Deep-Q-Learning-Based Proportional-Integral Event-Triggered Fuzzy Output Control of Variable Speed AGVs via Proportional-Integral Observer

This paper focuses on the T-S fuzzy output path tracking control issue for variable speed autonomous ground vehicles (VSAGVs) by using state observer and event-triggered mechanism. The model of VSAGVs with nonlinear dynamics and parameter uncertainty is represented by the uncertain T-S fuzzy system. To enhance the estimation performance of the unknown system state, a fuzzy proportional- integral observer containing an extra integral term of estimation error is designed. To reduce the network communication burden, a novel deep-Q-learning-based proportional- integral event-triggered mechanism by adding the integration of the normal triggering rule and introducing an optimization strategy of the triggering threshold. Compared with the classical normal and dynamic event-triggered mechanisms, a larger inter-event time can be generated by the proposed proportional- integral event-triggered mechanism. In addition, the deep-Q-learning-based optimization strategy is able to dynamically regulate the triggering rate according to the historical triggering performance and control performance. Some sufficient conditions are derived to solve the fuzzy observer and controller gains and the triggering matrix such that the VSAGV is asymptotically stable with given$H_{\infty }$level. Finally, the superiorities of the addressed approach are confirmed through some simulation results.