Single-Query Entropy-Guided Path Planning

Motion planning for robots with many degrees of freedom requires the exploration of an exponentially large configuration space. Single-query motion planners restrict exploration to regions of configuration space determined to be relevant to a particular planning query. The heuristics employed by existing single-query planners to estimate the relevance of a region, however, remain unchanged throughout the planning process. An incorrect estimate by the heuristic for a configuration space region will only be corrected by explicit exploration. As a result, unnecessary exploration is performed. In this paper we propose an alternative approach. We observe that every incremental sample improves the planner’s understanding of configuration space. This improved understanding can be exploited to inform the singlequery heuristic of a motion planner. We formalize the improvement in understanding by employing the notion of entropy from information theory and derive a principled method of con...

• Type: Conference
• Year: 2005
• Where: ICRA
• Authors: Brendan Burns, Oliver Brock
• Keywords: Configuration Space, Exploration, ICRA 2005, Motion Planner, Robotics