Rule-Constrained Hierarchical Fusion Path Planning for Unmanned Surface Vehicles

Abstract

Path planning for unmanned surface vehicles in dynamic waters requires coordination of global routing, collision avoidance, maneuverability, and maritime encounter rules. This study develops a rule-constrained hierarchical framework to integrate these requirements. The global planner combines adaptive weighted A* guidance, adaptive rapidly exploring random tree search, path pruning, and curvature-checked Bézier smoothing. The local planner integrates artificial potential field risk evaluation with the dynamic window approach, incorporating closest-point-of-approach indicators, asymmetric ship domains, and encounter-specific maneuver constraints. A coordinator manages risk-triggered avoidance, return to the global path, and replanning. Kinematic simulations evaluated static planning, smoothing, vessel encounters, and integrated navigation. Postprocessing produced acceptable paths for 98 of 120 inputs (81.7%), satisfying the prescribed clearance and turning-radius constraints. Under perturbed target conditions, 68 of 90 individual encounter trials (75.6%) and 19 of 30 integrated-navigation trials (63.3%) achieved complete success. Nominal integrated navigation completed avoidance and path return, while a separate blockage scenario demonstrated recovery through braking and global replanning. However, substantial detours and stalled trajectories limited navigation efficiency and completion. These findings support feasibility under the tested simulation conditions and identify trajectory efficiency and stall recovery as priorities for further development.