Nonfragile state estimation for memristor based competitive neural networks subject to sensor saturation and stochastic cyber attacks

This article presents an event-based framework for nonfragile state estimation in memristor-driven competitive neural networks, which are particularly vulnerable to stochastic cyber attacks and sensor saturation. Often, unnecessary signals burden the communication bandwidth when transmitted to the estimator. To address this, the article proposes a dynamic event-triggered mechanism that carefully transmits relevant data. Unlike a static event-trigger, this approach optimizes data transmission. By constructing proper Lyapunov-Krasovskii functional the authors establish delay-dependent criteria expressed as linear matrix inequalities to ensure asymptotic stability of the system. To showcase the advantages of the proposed criteria and confirm their practical relevance, two well-established benchmark problems including a quadruple-tank control system are examined.