A fast suboptimal information fusion scheme for federated filter

This paper develops a new information fusion scheme based on diagonal matrix weighting for the federated filter instead of covariance matrix weighting. In the global combination process, only the computation of each scalar component of the weighting diagonal matrixes is needed, so the computation of the inverse of the covariance matrixes can be avoided. The key to this method is that the computation load is largely reduced and the master filter can run at the same rate as the local filters do, combining outputs of the local filters, and almost not having to increase the additional computation burden. In addition, The accuracy of the global fusion estimate is higher than that of any local estimate. Therefore, it is highly advantageous to real-time applications. Simulation for a radar tracking system shows the effectiveness of this method.