Technical Report No. 160 - Abstract

Stefan Edelkamp
First Solutions to PDDL+ Planning Problems

In this paper we present the design and algorithmic details of a directed search planner to solve benchmark planning problems specified in PDDL+ syntax. The planner copes with durative actions, grounds and groups predicates and instantiates functions. The operators possibly span an infinite state space which is tackled by guided exploration. The heuristic estimates are inferred by plan relaxation, through pattern databases, or by state-to-goal differences of numerical values and are incorporated in scalable heuristic search planning algorithms.