Output details
11 - Computer Science and Informatics
University of Aberdeen
The statistical mechanics of belief sharing in multi-agent systems
<22>This paper appears in the premier journal for Information Fusion. The paper describes an analysis and design methodology for distributed algorithms that coordinate the exchange of information for extremely large groups of agents (e.g. sensor networks). We use statistical mechanics to create a probability distribution, which relates groups to pair-wise agent divergences and social temperature. We show that for a critical value of social temperature, the groups formed are simultaneously optimal wrt bandwidth efficiency and divergence from ground-truth. The research has significant implications for analysis of group behaviour in networked intelligent agents.