Search

International audience; Multi-agent-based simulation (MABS) is the processing of a multi-agent model of a complex system by a simulation platform that controls its execution. The objective is the understanding of the dynamic of this complex system with the experimenting of different configurations for the same multi-agent model. Following a scheduling process, an activated agent has to act according to his context, that is his current perceptible simulation state. In this paper, we propose to delegate the context computation process to the scheduling process. This approach has several advantages. The first is an optimization of the context computation, a single computation being used for several agents. The second advantage is a more configurable design process and a simplification of the reusability of agent behaviors in different simulations. The model that we propose gives a formal framework to support this context computation delegation while preserving agents’ autonomy. We describe a crisis situation to illustrate the benefits of our model and compare our approach with a classical simulation scheduling approach.

Nous presentons une approche pour la conception d’un systeme multi-agent de preparation des acteurs communaux a la gestion de crises, dans le contexte d’une interaction via des tables tangibles distribuees. Notre point focal est celui des normes qui regissent ces interventions. Le role du systeme est d’analyser la validite des interventions au sens de ces normes et de transmettre des feedbacks appropries. La perspective normative permet une modelisation unifiee des 3 facettes (production, coordination et communication) de la collaboration. Nous presentons les premieres etapes de conception, qui s’appuie sur la plateforme JaCaMo, ainsi qu’un exemple d’application a des scenarios simples.

International audience; Les communications multi-parties qui se déroulent au sein d'organisations nécessitent des modes d'interaction (in)directs entre rôles, (in)directs au sein de groupes, etc. La prise en compte de ces modes implique la mise en œuvre d'une modélisation et d'un support intégrant ces deux dimensions. Ainsi, ce papier propose d'étendre le modèle organisationnel normatif MOISE pour spécifier les différents modes d'interactions. Cette spécification a deux objectifs : (i) permettre à l'organisation multi-agent de vérifier que les agents utilisent correctement les modes d'interaction, (ii) permettre aux agents de raisonner sur ces modes comme ils peuvent le faire sur les normes. Seul le premier aspect est abordé dans ce papier. La traduction de cette extension pour le système multi-agent est assurée par le modèle d'interaction EASI. Un exemple de gestion de crise dans les transports est mis en œuvre pour illustrer les capacités d'expression de notre proposition.

The environment is a powerful first-order abstraction in Multi-Agent Systems (MAS), as well as a critical building block. The agents interact in their environment and the effects of their actions are observed and evaluated through this environment. The local complexity of the agents depends on its management of the interaction and action processes. If the environment carries out a part of this processes, the complexity of the agents is reduced. This delegating process implies a centralization of a part of the MAS computations inside the environment and therefore a flexible way to exchange information and to coordinate the agents.

This paper presents an environment-centered approach to design multi-agent solutions to transportation problems. Based on the property-based coordination principle (PbC), the objective of our approach is to solve three recurrent issues in the design of these solutions: the knowledge problem, the space-time dimension and the dynamics of the real environment. To demonstrate the benefits of our approach, two completely different applications, a demand-responsive transportation system and a simulator for crisis management, both based on this principle are presented. For each of them, we show how these recurrent issues have been solved.

When designing agent-based simulation, thechoice of a coordination model is a key issue, since oneof the difficulties is to link the activation of the agentswith their context efficiently. Current solutions separatethe activation phase from the action phase of theagents, and each action phase is based on local agentcontext analysis which is time-expensive. Moreover, becausethe link between the context and the action is aninternal part of the agent, it is more difficult to modifythe way the agent reacts to the context without alteringthe way the agent is implemented. Our proposal, calledEASS (Environment as Active Support for Simulation),is a new approach for agent activation, where the contextis analysed inside the environment and conditionsthe activation of the agents. The main result of contextualactivation is to simplify the achievement of complexsimulations and to decrease run-time. The EASS modelhas been implemented within the kernel of MadKit, amulti-agent platform, and the first results are given.