The logic of totally and partially ordered plans: A deductive database approach

The problem of finding effective logic-based formalizations for problems involving actions remains one of the main application challenges of non-monotonic knowledge representation. In this paper, we show that complex planning strategies find natural logic-based formulations and efficient implementations in the framework of deductive database languages. We begin by modeling classical STRIPS-like totally ordered plans by means of Datalog(1S) programs, and show that these programs have a stable model semantics that is also amenable to efficient computation. We then show that the proposed approach is quite expressive and flexible, and can also model partially ordered plans, which are abstract plans whereby each plan stands for a whole class of totally ordered plans. This results in a reduction of the search space and a subsequent improvement in efficiency.