Reinforcement Learning for UAV Autonomous Tracking Random Moving Target

A novel unmanned aerial vehicles (UAVs) autonomous target tracking control method based on Reinforcement Learning (RL) is presented in this article. The policy network controller trained by the RL algorithm replaces the trajectory planning and UAV outer loop controller in the traditional UAV target tracking architecture, and this controller has the advantages of high robustness and less computation. Besides, a model optimization is applied to turn the issue of UAV tracking random moving target to UAV tracking stationary target. In this way, the UAV target tracking question will be more comfortable to apply the RL algorithm to train, and the training result also has strong performance on UAV tracking a random moving target.