Searching Optimal Parallel Plans: A Filtering and Decomposition Approach

Guillaume Gabriel (1) , Stéphane Grandcolas (1)

1 LSIS - Laboratoire des Sciences de l'Information et des Systèmes

In the domain of planning, searching for optimal plans gives rise to many works. Most of the existing planners search for optimal parallel plans, that is plans which are optimal in the number of steps. FDP is an exception: it searches for optimal sequential plans, that is plans which are optimal in the number of actions. FDP implements a depth-first iterative-deepening search, using a CSP-like structure, decomposition rules, and filtering techniques similar to consistency-enforcing techniques. We propose a new planner based on FDP, which generates optimal parallel plans. This planner has been compared on a set of selected problems with the well-known optimal parallel planner SATPLAN, and with the planner FDP, so as to analyse the effects of the optimization criteria in the performances.

Domaines: Intelligence artificielle [cs.AI]

Guillaume Gabriel, Stéphane Grandcolas. Searching Optimal Parallel Plans: A Filtering and Decomposition Approach. ICTAI, Nov 2009, New York, United States. ⟨hal-02471093⟩