CAREER: An Analysis of the Dynamics of Adaptive Multiagent Systems, with Application to Global Information Exchange Systems

点击翻译按钮获取中文摘要

The goal of this research project is to develop agent-based systems--for electronic commerce, automated manufacturing, or supply-chains, for example--where the agents adapt their behaviors so as to maximize their individual rewards. A general predictive theory is developed, which can determine expected emergent behaviors from initial agent capabilities and interaction protocols. A methodology for the engineering of utility-based adaptive multi-agent systems is derived from these studies. Finally, the theories are applied to the development of an incentive-compatible information exchange protocol and its implementation in a prototype information exchange multi-agent system. Information serves as a currency in the protocol, and its value is determined via the agents' adaptive behavior. The aggregate of all the users' knowledge, opinions, and biases are immediately accessible to all users. Such research is critical to begin to predict the emergent behavior of systems and to move in the direction of integrating multi-agent systems theory and practice. New graduate curricula with an emphasis on decentralized systems and an undergraduate course on agent-based software engineering will be developed.