Policies and Emergent Behaviour Policies and Emergent Behaviour

A major current challenge in the area of complex multi-agent systems is the conceptual approach to, and practical treatment of, emergent behaviour. This paper indicates how Formal Methods may be applied to engineer multi-agent adaptive systems exhibiting emergence, by consideration of an ‘emergence predicate’ in the specification and the incremental derivation of interacting components that achieve that predicate. To perform that task, ‘policies’ are introduced as a novel technique for constraining agents in a hierarchical manner so that the multi-agent implementation as a whole exhibits the specified emergence. The ideas are demonstrated on models of the foraging behaviour of an ant colony, and used to analyse various designs appearing in the literature. Hu Jun received an M.Sc. in Computer Application from Kunming University of Science and Technology, Kunming, China, and a Ph.D. in Computer Science and Technology from Zhejiang University, Hangzhou, China. In 2008 he was a postdoctoral fellow at UNU-IIST working on multi-agent systems; he is currently Associate Professor of Computer Science and Technology at Hunan University, Hunan, China. His main research interest focuses on Artificial Intelligence, multi-agent systems and software engineering. Very recently, it has also included areas such as policy-based management and trustworthy software and other emergent phenomena. Jeff Sanders is Principal Research Fellow at UNU-IIST. His interests lie largely in Formal Methods. Copyright c © 2009 by UNU-IIST