Distributed auto disturbances rejection resilient control of permanent magnetic maglev trains based on the optimized deep deterministic policy gradient algorithm

Abstract

Due to time-varying external disturbances and uncertain system models, distributed cooperative controllers with poor adaptability are unable to meet the cooperative control requirements of multiple permanent magnetic maglev trains in virtual coupling mode. In this work, a new effective distributed auto disturbance rejection resilient controller based on the optimized deep deterministic policy gradient algorithm (DDPG) is proposed. The DDPG algorithm is used to improve the adaptability of the controller against the time-varying disturbances. An adaptive particle swarm optimization method (APSO) is also proposed to optimize the hyperparameters of DDPG in the search space. The simulation results show that, compared to the particle swarm optimization (PSO)-actor-critic (AC), PSO-policy gradient (PG), and PSO-DDPG algorithms, the proposed APSO-DDPG algorithm performs better during training and verification. The proposed method achieves adaptive online adjustment of the controller parameters effectively and greatly improves the stability of cooperative control.