Path optimization for marine vehicles in ocean currents using reinforcement learning

This study proposes a path planning algorithm for marine vehicles based on machine learning. The algorithm considers the dynamic characteristics of the vehicle and disturbance effects in ocean environments. The movements of marine vehicles are influenced by various physical disturbances in ocean environments, such as wind, waves, and currents. In the present study, the effects of ocean currents are the primary consideration. A kinematic model is used to incorporate the nonholonomic motion characteristics of a marine vehicle, and the reinforcement learning algorithm is used for path optimization to generate a feasible path that can be tracked by the vehicle. The proposed approach determines a near-optimal path that connects the start and goal points with a reasonable computational cost when the map and current field data are provided. To verify the optimality and validity of the proposed algorithm, a set of simulations were performed in simulated and actual ocean current conditions, and their results are presented.