Distributed fusion estimation from measurements with correlated random parameter matrices and noise correlation

Abstract

This paper addresses the distributed fusion estimation problem for discrete-time multi-sensor stochastic systems with random parameter matrices. It is assumed that the random parameter matrices in the observation equations are one-step autocorrelated and cross-correlated between the different sensors and the additive noises are also correlated. Under these assumptions, a recursive algorithm is proposed to obtain local least squares linear filters based on the measurements of each sensor, and the distributed fusion filter is designed as the matrix-weighted linear combination of these estimators which minimizes the mean squared estimation error. This research is illustrated by two numerical simulation examples where multi-sensor systems with randomly delayed measurements and missing measurements are considered, respectively, and the performance of the proposed estimators is analysed by comparing the estimation error variances of the distributed and centralized fusion filters.