Comparison of distributed receding horizon filtering for linear discrete-time systems with uncertainties

A distributed receding horizon filtering for linear discrete-time systems with uncertainties is presented. The choice of receding horizon strategy makes the estimation fusion algorithms robust against dynamic model uncertainties. All distributed fusion algorithms are based on the fusion formulas which represent weighted sums of local receding horizon Kalman estimates with matrix weights. The fusion weights are calculated by using four algorithms: convex combination, optimal fusion, covariance intersection, and median fusion. The comparison results of the fusion algorithms in terms of estimation accuracy and computation cost are discussed.