Unmanned Aerial Vehicles Path Planning Based on Deep Reinforcement Learning

Guoqiu Wang, Xuanyu Zheng, Haitong Zhao, Qidong Zhao, Changsheng Zhang, Bin Zhang

Abstract: Obstacle avoidance and path planning of unmanned aerial vehicles (UAVs) is an essential and challenging task, especially in the unknown environment with dynamic obstacles. To address this problem, a method of UAV path planning based on Deep Q-Learning is proposed. The experience replay mechanism is introduced in the deep reinforcement learning (DRL) process, and a value network is established to calculate the optimal value for the action of the UAV. The optimal flight policy of the UAV is determined through the $$\epsilon $$ -greed algorithm. The experimental results show that the UAV with well-trained model can avoid the obstacles in motion perfectly, and the cruise time is reduced by half compared with the untrained UAV.