Distributed nonlinear estimation over time-varying directed networks

Abstract

In this paper, we investigate distributed estimation with nonlinear sensing functions over time-varying directed networks. Each agent sequentially makes nonlinear and noisy measurements of the objective parameter, and aims to estimate the parameter through local information. To this end, an algorithm is designed by introducing two time-varying matrices corresponding to the network graph. Under the condition of B-bounded strong connectivity, and some mild assumptions on sensing functions, the distributed estimator achieves consistent parameter estimates by selecting an appropriate stepsize. Furthermore, we also analyze the asymptotic property of the weighted estimate error sequence, including the asymptotic mean and asymptotic covariance. It is shown that the performance is consistent with that of a widely used centralized estimator and a distributed estimator over a fixed directed communication topology. Finally, we give simulation results to verify the effectiveness of the devised algorithm.