A Hexagonal Grid Cell-inspired Collision Avoidance Motion Planning Algorithm for USVs in a Velocity Domain-based Marine Environment

Optimal motion planning is a challenging problem for unmanned surface vehicle (USV) systems, especially in complex ocean environments, with inevitable dynamic obstacles induced by currents or wind flows. In this work, we propose a new motion planning algorithm for USVs considering flowing obstacles, path distance and time consumption optimization, inspired by neuron structure of entorhinal grid cells in mammals' brains, and how these grid cells learn and build cognitive maps. The proposed collision avoidance motion planning algorithm comprises three key components. First, a hexagonal grid cell-based neural network is designed to provide references of paths with optimal distance for vehicles, where one of the six neurons surrounding the vehicle's current position will be selected due to their residual lengths to the target. Second, a velocity domain that collects the kinematics of environmental flows, dynamic obstacles and the vehicle is constructed. The optimal reciprocal collision avoidance (ORCA) method is then applied to obtain the most efficient collision avoidance velocity vector for the USV, and an adjustment velocity component can be generated to eliminate both flow effects and collisions. The effectiveness of the proposed approach is demonstrated through extensive simulations in 2D conditions.