Frugal Propositional Encodings for Planning

Casting planning as a satis ability problem has recently lead to the development of signi cantly faster planners. One key factor behind their success is the concise nature of the encodings of the planning problems. This has lead to an interest in designing schemes for automated generation of propositional encodings for the planning problems. These encodings are inspired by traditional planning algorithms like state space and causal link planning. We examine the existing schemes for generating encodings for causal planning and suggest several improvements to them. We show that these improvements reduce the number of clauses and/or variables in the encoding. Since the number of clauses and variables are related to the hardness of solving a constraint satisfaction problem, our results provide important directions in e ciently encoding planning problems. Our improvements derive plan encoding that is smaller than all other previously developed plan encodings.