A Spatial-Temporal-Performance Adaptive Hybrid Sampling Strategy for Mobile Robot Path Planning
Abstract
RRT* is a representative sampling-based path planning algorithm, but its convergence efficiency is limited by global uniform sampling. Existing hybrid sampling methods combine multiple samplers to balance exploration and exploitation, yet their fixed sampling proportions make them difficult to adapt to environments with different structural complexities. To address this issue, this paper proposes TDA-RRT*, a spatial-temporal-performance adaptive variant of RRT*. The proposed method dynamically adjusts the weights of uniform sampling, normal-distribution sampling, and goal-biased sampling from the spatial, temporal, and performance perspectives. Specifically, the spatial dimension extracts environmental features to initialize sampling weights, the temporal dimension guides the transition from global exploration to local exploitation, and the performance dimension provides online feedback based on the actual effectiveness of each sampler. Comparative experiments in narrow passage, maze, and forest scenarios show that TDA-RRT* reduces the average number of convergence iterations by 65%, 48%, and 20% compared with RRT*, Informed-RRT*, and Hybrid-RRT*, respectively, while maintaining comparable path quality. The results verify the effectiveness of the proposed adaptive hybrid sampling strategy.