Visualization of Dynamic Behaviour of Multi-Agent Systems

An extension of the research on analysis of a single agent or agent communities combining advanced methods of visualization with traditional AI techniques is presented, which addresses the problem of not suitable well to present the dynamics of the MAS by visualizing the changes of theMAS along with their quality and context.

Abstract

The extension of our research on analysis of a single agent or agent communities combining advanced methods\nof visualization with traditional AI techniques is presented in this paper. Even though this approach can be used\nfor arbitrary Multi-Agent System (MAS), it was primarily developed to analyze systems falling into Artificial\nLife domain. Traditional methods are becoming insufficient as Multi-Agent Systems (MAS) are becoming more\ncomplex and therefore novel approaches are needed. In this paper we present an extension of our recent\nvisualization tools suite. The previous approach was not suitable well to present the dynamics of the MAS, even\nthough the development of MAS state parameters in time was presented. Our new technique, which is presented\nin this paper, addresses this problem by visualizing the changes of the MAS along with their quality and context.\nThis transparent approach emphasizes MAS dynamics by providing means for discovery of changes in its\ntendencies or in behaviour of either single agent or agent communities. A simulated artificial life environment\nwith intelligent agents has been used as a test bed. We have selected this particular domain because our longterm\ngoal is to model life as it could be so as to understand life, as we know it.