Intellectics Group: Technical Report 00-02
Extracting State Constraints from PDDL-like Planning Domains.
In Workshop on Analyzing and Exploiting Domain Knowledge for Efficient
Planning. AIPS'2000, Breckenridge, Colorado, USA.
We describe a new method for inferring state constraints from PDDL-style
planning problems. Planning systems can use state constraints as valuable
domain-dependent knowledge to cut the search space and to speed up planning.
A large and important class of state constraints are c-constraints which are
prevalent among the constraints inferred by current techniques. These
techniques are restricted to c-constraints of a certain structure. The
described method finds many of the c-constraints which are found by other
methods as well as complex c-constraints which cannot be inferred by current
techniques. The presented method takes PDDL input, is domain-independent,
and does not use the initial state of a planning problem.