An Accessibility graph learning approach for task planning in large domains

In the stream of research that aims to speed up practical planners, we propose a new approach to task planning based on Probabilistic Roadmap Methods (PRM). Our contribution is twofold. The rst issue concerns an extension of GraphPlan[1] specially designed to deal with \local planning" in large domains. Having a reasonably e cient \local planner", we show how we can build a \global task planner" based on PRM and we discuss its advantages and limitations. The second contribution involves some preliminary results that allow to exploit to domain symmetries and to reduce in drastic manner the size of the \topological" graph. The approach is illustrated by a set of implemented examples that exhibit signi cant gains.