Research on Dynamic Path Planning Algorithm for Unmanned Underwater Vehicles Based on Multi-step Mechanism DDQN

The dynamic path planning capability of UUV in complex underwater environment is difficult to meet the needs of advanced nature. In this paper, Q(lambda) algorithm and DDQN algorithm are combined, the updated iteration strategy of DDQN algorithm is improved, and a MS-DDQN algorithm is designed to obtain more accurate Q value through continuous multi-step interaction and instant reward. According to this algorithm, the input information and reward function of the algorithm are redesigned, and the UUV path planning is divided into two network modules: navigation and obstacle avoidance, which improves the dynamic path planning ability of UUV. The simulation results show that the MS-DDQN algorithm has a high planning success rate and can meet the dynamic operation requirements of UUV.