Work

Accelerating Informed-RRT* through Goal-Projected Sampling and Adaptive Informed-Region Refinement

Year: 2026
Type: article
Abstract: Efficient global path planning in complex environments remains a major challenge for sampling-based planners, particularly during the initialization phase. Although Informed-RRT* guarantees asymptotic optimality once a feasible path is found, its performance is often limited by the slow discovery of initial solutions caused by undirected random sampling. To address this limitation, this paper proposes Adaptive goal-projected informed-RRT*, designed to accelerate the initialization phase without sacrificing probabilistic completeness or asymptotic optimality. The main contribution of this work is a hybrid adaptive sampling strategy that integrates directed projection toward the goal with adaptive informed-region refinement. Before an initial solution is obtained, random samples are adaptively projected toward the goal under line-of-sight constraints to prevent exploration from becoming trapped in dead-end corridors. After a feasible path is found, sampling is restricted to an informed elliptical region defined by the current best path cost, allowing the search to focus on states that can potentially improve the solution. Within this region, adaptive objective projection and LOS-based filtering further guide the sampling process while maintaining sampling diversity. Experiments conducted in six environments with varying obstacle densities show that the proposed method reduces planning time by 7.7%, 59.5%, and 56.0% compared with Fast-RRT, Smart-RRT, and Informed-RRT*, respectively. In addition, the proposed approach generates shorter trajectories while maintaining a 100% success rate, resulting in an average overall efficiency improvement of 9.1 % relative to all baseline methods. These results indicate that accelerating the initialization phase significantly improves computational efficiency and enhances the practical applicability of Informed-RRT* for real-world robotic navigation
Language: English
Open Access status: diamond
OpenAlex ID: W7224498478
FWCI:
Cites: 78
Cited by:
Related to: 10