Performance analysis of federated filter for SAR/TRN/GPS/INS integration

Abstract

Decentralized integration of global positioning system (GPS), synthetic aperture radar (SAR), and terrain referenced navigation (TRN) system into an inertial navigation system (INS) is analysed in comparison to the centralized integration method in this paper. Many publications have shown different decentralized integration methods. The main advantage of the decentralized filter architecture is flexibility and modularity. Especially the federated no-reset filter has received considerable attention in literature because of improved fault detection and isolation (FDI) capability, which is an important feature in multi-sensor navigation systems. Fault detection capability of the federated no-reset filter is well-established. However, this filter type combines sensor data in a suboptimal way and a loss of accuracy compared to the centralized Kalman filter is generally observed. Other publications mentioned this loss of accuracy, but did not give the amount of accuracy loss for SAR/TRN/GPS/INS systems. In this paper additionally a slight modification of the federated filter is presented. Typically a GPS/INS filter tends to be overoptimistic due to long-term timecorrelated atmospheric errors of the GPS pseudo-range measurements. A rather simple modification of the federated filter is given to take into account this overoptimistic behaviour. A covariance correction yields a more reasonable covariance of the GPS/INS local filter and an improved overall position estimation of the federated filter.