A Framework of Simplifications in Learning to Plan

Learning shows great promise to extend the generality and effectiveness of planning techniques. Research in this area has generated an impressive battery of techniques and a growing body of empirical successes. Unfortunately the formal properties of these systems are not well understood. This is highlighted by a growing corpus of demonstrations where learning actually degrades planning performance. In this paper we view learning to plan as a search problem. We argue that the complexity of this search precludes a general solution and can only be approached by making simplifying assumptions. We discuss the frequently unarticulated commitments which underly current learning approaches. From these we assemble a framework of simplifications which a learning planner can draw upon. These simplifications improve learning efficiency but not without tradeoffs.