Extended dissipative resilient estimator design for discrete-time switched neural networks with unreliable links

Abstract

This paper addresses the problem of extended dissipativity-based resilient state estimation for discrete-time switched neural networks in the presence of unreliable links. To overcome the difficulty of random fluctuation and communication constraints, a novel and highly flexible form of estimator is designed by considering random uncertainty, signal quantization and data packet dropout phenomena simultaneously. The considered mode-dependent average dwell time (MDADT) switching law is shown to be more general than the traditional ADT for permitting each subsystem to has its own average dwell time. By constructing the MD Lyapunov–Krasovaskii functional (LKF), sufficient conditions are given to ensure the exponential mean-square stability and extended stochastic dissipativity of the augmented system and an explicit expression of the desired estimator is presented. Finally, a example with two cases is provide to illustrate the feasibility and effectiveness of the developed theoretical results.