Dynamic average consensus over random networks with additive noise

In this paper, we consider distributed dynamic average consensus problem in the presence of uncertainties on information exchange. Two categories of noise are used to characterize these uncertainties: the first is multiplicative noise that captures the randomness of network connections, while the second is additive noise that describes several uncertainty sources. We propose an iterative algorithm that allows each agent to compute/track the average of their private dynamic signals in the presence of both kinds of noise. This algorithm relaxes restrictive assumptions on consensus over random directed network topologies, such as doubly stochastic weights, symmetric link switching styles, etc, and introduces new mechanisms for mitigating effects of communication uncertainties on information aggregation.