关键词:
A
global path planning
robot
routing planning
scene graph
spatial connectivity
TP242.6
摘要:
Real-time global path planning for large-scale scenarios is a challenging problem due to the difficulty of maintaining high-definition maps and the high computational cost for real-time planning in changing environment. In this paper, we propose a global path planning framework based on route scene graph (RSG). Based on the hierarchical scene graph framework, we focus on mining spatial connectivity and extracting RSG representing scene road connections from precise obstacle information. The method updates the connection relationships between nodes locally through the minimum spanning tree to make the graph connections sparse. Loop processing ensures the optimality of planning without precise obstacle maps. With the abstract road information in RSG, road pre-planning and multi-level planning can be used to accelerate global path planning. We evaluate the method in both simulation and real environments. Compared to the search-based methods Far Planner (1–10ms), A*, D* Lite (10–100 ms), and the random sampling-based methods BIT*, SPARS (10–100 ms), the method achieved the sub-millisecond level (0.1–1 ms) planning speed in over 10 000 m2 scenarios. (Figure presented.) © Shanghai Jiao Tong University 2025.