An Agent Architecture to Design Self-Organizing Collectives: Principles and Application

Abstract : Designing teams which have a task to execute in a very dynamic environment is a complex problem. Determining the relevant organization of these teams by using group or role notions might be very difficult and even impossible for human analysts. Although an organization can be found or approximated, it becomes complicated to design entities, or agents in our case, that take into account, at the conception and design phases, all possible situations an agent could face up to. Emergent and self-organizing approaches to model adaptive multi-agent systems avoid these difficulties. In this paper, we propose a new approach, to design Adaptive Multi-Agent Systems with emergent functionality, which enables us to focus on the design of agents that compose the system. In fact, self-organization of the system is led by the environmental feedback that each agent perceives. Interactions and organization evolve, providing an adequate function to the system, which fits to its environment as well. Such functions have enough properties to be considered as emergent phenomena. First, we briefly present the Adaptive Multi-Agent Systems theory (AMAS) and our view of self-organization. In the second part, a multi-level architecture is proposed to model agents and to consider groups of agents as self-organizing teams. In the third part, we describe a sample robot group behaviour, the setting up of traffic in a constrained environment. Our architecture allows the emergence of a coherent collective behaviour: the dedication of corridors to specific directions. Finally, we show what is emergent by the analysis of results arising from measurements of collective phenomena.
Complete list of metadatas
Contributor : Gauthier Picard <>
Submitted on : Wednesday, March 4, 2015 - 9:36:55 PM
Last modification on : Thursday, June 27, 2019 - 4:27:48 PM

Links full text




Gauthier Picard, Marie-Pierre Gleizes. An Agent Architecture to Design Self-Organizing Collectives: Principles and Application. Eduardo Alonso; Daniel Kudenko; Dimitar Kazakov. Adaptive Agents and Multi-Agent Systems, 2636, Springer, pp.141-158, 2002, Lecture Notes in Computer Science, 978-3-540-40068-4. ⟨10.1007/3-540-44826-8_9⟩. ⟨⟩. ⟨hal-01123459⟩



Record views