Work

A Dynamic Hybrid Sampling Improved RRT* Path Planning Method of the Manipulator for Diverse Map Environments

Year: 2026
Type: preprint
Abstract: To address the core limitations of the original rapidly-exploring random tree (RRT) algorithm in adapting to diverse map environments, including slow convergence, poor path quality, and weak environmental adaptability, a dynamic hybrid sampling improved RRT* (DHSIRRT*) algorithm is proposed. It quantifies environmental complexity through dual metrics of obstacle density and relative channel width, and achieves continuous, smooth adjustment of the weights for uniform sampling, Gaussian sampling, and Cauchy sampling to enable adaptive adjustment of sampling weights across open, normal, dense, and maze typical maps. A two-factor dynamic step size model integrating obstacle distance and path curvature is built to maximize expansion efficiency while guaranteeing collision-free motion, and a two-step strategy of pruning and node fine-tuning improves path smoothness and reduce path length. Statistical experiments with 100 independent runs are conducted in 2D, 3D, and manipulator workstation scenarios. The results show that, compared with the original RRT, RRT*, and goal-biased RRT* algorithms, the proposed method achieves notable reductions in average iterations, planning time, and path length in dense and maze environments, while maintaining a planning success rate above 99%, which verifies the robustness and superiority of the algorithm across diverse map environments.
Language: English
Open Access status: green
OpenAlex ID: W7237209729
Cites: 26
Cited by:
Related to: 10