This paper describes an enhanced version of our previously proposed adaptive framework for single shot motion planning. This framework is versitile, and particularly suitable for crowded environments. Our iterative strategy analyzes the characteristics of the query and adaptively selects planners whose strengths match the current situation.

Contributions in this paper include an automatic method for setting and adaptively tuning planner characterizations, and reducing the reliance on programmer expertise present in the original framework. The adaptive re nement enables the system to evolve parameters specically suited for particular classes of applications. The system now supports articulated robots, which were not supported previously. Our experimental results in complex 3D CAD environments show that our strategy solves queries that none of the planners could solve on their own.