Fixed‐Time Formation Tracking of Multiple USVs With DoS Attacks: A Reinforcement Learning Control Method

This article studies the fixed‐time secure formation tracking problem of multiple unmanned surface vehicles (USVs) suffering from two types of denial‐of‐service (DoS) attacks, namely connectivity‐maintained attacks and connectivity‐broken attacks. A novel flexible fixed‐time observer is proposed to successfully estimate the guidance trajectories and mitigate the negative impacts caused by DoS attacks, such that a hierarchical algorithm with a fixed‐time reinforcement learning (RL) controller is advanced to achieve fixed‐time formation tracking under input saturation. Note that the actor neural networks (ANNs) are utilized in the algorithm to generate control strategies, and the critic neural networks (CNNs) are employed to assess the execution cost and optimize control inputs. Based on Lyapunov stability theory, sufficient conditions for system convergence are derived. Finally, the effectiveness of the theoretical calculation results is verified through simulation experiments.