Vallati, Mauro, Chrpa, Lukáš and Crampton, Andrew (2013) Underestimation vs Overestimation in SAT-based Planning. In: Proceedings of the XIII Conference of the Italian Association for Artificial Intelligence (Ai*iA-13). Lecture Notes in Computer Science, 8249 . Springer, Turin, Italy, pp. 276-287. ISBN 978-3-319-03524-6

Planning as satisfiability is one of the main approaches to finding parallel optimal solution plans for classical planning problems. Existing high performance SAT-based planners are able to exploit either forward or backward search
strategy; starting from an underestimation or overestimation of the optimal plan length, they keep increasing or decreasing the estimated plan length and, for each fixed length, they either find a solution or prove the nsatisfiability of the corresponding SAT instance.
In this paper we will discuss advantages and disadvantages of the underestimating and overestimating techniques, and we will propose an effective online decision system for selecting the most appropriate technique for solving a given planning problem. Finally, we will experimentally show that the exploitation of such a decision system improves the performance of the well known SAT-based planner SatPlan.

draft0.pdf - Accepted Version

Download (227kB)


Downloads per month over past year