Planning in probabilistic domains using a deterministic numeric planner
Sergio Jiménez, Andrew J. Coles, Amanda J. Smith, Iii Madrid · Strathprints: The University of Strathclyde institutional repository (University of Strathclyde) · 2006
In the probabilistic track of the IPC5 —the last International Planning Competitions — a probabilistic planner based on combining deterministic planning with replanning —FF-REPLAN — outperformed the other competitors. This probabilistic planning paradigm discarded the probabilistic information of the domain, just considering for each action its nominal effect as a deterministic effect. Thus, in certain domains, the plans proposed by this approach are not robust, so replanning occurs too much frequently. This paper describes a new approach to solve probabilistic planning problems, also based on deterministic planning and replanning; but without rejecting the probabilistic information of the domain. In this approach, the probabilistic domain is compiled into a new deterministic domain and the probabilistic information is translated to an action ‘cost model’, used by a numeric planner to improve the robustness of the plans found, reducing the frequency with which replanning occurs.