Optimizing Initial Path Finding in Informed-RRT* with a Novel Map-Adaptive Sampling Technique

Rapidly-exploring Random Trees (RRT) are extensively employed in robotics motion planning due to their efficacy in solving single-query problems. Informed-RRT* enhances RRT* by sampling within a hyperellipsoid to refine the current best solution. However, this method often encounters inefficiencies, particularly in the initial identification of a feasible path, and its subsequent refinement, due to the expansive size of the hyperellipsoid. This paper introduces an improved Informed-RRT* algorithm to address these specific challenges and enhance planning efficiency. The novel Probabilistic Ellipsoid Informed-RRT* (PEI-RRT*) accelerates the discovery of an initial solution through a probabilistic ellipsoid sampling technique. We also propose the Adaptive Probabilistic Ellipsoid Informed-RRT* (APEI-RRT*), which dynamically adjusts the ellipsoid size based on the environmental context. Numerical simulations comparing state-of-the-art Informed-RRT* with the proposed algorithms demonstrate that PEI-RRT* effectively identifies the initial solution, while APEI-RRT* excels in edge cases involving straight paths or complex environments. The results confirm that the proposed algorithms significantly enhance performance in terms of convergence rate.