A Quantitative Theory for Plan Merging

Merging operators in a plan can yield significant savings in the cost to execute a plan. Past research in planning has concentrated on handling harmful interactions among plans, but the understanding of positive ones has remained at a qualitative, heuristic level. This paper provides a quantitative study for plan optimization and presents both optimal and approximate algorithms for finding minimum-cost merged plans. With worst and average case complexity analysis and empirical tests, we demonstrate that efficient and wellbehaved approximation algorithms are applicable for optimizing general plans with large sizes.