Fuzzy weight-based secure formation control for two-order heterogeneous multi-agent systems via reinforcement learning

Combining with reinforcement learning (RL) technique and optimized backstepping strategy, this article investigates the distributed fuzzy formation control problem for two-order heterogeneous nonlinear multi-agent systems (MASs) against denial-of-service (DoS) attacks. Based on fuzzy logic system approximations, RL is utilized to achieve optimized control under an identifier- critic-actor architecture. As data transfer between agents is interrupted by DoS attacks, leader's state information cannot be used in optimized formation control design. Therefore, a novel distributed formation filter is constructed to obtain the unknown leader state. Additionally, the designed dynamically adjusted event-triggered condition can effectively reduce communication burden of control channel. As a consequence, the proposed optimized secure formation control can smoothly steer the heterogeneous MASs to implement multiple control tasks, i.e., system performance optimization, communication resource saving, followers-leader formation and DoS attacks defense. Ultimately, the validity of the proposed method is demonstrated through Lyapunov stability theory and digital simulations.