Path Updating Tree based fast path planner for unpredictable changing environments

For changing environments, although lots of planning algorithms have focused on how to get a valid and short path, seldom planning algorithms can be employed to extract a sustaining valid path. Especially for large-scale environments, getting a sustaining valid path is required to make intelligent decisions for robots. In this paper, a method of Path Updating Tree (PUT) is proposed to get such a sustainingly valid region path, which approximately estimates the environment using extended nodes in the projected Cspace. Each extended node is used to reflect local crowding information in this paper. An approximate path, which connected goal and start, is generated based on each region's collision possibility, reachable possibility and path length. With environment changing, nodes of PUT are updated and new region paths are generated by regrowing if necessary. Then a two-level path planner is introduced with PUT and local path generator to distribute the computational resource properly. The proposed method has shown to be efficient in large-scale scenarios with crowded obstacles.