On transformation of conditional action planning to linear programming

Abstract

Planning in Artificial Intelligence is a problem of finding a sequence of actions that transform given initial state of the problem to desired goal situation. In this work we consider computational difficulty of so called conditional planning. Conditional planning is a problem of searching for plans that depend on sensory information and succeed no matter which of the possible initial states the world was actually in. Finding a plan of such problems is computationally difficult. To avoid this difficulty a transformation to Linear Programming Problem, illustrated by an example, is proposed.