Dynamic average consensus estimation over stochastically switching network via quantization communication

In this paper, we consider the problem that a group of agents aims to compute the average of individually estimated noisy parameters by sharing information among a random network of digital links. In this scenario, the average consensus seeking is involved in a two-step procedure. First, each agent estimates the local time-varying parameters individually, and then agents average their estimations by interaction with neighbors through quantized communication. Impact of quantization on the performance of the proposed distributed algorithm is investigated. We prove that the agents´ states converge to a random variable that deviates from the average of the estimated parameters. We derive an upper bound for the asymptotic residual mean square error of the states, which captures effects of the quantization precision and the structure of the random communication networks.