Path Planning of Autonomous Underwater Vehicle Departing from Port Based on Improved Deep Q-Network

This paper presents a path planning scheme for autonomous underwater vehicle (AUV) in port environment based on improved Deep Q-Network (DQN) algorithm. First, port map is gridded and binarized, and port area is separated from other areas according to port shape. Secondly, in order to enable the agent to leave port area, the range of agent exploration randomness is increased in port area. Finally, to solve the problem that the traditional DQN algorithm is difficult to make the agent find the target point after leaving port area quickly, the reward function of DQN is improved. Simulation results verify the effectiveness and efficiency of the proposed scheme.