AIRRT*: An improved fast path planning algorithm for manipulators

Abstract

Path planning algorithms based on sampling have been widely studied and applied in recent years, among which the Informed-RRT* algorithm is a typically progressive optimization algorithm. To overcome the weakness of low optimization efficiency of the Informed-RRT* algorithm, an improved algorithm AIRRT* (Adaptive Informed-RRT*) is proposed, which can greatly improve optimization efficiency and improve the quality of optimized paths. According to the node distribution of the initial path, the appropriate node construction areas are dynamically selected first in the process of direct local sampling. Then collision detection and reconstruction are performed. Unlike the traditional algorithm, the AIRRT* algorithm determines the area to be reconstructed firstly, then performs direct local sampling in the reconstructed area, and finally reconstructs the path to improve the sampling efficiency. In addition, to reduce the searching difficulty of global path nodes, the invalid nodes are removed in reconstruction, which limits the growth of nodes in optimization. AIRRT* algorithm is compared with the common algorithms through simulation experiments, which proves that AIRRT* algorithm has obvious advantages in algorithm efficiency and path quality and has good applicability in varied scenarios.