2016 Galland, Stéphane and Balbo, Flavien and Picard, Gauthier and Boissier, Olivier and Gaud, Nicolas and Rodriguez, Sebastian, Environnement Multidimensionnel Pour Contextualiser Les Interactions Des Agents. Application à La Simulation Du Trafic Routier Urbain., Revue d Intelligence Artificielle, 30(1-2): 81—108, 2016.
Resumen: The environment, as a space shared between agents is essential to multi-agents systems (MAS). Depending on the systems, it responds to di erent points of view. It is described as physical or communicative as that agents interact through situated actions or exchanging messages, or social if a social model governs the interactions. Each of these views could be declined into a dimension with its own model. For a complex system in which di erent dimensions should be combined, there are only ad-hoc solutions for a global environment model. The consequence is a limited reusability and modularity of the use of environment models. An alternative is to consider the environment as the juxtaposition of its dimensions and make the agent the place for combining the information conveyed in these dimensions. This choice of design increases the agent complexity, and makes them dependent upon the mechanisms for managing the interactions between the dimensions. Finally, the implementation of contextual interactions, i.e. constrained by the rules of the MAS, requires a management of the interactions that is independent of the agents. In this paper, a model is proposed for ensuring the combination of the environment dimensions for the modeling and implementation of contextualized interactions between agents. This model is developed with the SARL agent-oriented programming language. This proposal is illustrated with a road trafic simulation application to the city of Belfort.