On Finding Narrow Passages with Probabilistic Roadmap Planners

A probabilistic roadmap is a network of simple paths connecting collision-free conngurations obtained by sampling a robot's connguration space at random. Several probabilistic roadmap planners have solved unusually diicult path-planning problems, but their ee-ciency remains disappointing when the free space contains narrow passages. This paper provides foundations for understanding the eeect of passages on the con-nectedness of probabilistic roadmaps. It also proposes a new random sampling scheme for nding such passages. An initial roadmap is built in a \dilated" free space allowing some penetration distance of the robot into the obstacles. This roadmap is then modiied by resampling around the links that do not lie in the true free space. Experiments show that this strategy allows relatively small roadmaps to reliably capture the free space connectivity.