Abstract
In recent years, the rapidly exploring random tree star (RRT*) algorithm has continued to attract significant research attention in the field of path planning due to its simple implementation and asymptotic optimality. However, RRT* still suffers from limitations such as slow convergence speed and insufficient adaptability to complex environments. Its efficiency significantly decreases when navigating environments containing narrow passages. To address these limitations, this paper proposes an adaptive RRT* algorithm (A-RRT*) that achieves faster convergence and superior environmental adaptability than the standard RRT*. Firstly, we introduce the expansion superposition method to accurately detect narrow passages, enhancing the algorithm’s ability to handle them. This method facilitates swift traversal through the strategic placement of subtree root nodes within these identified regions. Furthermore, these subtrees are dynamically activated or deactivated based on spatial distance, effectively mitigating the computational explosion associated with maintaining multiple trees. After obtaining an initial solution, we design a circular sampling region and a specific node connection strategy to improve sampling efficiency and reduce the computational complexity of achieving path optimality. Comparative simulation experiments against the standard RRT* demonstrate the superiority of the proposed algorithm in terms of convergence speed, environmental adaptability, and computational efficiency.