Prise en compte des comportements anticipatifs dans la coordination multi-agent : application à la simulation de trafic en carrefour

Abstract : Multi-agent systems allow the simulation of complex phenomena which are not easily describable in an analytical way. This approach is often based on the coordination of agents whose actions and interactions involve the emergence of the phenomenon to be simulated. When the aim is to mimic the human behaviour, it is sometimes necessary to use a competitive coordination and to be able to reproduce anticipatory behaviours. This thesis addresses the issue of recognition and anticipation of coordination context. Our work is based on an approach of the anticipation suggested by Rosen : "preventive anticipation" which consists in adapting its actual behaviour to avoid some undesired states. We propose a formalization of this concept based on constraints networks processing. The underlying decisional model uses constraints propagation techniques which allow each agent to gather the effects of an action and thus to detect some undesired states. The proposed algorithm is generic and filters out, before any coordination, all actions leading to one of the non-desired states of the system. Implemented within the framework of a simulation of traffic developed at INRETS and called ARCHISIM, our proposal enable to introduce opportunistic behaviours without any risk of having deadlock in the center of the intersection. The various evaluations validate the emergence of traffic phenomena and confirm the relevance of the approach.
Complete list of metadatas

https://hal.archives-ouvertes.fr/tel-01992947
Contributor : Arnaud Doniec <>
Submitted on : Thursday, January 24, 2019 - 4:37:22 PM
Last modification on : Saturday, February 2, 2019 - 1:22:03 AM
Long-term archiving on : Thursday, April 25, 2019 - 2:51:14 PM

File

these.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01992947, version 1

Collections

Citation

Arnaud Doniec. Prise en compte des comportements anticipatifs dans la coordination multi-agent : application à la simulation de trafic en carrefour. Intelligence artificielle [cs.AI]. Université de Valenciennes, 2006. Français. ⟨tel-01992947⟩

Share

Metrics

Record views

24

Files downloads

40