Dynamic Obstacle Avoidance Navigation for USVs Based on the Multihead Attention Mechanism and Distributional Reinforcement Learning

Abstract

AbstractAutonomous navigation of unmanned surface vehicles (USVs) is challenged by environmental uncertainty and safety risks caused by eddies and unknown obstacles. Traditional deep reinforcement learning based on expected returns is prone to decision ...Practical ApplicationsThis paper proposes a method to enhance the navigation of unmanned surface vehicles in complex marine environments, particularly those affected by unpredictable currents and obstacles. The method combines distributional reinforcement ...