Observer-Based Event-Triggered Optimal Control for Nonlinear Multiagent Systems With Input Delay via Reinforcement Learning Strategy

Ever since the reinforcement learning method has been proposed, the optimal control problem for multiagent systems has been intensively explored in light of the energy consumption growth. However, most of the consequences only focused on ideal communication cases without input delay. Distinguished from the existing optimal control mechanisms, the paper concentrates on the event-triggered optimal tracking control problem for nonlinear multi-agent systems with unmeasurable states and input delay. A state observer taking into account the universal approximation capability of neural networks is proposed to tackle the unmeasurable state variables which are often encountered in practical engineering scenarios. Subsequently, the optimal backstepping control mechanism is developed according to the reinforcement learning method, where the actor-critic networks are utilized to obtain the modified control parameters. Finally, an event-triggered strategy, which relies both on the latest sampled state and a non-negative threshold, is presented to relieve the communication frequency of the control signals with guaranteed performance requirements. The illustrated simulations are presented to confirm the feasibility of the proposed optimal tracking control algorithms.