Distributed information filtering using consensus filters

In this paper we present a new information consensus filter for distributed dynamic-state estimation. Estimation is handled by the traditional information filter, while communication of measurements is handled by a consensus filter. First and second-order statistics of local estimates are discussed. It is shown that local information consensus filter estimates are unbiased, and the actual variance of the local estimation errors is comparable to a centralized estimate. However, local agents believe their local estimates are less accurate.