Additive Merge-and-Shrink Heuristics for Diverse Action Costs

Abstract

In many planning applications, actions can have highly diverse costs. Recent studies focus on the effects of diverse action costs on search algorithms, but not on their effects on domain-independent heuristics. In this paper, we demonstrate there are negative impacts of action cost diversity on merge-and-shrink (M&S), a successful abstraction method for producing high-quality heuristics for planning problems. We propose a new cost partitioning method to address the negative effects of diverse action costs on M&S. We investigate nonunit cost IPC domains, especially those for which diverse action costs have severe negative effects on the quality of the M&S heuristic. Our experiments demonstrate that in these domains, an additive set of M&S heuristics using the new cost partitioning method produces much more informative and effective heuristics than creating a single M&S heuristic which directly encodes diverse costs.

Merge-and-Shrink: A Compositional Theory of Transformations of Factored Transition Systems

“…As mentioned in the introduction, the purpose of this paper is to develop a solid and extensible theory of merge-and-shrink, both to consolidate and unify existing work and to provide a firm foundation for future developments. In the interest of maintaining focus, this paper is not concerned with experimental evidence of the utility of merge-and-shrink, which has already been established in the literature (e.g., Dräger et al, 2006;Helmert et al, 2007;Helmert, Haslum, & Hoffmann, 2008;Nissim et al, 2011;Hoffmann et al, 2014;Sievers et al, 2014;Fan et al, 2014;Sievers et al, 2015;Torralba & Hoffmann, 2015;Torralba & Kissmann, 2015;Sievers et al, 2016;Eriksson et al, 2017;Fan, Müller, & Holte, 2017;Eriksson, Röger, & Helmert, 2018;Fan, Holte, & Müller, 2018;Sievers, Pommerening, Keller, & Helmert, 2020). Instead, this last section before the conclusions highlights applications of our theory, both in relation to previously published work and in a wider context including possible future research directions.…”

“…However, the available evidence suggests that the heuristic search aspects of SymBA* only play a minor role in its overall performance, and its blind symbolic search configurations perform best in many common benchmarks (Torralba, 2015). Fan et al (2017) consider the use of merge-and-shrink abstractions within the framework of cost partitioning . Cost partitioning is a conceptual framework for the additive combination of admissible heuristics.…”