Dynamo: a behavioural analysis model for multi-agent systems

The aim of multi-agent system behavioural analysis is to understand the individual and collective behaviour of agents. This is useful for improving the agents' efficiency or for studying an unknown system. It is a difficult problem for many reasons, such as system heterogeneity and the large set of events that arise at any time. In this paper, we propose a DYNamic Agent MOdel (Dynamo), relying on the notion of semi-structured data. This paradigm allows one to store and query, in a homogeneous form, information about the structure, contents and evolution of data. We show how to query knowledge about the behaviour of entities, and how to classify them by detecting their respective roles. We also discuss the possibility of integrating the model into an intelligent architecture.