A Multi-Agent System for Behavioral Studies

In this paper, we propose a multi-agent system (MAS) to simulate a community of beings that move, eat, reproduce, trade and auction, among other activities. We demonstrate how beings using primitive rules result in emerging, non-primitive behavior. Heterogeneity of real world entities is captured by allowing the user to give the beings different biological characteristics (e.g. level of intelligence and risk taking) using a control panel. Our paper focuses on our modifications and additions to the work inspired from the sugarscape problem [2]. Interesting statistics are obtained from test-runs. The MAS and the GUI that allows the user to control and observe the model were developed using our Agent Base Environment built on top of the Mozart Programming System; a general-purpose development platform that supports concurrency, distribution, resource-aware and symbolic computation.