Automatically Modeling Group Behavior of Simple Agents

Mathematical modeling and analysis of collective behavior of multi-agent systems is an important tool that will enable researchers to study even very large systems, validate agent models and get insight into multi-agent system design. Biologists, for example, can compare the model’s predictions to the observed collective behavior of simple organisms, such as social insects, to understand individual organism behavior. We describe the process of automatic construction of models of collective behavior. This process consists of 1) observing the sequence of agent behaviors, 2) inducing a model of the agent’s behavior from these observations, 3) translating the model into equations describing group behavior, 4) analyzing the equations to learn more about the system. The focus of this paper is the last two steps, including a “recipe” for creating equations of collective behavior from the details of the individual agent controller.