Classifying Agent Systems

To select an appropriate tool or tools to build an agent-based system we need to map from features of agent systems to implementation technologies. In this paper we propose a simple scheme for classifying agent systems. Starting from the notion of an agent as a cluster concept, we motivate an approach to classification based on the identification of features of agent systems, and use this to generate a high level taxonomy. We illustrate how the scheme can be applied by means of some simple examples, and argue that our approach can form the first step in developing a methodology for the selection of implementation technologies.