Coverage Path Planning for Maritime Search based on Reinforcement Learning with Path Expansion Strategy

In maritime search and rescue (SAR) operations, the planning of search paths directly impacts the efficiency of rescuing individuals in distress at sea. This study proposes a path-expanding reinforcement learning model designed to achieve complete coverage of maritime accident areas. Multiple scenarios are considered in conjunction with reward functions, aimed at maximizing the cumulative success rate of rescue operations within a finite time frame. The strategy allows rescue units to utilize diagonal movement to avoid redundant search efforts. Through comparative analysis, superior starting points for rescue operations are identified. The adoption of the Upper Confidence Bound (UCB) strategy within the reinforcement learning algorithm effectively reduces the likelihood of redundant search efforts and shortens path lengths. Rescue units equipped with path-expanding capabilities demonstrate higher search efficiency compared to conventional rescue path planning units. The path-expanding strategy is combined with the UCB strategy to improve the efficiency and success rate of maritime search and rescue operations through the exploration of diverse search starting points.