Neural-Network-Based Set-Membership Filtering Under WTOD Protocols via a Novel Event-Triggered Compensation Mechanism

This article investigates the neural-network-based (NN-based) set-membership filtering issue for nonlinear systems. In order to lighten the network transmission burden and avoid data collisions, the weighted try-once-discard (WTOD) protocol is employed to regulate the signal transmission process, which provides higher transmission priority to the most needed data. Considering the data discarding problem of the WTOD protocol, a novel event-triggered compensation mechanism is proposed to compensate the measurement output processed by the WTOD protocol, thereby improving the filtering performance. Next, considering the nonlinear dynamics of the system and the unknown-but-bounded (UBB) noise interference, an NN-based set-membership filter is designed to solve the state estimation problem. In a unified set-membership framework, an neural-network (NN) weight adaptive tuning law and a state estimation algorithm are designed. Sufficient conditions are derived for the existence of the adaptive NN parameters and the NN-based set-membership filter, and two optimization problems are put forward to seek the optimal NN parameters and filtering parameters that make the filter performance optimal. Finally, illustrative examples demonstrate the effectiveness of the proposed compensation mechanism and filtering algorithm.