An exploratory analysis for designing robust interactions in multi-agent systems

Abstract

In large multi-agent system (MAS), the definition of protocols between pairs of agents is not enough for guaranteeing the absence of undesirable communication organizations and the presence of desirable ones. This paper presents a technique and a tool for designing robust MAS communication architectures. In particular, in the proposed technique, designers are not only recommended to design the desired communication protocols, but also the undesired ones and the organization structures. The proposed tool performs data mining and on-line analytical processing (OLAP) on the logs of execution of MAS that use the Foundation for Intelligent Physical Agents (FIPA) protocol, in order to debug MAS communication architectures. The experimental results advocate that this approach is useful for MAS debugging and testing. © 2011 Springer-Verlag Berlin Heidelberg.