ABSTRACT In this paper, a hierarchical planning algorithm for complex urban environment is proposed, which combines the DAB‐RRT* algorithm with the dynamic window algorithm to better guide and optimize the global path. At the global level, adopt multi‐modal optimization strategy to improve the planning efficiency and dynamic obstacle avoidance ability: connecting the starting point and the target by biased random sampling, adjusting the sampling probability according to the obstacle density, and introducing dynamic target deviation and reverse deviation strategies to guide the efficient expansion of the bidirectional tree structure. In the process of expansion, the methods of dynamic direct connection, candidate node rotation and artificial potential field are combined to realize the adaptive adjustment of avoiding obstacles and step size. The post‐processing of the path adopts deleting redundant nodes and B‐spline smoothing to make the path smoother. At the level of local action planning, we put forward the mechanism of “global–local collaborative dynamic guidance domain,” which turns a fixed global path into a flexible space area that can be dynamically adjusted, and its constraint range will change in real time with the complexity of the environment, and relax the restrictions in crowded areas and keep the route consistent in open areas through two‐way feedback. At the same time, an adaptive target attraction strategy is introduced to balance the needs of approaching the target and local obstacle avoidance based on the distance gradient. In the evaluation function of the local planner, the guide domain fitting term is added, and the trajectory conforming to the global reference path is selected first, so that the coordination of local behavior and global path can be realized on the premise of ensuring the overall task goal.
Zhou et al. (Tue,) studied this question.