Homotopy conscious roadmap construction by fast sampling of narrow corridors

Abstract

Probabilistic Roadmaps are increasingly being used for robot motion planning. The method makes use of an offline construction of a roadmap. Even though the method is offline, it needs to be initially constructed as quickly as possible for an efficient and near-real time initial motion of the robot. The challenge lies in sampling of multiple narrow corridors wherein the probability of samples is very low. It is important to discover all homotopic groups very early to make good initial decisions from the roadmap. Missing out of even a single homotopic group can lead to no solution or poor solutions. The proposed method uses a multi-strategized approach for sampling of the initial points and then intelligently constructs edges between the points in a multi-strategized manner. The aim is to increase sampling at the narrow corridors and then to facilitate edge connectivity of nodes inside the corridor with the rest of the roadmap, so as to lead to the discovery of all possible homotopic groups between any pair of sources and goals. The approach results in a better performance as compared to uniform sampling and obstacle based sampling.