Application of Deep Reinforcement Learning in Reconfiguration Control of Aircraft Anti-Skid Braking System

The aircraft anti-skid braking system (AABS) plays an important role in taking off, taxiing, and safe landing. In addition to the disturbances from complex runway environment, potential component faults, such as actuators can also reduce safety reliability of AABS. To meet increasing performance requirements AABS under fault disturbance conditions, a novel reconfiguration controller based on linear active rejection control combined with deep reinforcement learning was proposed this paper. treated external perturbations, measurement noise total disturbances. twin delayed deterministic policy gradient algorithm (TD3) introduced realize parameter self-adjustments both extended state observer error feedback law. action space, reward function, network structure for training were properly designed, so that could be estimated compensated more accurately. simulation results validated environmental adaptability robustness controller.