Search in abstract spaces has been shown to produce useful admissible heuristic estimates in deterministic domains. We show in this paper how to generalize these results to search in stochastic domains. Solving stochastic optimization problems is significantly harder than solving their deterministic counterparts. Designing admissible heuristics for stochastic domains is also much harder. Therefore, deriving such heuristics automatically using abstraction is particularly beneficial. We analyze this approach both theoretically and empirically and show that it produces significant computational savings when used in conjunction with the heuristic search algorithm LAO*. © Springer-Verlag Berlin Heidelberg 2005.
CITATION STYLE
Beliaeva, N., & Zilberstein, S. (2005). Generating admissible heuristics by abstraction for search in stochastic domains. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 3607 LNAI, pp. 14–29). Springer Verlag. https://doi.org/10.1007/11527862_2
Mendeley helps you to discover research relevant for your work.