Path Planning Method for Rrt Algorithm Based on Dynamic Region-Constrained Sampling

The traditional Rapidly-exploring Random Tree (RRT) algorithm suffers from excessive randomness, a high proportion of invalid samples, and slow convergence. To address these issues, this paper proposes an improved RRT-based path planning method that incorporates dynamic region-constrained sampling. A dynamic elliptical region is employed to restrict the sampling space, effectively narrowing the sampling range. Combined with a dynamic goal biasing strategy, this approach increases the likelihood of sampling near the target and guides the sampling direction, thereby reducing the number of ineffective samples. During the tree expansion phase, an adaptive step-size mechanism is introduced, which adjusts the expansion step according to the distance to the target and the proximity of nearby obstacles, thereby accelerating convergence. Additionally, a pruning strategy is applied during the search process to eliminate redundant branches and maintain a compact tree structure. Once an initial path is generated, point-wise optimization and B-spline smoothing are performed to enhance path quality. Experimental results demonstrate that the proposed algorithm outperforms the traditional RRT in terms of planning time and path length.