Non‐fragile set‐membership estimation for sensor‐saturated memristive neural networks via weighted try‐once‐discard protocol

Abstract

This study is concerned with the non-fragile set-membership estimation problem for a class of delayed sensor-saturated memristive neural networks under the premise of communication protocol transmissions. The exogenous noises are unknown but bounded and the activation function satisfies the sector-bounded condition. In order to schedule the limited network resources, the node's utilisation right of the communication channel at the current instant is determined by the weighted try-once-discard protocol. Besides, the estimator gain perturbations are considered to enhance the robustness of estimation method. The major focus of the study is to a design a non-fragile estimator such that, for all measurement saturation, mixed time-delays and estimator gain perturbation, the estimation error exists in the ellipsoid by providing a sufficient criterion and the optimal semi-axes length of the ellipsoid is found by solving a convex optimisation problem. Finally, a simulation example shows that the proposed non-fragile estimation strategy is effective.