Abstract
In this paper, in order to solve the 3D path planning problem of the unmanned aerial vehicle (UAV) in unknown environments with obstacles, an improved Double-deep Q network (DQN) algorithm is proposed. In order to solve the problem of sparse rewards in the process of deep reinforcement learning training, a mechanism of energy efficiency reward is designed, and then the 3D path of the VAV is shortened and smoothed. In addition, a mechanism of priority experience replay based on the queue is proposed to improve the sampling probability of relatively good experience, so that the training time is reduced and the success rate is improved. Simulation results show that, compared with the traditional method, the algorithm proposed in this paper can shorten the average number of steps required to reach the maximum score and improve the training success rate.
Funding Information
  • National Natural Science Foundation of China (61903198)