Exploiting advanced collision detection libraries in a probabilistic motion planner

Abstract

Motion planning is a fundamental problem in a number of application areas, including robotics, automation, and virtual reality. The performance of motion planning is largely affected by the underlying collision detection technique. In this paper we report the results of an experimental evaluation of several recent collision detection libraries within the context of motion planning for rigid and articulated robots in 3D workspaces. The libraries investigated have been chosen based also on their free availability to the research community. Results reported in this paper show that some of the collision detection packages investigated are very sensitive to the type of problem to be solved, possibly determining the best performance on certain problems and proving very inefficient or even not applicable on different problems. Other collision detection libraries are much less sensitive to the type of problem, although they do not necessarily exhibit the best performance on any given problem. These considerations suggest that a motion planner could take advantage from the ability to select one among a range of collision detection libraries based on characteristics of the problem to be solved which could be known a priori.