Distributed belief averaging using sequential observations

Yang Liu, Ji Liu, Tamer Basar, Mingyan Liu

Abstract: This paper considers a distributed belief averaging problem with sequential observations in which a group of n > 1 agents in a network, each having sequentially arriving samples of its belief in an online manner, aim to reach a consensus at the average of their beliefs, by exchanging information only with their neighbors. The neighbor relationships among the n agents are described by a time-varying undirected graph whose vertices correspond to agents and whose edges depict neighbor relationships. A distributed algorithm is proposed to solve this problem over sequential observations with O(1/t) convergence rate. Extensions to the case of directed graphs are also detailed.