Improved Q-Learning-Based Path Planning Method for Unmanned Surface Vehicles Using Electronic Navigational Chart

Abstract

This paper proposes an improved Q-learning-based path planning method for unmanned surface vehicle (USV) in obstacle-dense marine environments using electronic navigational chart (ENC). First, a dynamic reward function integrating safety distance constraints and directional exploration is designed, ensuring efficient navigation towards the target destination while enabling safe obstacle avoidance. Second, a path short-cutting strategy based on Bresenham algorithm is introduced to eliminate redundant nodes on the path, improving the conciseness of the path. Third, a path expansion method based on the planned path is proposed to expand a single path into multiple paths. Simulation results demonstrate the feasibility of the proposed method.