To address path planning for unmanned surface vessels(USVs) in ocean environments affected by physical disturbances, such as ocean currents and obstacles, an energy-optimal algorithm based on improved reinforcement learning is proposed. First, a two-dimensional ocean-current model comprising multiple random vortices and a plane kinematic model of the USV is established. Then, the study determined whether a waypoint is reachable using the relative velocity relationship between the USV and the ocean current. An improved reward function, an action set, and a state set are used to obtain a global optimal path, and the B-spline method is applied to smooth the plan. Finally, numerical simulations are performed in two typical environments. The simulation results show that a path with optimal energy consumption and smoothness can be planned based on the proposed algorithm.
