从而得到更少的保守性准则。为了便于使用Matlab的LMI工具箱，将所提出的充分条件简化为线性矩阵不等式的形式。最后，两个数值例子被用来证明所提出的理论结果的有效性和适用性。
This paper is concerned with the problem of global exponential stability for uncertain memristive-based neural networks (UMNNs) with time-varying delays and switching parameters subject to unstable subsystems. Different from most of the existing papers, the considered uncertain switched MNNs with discrete-delays are modeled as switched neural networks (SNNs) with uncertain time-varying parameters. Based on multiple Lyapunov–Krasovskii functional (MLF) approach, average dwell time (ADT) technique and mode-dependent average dwell time (MDADT) method, some LMIs-based stability criteria are derived to design the switching signal and guarantee the exponential stability of the considered uncertain switched neural networks. By exploring the mode-dependent property of each subsystem, all the subsystems are categorized into stable and unstable ones. The concerned SNNs with both stable and unstable subsystems are more general and applicable than the existing models of SNNs only view all subsystems being stable, thus getting less conservatism criteria. The proposed sufficient conditions can be simplified into the forms of LMIs for conveniently using Matlab LMI toolbox. Finally, two numerical examples are exploited to demonstrate the effectiveness and applicability of the proposed theoretical results.