Abstract:

With the widespread application of mobile robots, path planning efficiency has become a key factor influencing their operational performance. To address the path planning efficiency challenge for mobile robots in dynamic environments, this paper proposes an improved Batch Informed Trees (BIT^*) algorithm based on enhanced sampling and local planning. The Kennard-Stone algorithm is employed to achieve uniform sampling. By refining the sampling distribution to reduce point clustering, the algorithm's global exploration capability is enhanced. During the pathfinding phase, ray detection is introduced to control the expansion of the search tree, actively avoiding obstacles when generating new edges, thereby reducing the impact of inefficient expansions on path planning. Rapid path optimization is performed during the search tree exploration. Post-processing strategies are applied to effectively minimize the number of turns and optimize the global path. For dynamic obstacles, a hierarchical path planning framework is designed. The upper layer generates reference paths based on the improved BIT^* algorithm (K-BIT^*), while the lower layer employs an enhanced Informed RRT^* algorithm for local replanning around dynamic obstacles. A simulation environment with obstacle spaces is constructed in MATLAB. Global path planning is conducted using K-BIT^*, BIT^*, D-BIT^*, and Informed RRT^* algorithms. In dynamic environments, the dynamic planning efficiency of K-BIT^* is compared with that of D-BIT^* and DGF-APF. The results demonstrate that the K-BIT^* algorithm achieves shorter path lengths compared to D-BIT^*, along with a 13.7% reduction in planning time and 9.7% fewer waypoints. In dynamic environments, the locally replanned path of K-BIT^* is 0.368 m shorter than that of D-BIT^*, with a lower standard deviation. These findings validate the efficiency of the K-BIT^* algorithm for path planning in dynamic obstacle spaces.

Key words: mobile robot, Batch Informed Trees algorithm, Informed RRT^*, Kennard-Stone sampling, dynamic environment, path planning