Refinement Search as a Unifying Framework for Analyzing Planning Algorithms

Despite the long history of classical planning, there has been very little comparative analysis of the performance tradeoffs offered by the multitude of existing planning algorithms. This is partly due to the many different vocabularies within which planning algorithms are usually expressed. In this paper, I show that refinement search provides a unifying framework within which various planning algorithms can be cast and compared. I will provide refinement search semantics for planning, develop a generalized algorithm for refinement planning, and show that all planners that search in the space of plans are special cases of this algorithm. I will then show that besides its considerable pedagogical merits, the generalized algorithm also (i) allows us to develop a model for analyzing the search space size, and refinement cost tradeoffs in plan space planning, (ii) facilitates theoretical and empirical analyses of competing planning algorithms and (iii) helps in synthesizing new planning algorithms with more favorable performance tradeoffs. I will end by discussing how the framework can be extended to cover other planning models (e.g. state-space, hierarchical), and richer behavioral constraints.