A Multi-Agent Approach for Hybrid and Dynamic Coevolutionary Genetic Algorithms: Organizational Model and Real-World Problems Applications

Abstract : In this dissertation we assert that modeling Coevolutionary Genetic Algorithms (CGAs) as organizational multi-agent systems overcomes the lack of explicitness at the level of the algorithms structure, interactions and adaptation to existing models and platforms. We therefore introduce MAS4EVO, Multi-Agent Systems for EVolutionary Optimization, a new agent (re-)organizational model based on Moise+ and dedicated to evolutionary optimization. This model was used to describe existing CGAs as well as to build two new variants, hybrid and dynamic, of a competitive CGA. MAS4EVO is implemented in DAFO (Distributed Agent Framework for Optimization) which permits the use, the manipulation and the distribution of these CGAs, on hard optimization problems. The CGAs experimentations were conducted on two business problems, the first one being an inventory management problem and the second one being a new topology control problem in wireless ad hoc networks.
Document type :
Theses
Complete list of metadatas

Cited literature [215 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00785695
Contributor : Florent Breuil <>
Submitted on : Wednesday, February 6, 2013 - 4:59:26 PM
Last modification on : Tuesday, October 23, 2018 - 2:36:03 PM
Long-term archiving on : Saturday, April 1, 2017 - 5:13:15 PM

Identifiers

  • HAL Id : tel-00785695, version 1

Citation

Gregoire Danoy. A Multi-Agent Approach for Hybrid and Dynamic Coevolutionary Genetic Algorithms: Organizational Model and Real-World Problems Applications. Multiagent Systems [cs.MA]. Ecole Nationale Supérieure des Mines de Saint-Etienne, 2008. English. ⟨NNT : 2008EMSE0017⟩. ⟨tel-00785695⟩

Share

Metrics

Record views

809

Files downloads

1129