Consensus-based distributed estimation in homogeneous sensor networks

Abstract

This paper studies distributed estimation of linear dynamical target in homogeneous sensor networks. Just some sensors can get observations of the target. A distributed estimator composing of Luenberger observers and consensus strategy is applied. Based on spectral radius analysis and algebraic graph theory, necessary conditions of distributed detectability are given for time-invariant and randomly switching networks, respectively. By applying Lyapunov functional analysis and solving a special LQR problem, sufficient detectability conditions are proposed for both networks. The sufficient conditions are composed of an explicit inequality concerning weights, topology switching probabilities and the eigenvalues of the target's system matrix. Simulation examples are given to illustrate the results.