关键词: 拓扑路径, 混合A*算法, 维诺势场, 路径规划

Abstract: Path planning is aimed at finding an optimal or feasible path for a mobile object from a start point to a goal in a given environment. However, existing methods are limited in environmental modeling, feasible path search, and obstacle collision risk assessment in underground spaces, resulting in insufficient generalization, efficiency, and motion feasibility. A Topology Information-Guided Spatiotemporal State Graph Search (TG-SGS) algorithm is proposed in this paper. First, a topology-guided path generation mechanism based on the Generalized Voronoi Diagram (GVD) is introduced, where environmental connectivity is utilized to derive a prior path and blind search is avoided. Then, by considering nonholonomic kinematic constraints, incremental spatiotemporal state expansion is performed using motion primitives, enabling variable-step node growth. A virtual channel constraint and a Reeds-Shepp curve acceleration strategy are incorporated to reduce redundant searches and improve efficiency. Furthermore, a Voronoi potential field function is designed to assess collision risks, where obstacle avoidance and path optimization are balanced for enhanced safety. Experimental results demonstrate that the proposed TG-SGS algorithm outperforms existing methods in simple, maze-like, and real-world indoor environments, with higher efficiency and path quality being achieved.

Key words: topological path, hybrid A* algorithm, Voronoi potential field, path planning