Intellektik: Technical report 95-04
Solving Deductive Planning Problems Using Program Analysis and Transformation
Andre de Waal and Michael Thielscher
Two general, problematic aspects of deductive
planning, namely, detecting
unsolvable planning problems and solving a certain kind of postdiction
problem, are investigated.
The work is based on a resource oriented approach to reasoning about
actions and change using a logic programming paradigm.
We show that ordinary resolution methods are insufficient for solving
these problems and propose program analysis and transformation as a
more promising and successful way to tackle them. In particular,
a recently developed specialization technique based on partial
evaluation and regular approximation is refined and adapted for the
specialization of the underlying logic program modeling actions and
change. Furthermore, a regular approximation is proposed as a finite
description for representing and collecting infinitely many combinations of
resources. The usefulness of the developed specialization techniques are
demonstrated on some prototypical examples.