Date of this Version
Department of Computer Science & Engineering, University of Nebraska-Lincoln, Technical Report, TR-UNL-CSE-2013-0001
Information sharing is important in agent-based sensing, especially for large teams where only a small subset of the agents can directly observe the environment. We consider the impact of non-stationarity in the observed phenomenon on the collective beliefs of such large teams. Non-stationarity is challenging because not only must the team converge to consistent, accurate beliefs (as studied previously), but, most importantly, the team must also frequently revise its beliefs over time as the phenomenon changes values. We analytically and empirically demonstrate the difficulty in revising beliefs over time with the standard model and pro-pose two novel solutions for improving belief convergence when observing non-stationary phenomenon: (1) a change detection and response algorithm for cooperative environments, and (2) a for-getting-based solution for non-cooperative environments.