Cost-Optimal Planning with Constraints and Preferences in Large State Spaces

This paper deals with planning in the presence of constraints and preferences as proposed for the 5 International Planning Competition. State trajectory constraints are translated into LTL formulae and are compiled into Büchi automata in PDDL format. Preference constraints are compiled into numerical fluents. Values of these fluents are changed by grounded operator effects upon violation. We propose two exploration strategies for optimal planning in PDDL3 domains: (i) a best-first branchand-bound weighted heuristic search; (ii) an external breadth-first search exploration algorithm that exploits secondary memory, such as harddisk, to save the open and closed lists. We prove an upper bound on the locality of the search in planning graphs that dictates the number of layers that have to be kept to avoid re-openings. For non-optimal planning, we present an external variant of enforced hill climbing.