Neural‐network‐based distributed security filtering for networked switched systems

Abstract In this paper, distributed security filtering is investigated for networked switched systems under the round‐robin protocol (RRP). In order to fully take the practical situation into account, a cyber attack without additional bounded constraint is considered by introducing the neural network approximate technique. Moreover, the scattered sensor nodes are universally utilized to transmit and collect system information. The distributed filtering is proposed for the networked switched systems to estimate the system output. In order to ease data collisions, RRP is taken to schedule signal communication through the shared network. By constructing a token‐dependent and mode‐dependent Lyapunov function, sufficient conditions are given to ensure the filtering error system under cyber attack is finite‐time bounded. The filtering gains are determined by solving a set of linear matrix inequalities. Finally, the availability of the acquired methods is verified by a PWM‐driven boost converter model.