Scoring functions of approximation of STRIPS planning by linear programming - Block world example

STRIPS planning is a problem of finding of a set of actions that transform given initial state to desired goal situation. It is hard computational problem. In this work an approximation of STRIPS block world planning by linear programming is shown. The cost of such approach is that algorithm can results in non-interpretable solutions for some initial states (what is followed by assumption P ≠ NP). This is because the discrete domain (truth or false) is transformed to continuous domain (LP program). Additionally two scoring functions have been introduced to estimate quality of the plan. Proposed approach is illustrated by exemplary simulation.