Filtering, decomposition and search space reduction for optimal sequential planning

  • Authors:

  • Stéphane Grandcolas;Cyril Pain-Barre

  • Affiliations:

  • LSIS - UMR CNRS, Domaine Universitaire de Saint-Jérome, Marseille, Cedex 20, France;LSIS - UMR CNRS, Domaine Universitaire de Saint-Jéme, Marseille, Cedex 20, France

  • Venue:
  • AAAI'07 Proceedings of the 22nd national conference on Artificial intelligence - Volume 2
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present in this paper a hybrid planning system Which combines constraint satisfaction techniques and planning heuristics to produce optimal sequential plans. It integrates its own consistency rules and filtering and decomposition mechanisms suitable for planning. Given a fixed bound on the plan length, our planner works directly on a structure related to Graphplan's planning graph. This structure is incrementally built: Each time it is extended, a sequential plan is searched. Different search strategies may be employed. Currently, it is a forward chaining search based on problem decomposition with action sets partitioning. Various techniques are used to reduce the search space, such as memorizing nogood states or estimating goals reachability. In addition, the planner implements two different techniques to avoid enumerating some equivalent action sequences. Empirical evaluation shows that our system is very competitive on many problems, especially compared to other optimal sequential planners.