Fixed-time optimal time-varying formation control for unmanned surface vehicle systems based on reinforcement learning.

Abstract

This article proposes the distributed fixed-time optimal time-varying formation control (TVFC) strategy based on reinforcement learning (RL) for unmanned surface vehicle systems (USVSs) with partially unmeasurable states and unknown dynamics. The fixed-time adaptive neural network state observer (FANNSO) is introduced for reconstructing unknown dynamics and unmeasurable states of the system. On this basis, a distributed optimization performance index function containing exponential terms is proposed, and a distributed fixed-time optimal TVFC strategy is developed by combining the actor-critic structure. This strategy achieves the dual objectives of formation control and cost optimization by adaptively adjusting the controller through the RL algorithm. Theoretical analyses show that the proposed control strategy can make the error signals bounded within a fixed time. Simulation results demonstrate the effectiveness and superiority of the method.