Design of Model Tracking Control Law Based on DDPG Algorithm

A design method based on deep reinforcement learning is proposed for the problem of designing a model tracking control law with anti-disturbance capability. Firstly, the basic control structure based on the model tracking method is determined. Then an open source high fidelity arithmetic aircraft dynamics model is selected to carry out the control law design through simulation. The intelligent body model is built by deep deterministic policy gradient algorithm (DDPG), and the state quantity, control quantity and reward are set according to the target model tracking control problem, and the intelligent body is trained. The trained intelligent body is simulated and tested to verify its model tracking control capability with high anti-disturbance capability. By comparing with the control effect of traditional PID algorithm under perturbation, it is further demonstrated that the intelligent body trained based on deep deterministic policy gradient algorithm for flight control can achieve the model tracking control target with good anti-disturbance capability at the same time.