A Review on Event-Triggered State Estimation for Time-Delay Neural Networks

Abstract

In practical engineering applications, time delays are inherent in neural networks due to signal transmission lags, hardware response limits and network communication constraints, which may induce oscillation, chaos and even system instability. Meanwhile, limited measurable outputs and external noise make it impossible to obtain full neuron states directly, so state estimation has become an indispensable technique. However, traditional time-triggered transmission leads to massive redundant data and unnecessary communication consumption. Event-triggered mechanism (ETM) transmits data only when system states violate preset trigger conditions, which effectively reduces communication load while ensuring satisfactory estimation performance. This paper systematically reviews the research progress of event-triggered state estimation for time-delay neural networks. Firstly, basic concepts, system modeling and core problems are introduced. Then, key theoretical tools including Lyapunov-Krasovskii (L-K) functional, integral inequalities and stability analysis methods are summarized. Furthermore, typical event-triggered protocols, estimator design approaches and unified performance indices are elaborated. Finally, existing challenges and future research directions are discussed, aiming to provide a clear reference for researchers in related fields.