A Flexible Agent Architecture in SPADE

Javier Palanca, Jaime Andres Rincon, Carlos Carrascosa, Vicente Julián, Andrés Terrasa

Abstract: For some years now, multi-agent systems have demonstrated their potential to be used as a base technology to address the development of distributed, intelligent and autonomous systems. However, existing agent platforms still require a higher degree of flexibility in order to effectively combine decision-making processes into the same system. In this context, the main goal of this paper is to improve flexibility in the development of behavior-based agents, by means of introducing the Flexible Agent Architecture (FAA). The FAA defines agents in such a way that each agent may integrate and combine different types of reasoning processes (for example, procedural-based and BDI processes) as behaviors. This definition remarkably empowers the global behavior of the agent, by allowing it to combine different, more or less deliberative techniques in its decision making process, in a transparent and flexible way. In addition, the paper presents a full implementation of the proposal in a real agent middleware called SPADE.