Approximation of action theories and its application to conformant planning

Problem solving in the context of artificial intelligence (heuristics, search strategies, etc.) (68T20) Logic in artificial intelligence (68T27) Reasoning under uncertainty in the context of artificial intelligence (68T37)

Review text

Conformant planning is an approach that devises a plan which reaches a given goal from a set of possible initial states of the world. In this article, approximation theories for the action language AL are studied and applied to the problem of generating conformant plans. Approximations replace the non-deterministic transition diagrams caused by causal laws in AL, where actions can have indirect effects, by a deterministic transition diagram with partial states. The article shows that a concise description of such an approximation can often be given by a logic program under the answer sets semantics. Complex initial situations and constraints on plans are expressed by logic programming rules. With this technique, the problem of finding a conformant plan can be reduced to computing the answer sets of the logic program. For this purpose, general answer set solvers can be used. The article describes several planning algorithms based on these solvers and gives a sufficient condition for their completeness, i.e., for finding a plan if one exists. Experiments are presented where the algorithms are compared to other state-of-the-art conformant planners on a number of well-known benchmarks and two new benchmarks which are rich in static causal laws.

Recommendations

Cites work

Cited in

(18)

Describes a project that uses

Uses Software

This page was built for publication: Approximation of action theories and its application to conformant planning

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q543583)