BibTeX
@MISC{Mathieu_universitédes,
author = {Ph. Mathieu and S. Picault},
title = {Université des Sciences et Technologies de Lille, SMAC team, LIFL.},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
In this paper we defend the advantages of pushing the interaction as a concrete means (not just an abstraction) in the representation of agent behaviors. While the behavior of an agent is usually deeply encoded in its architecture and relies on its very abilities, we rather dissociate agents from interactions, regarding as well the conceptual viewpoint as the implementation. Indeed, we associate both agents and interactions with a specific ontology. This approach is especially valuable for increasing the reutilisability of the interactions, which can be written very often in a generic way even when agents are bound to specific domains. We show here that this approach is particularly well adapted to broad simulation scales, regarding as well the number of individuals as the diversity of the behaviors. This paper clearly addresses issues at the intersection between software engineering and knowledge representation in agent simulation systems; thus we present the IODA methodology which provides guidelines in the design of such simulations.
