Heuristics for planning with SAT

Abstract

Generic SAT solvers have been very successful in solving hard combinatorial problems in various application areas, including AI planning. There is potential for improved performance by making the SAT solving process more application-specific. In this paper we propose a variable selection strategy for AI planning. The strategy is based on generic principles about properties of plans, and its performance with standard planning benchmarks often substantially improves on generic variable selection heuristics used in SAT solving, such as the VSIDS strategy. These improvements lift the efficiency of SAT based planning to the same level as best planners that use other search methods. © 2010 Springer-Verlag.