DAE: Planning as Artificial Evolution -- (Deterministic part)

Jacques Bibai 1, 2 Pierre Savéant 1 Marc Schoenauer 2 Vincent Vidal 3
2 TAO - Machine Learning and Optimisation
LRI - Laboratoire de Recherche en Informatique, UP11 - Université Paris-Sud - Paris 11, Inria Saclay - Ile de France, CNRS - Centre National de la Recherche Scientifique : UMR8623
Abstract : The sub-optimal DAE planner implements the stochastic approach for domain-independent planning decomposition. The purpose of this planner is to optimize the makespan, or the number of actions, by generating ordered sequences of intermediate goals via a process of artificial evolution. For the evolutionary part we used the Evolving Objects (EO) library, and to solve each intermediate subproblem we used the constraint-based optimal temporal planner CPT. Therefore DAE can only solve problems that CPT can solve. Compression of subplans into a global solution plan is also achieved efficiently with CPT by exploiting causalities found so far. Because the selection of predicates for intermediate goal generation is still an open question, we have submitted two planners DAE1 and DAE2 that use different strategies for the generation of intermediate goals. An empirical formula has been defined to set a limit on the number of backtracks allowed for solving the intermediate subproblems.
Document type :
Conference papers
Complete list of metadatas

Cited literature [3 references]  Display  Hide  Download

https://hal.inria.fr/inria-00354282
Contributor : Jacques Bibai <>
Submitted on : Monday, January 19, 2009 - 3:16:47 PM
Last modification on : Thursday, April 5, 2018 - 12:30:12 PM
Long-term archiving on : Tuesday, June 8, 2010 - 8:45:58 PM

File

deterministic.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : inria-00354282, version 1

Collections

Citation

Jacques Bibai, Pierre Savéant, Marc Schoenauer, Vincent Vidal. DAE: Planning as Artificial Evolution -- (Deterministic part). The sixth international planning competition ( IPC-6), International Conference on Planning and Scheduling (ICAPS),, Sep 2008, Sydney, Australia. ⟨inria-00354282⟩

Share

Metrics

Record views

461

Files downloads

753