Simplified Federated Filtering Algorithm With Different States in Local Filters

Abstract

In this paper, to reduce the computation load of federated Kalman filters, a simplified federated filtering algorithm for integrated navigation systems is presented. It has been known that the per-cycle computation load grows roughly in proportion to the number of states and measurements for a single centralized Kalman filter. Hence, the states that have poor estimation accuracies are removed from local filters, so that the per-cycle computation load is reduced accordingly. Local filters and master filter of the federated Kalman filter may have different states, so the transition matrices are required to combine the outputs from the local filters and the master filter properly and to reset the global solution into the local filters and the master filter correctly. An experiment demonstrates that the proposed algorithm effectively reduces the computation load, compared with the standard federated Kalman filtering algorithm.

Adaptive weighted federated Kalman filtering based on Mahalanobis distance and its application in navigation

“…The basic idea of FKF is to disperse first and then fusion globally. In the federated filter, the subsystem includes sensor 1, sensor 2, …, sensor n, and Kalman filter corresponds to different sensors respectively, forming multiple local filters, each local filter works in parallel, and the global filter is used for information synthesis and sequential processing, and the filtering results generated by all local outputs are fused to give the global optimal state estimation [15][16][17].…”