Encoding HTN Heuristics in PDDL Planning Instances (Extended Abstract)


The paper sketches a transformation process that allows Hierarchical Task Network (HTN) based domain-dependent planning heuristics to be encoded into Planning Domain Definition Language (PDDL) [1] planning instances. SHOP2 [2] is chosen as a representative of the many different HTN planning approaches in the literature, but the paper discusses modifications of the transformation applicable to other HTN based planning approaches, too. The evaluation, based on five state-of-the art PDDL planners from the public domain, shows that incorporating HTN heuristics accelerates the planning process in terms of computation time and improves the quality of the resulting plans.

In Ingo J. Timm, Matthias Thimm (Eds.): KI 2013: Advances in Artificial Intelligence - 36th Annual German Conference on AI, pp. 292-295. Heidelberg (Springer).