Reinforcement learning-based model predictive control method for high-dynamic electromagnetic aircraft recovery platform

This study aims to address the insufficient adaptability of traditional tracking control algorithms in real-time trajectory tracking tasks for fixed-wing UAV recovery with high dynamics, strong external interference, and multivariable coupling effects. It first establishes models of the aircraft and the linear motor, designs the DDPG algorithm and training strategies for the intelligent RL Agent, and proposes a reinforcement learning-based model predictive control(RL-MPC) framework. The control effects of RL-MPC are compared with traditional algorithms. Results show that the intelligent RL Agent adjusts controller parameters in real time through interaction with the controlled model and the environment, realizing complex trajectory tracking of the aircraft during the recovery phase and adaptive adjustment of the controller for varying recovery circumstances, significantly enhancing the performance and robustness of the original algorithms. The proposed algorithm realizes high-performance tracking control for complex trajectory tracking systems, offering a feasible and efficient method for electromagnetic recovery of future fixed-wing UAVs.