Case-based Planning for Problems with Real-valued Fluents: Kernel Functions for Effective Plan Retrieval

Case-based planning (CBP) re-uses existing plans as a starting point to solve new planning problems. In this work, we address CBP for planning in PDDL domains with real-valued fluents, that are essential to model real-world problems involving continuous resources. Specifically, we propose some new heuristic techniques for retrieving a plan from a library of existing plans that is promising for a new given planning problem, i.e., that can be efficiently adapted to solve the new problem. The effectiveness of these techniques, which derive much of their power from the use of the numerical information in the planning problem specification and in the library plans, is then evaluated by an experimental analysis.