Invariant filtering versus other robust filtering methods applied to integrated navigation
Erickson Robert, T. Perrot · 2017
Although the Extended Kalman Filter (EKF) achieves satisfactory estimations in general, inconsistencies can appear when the filter is used in adverse conditions. After explaining theoretically why such inconsistencies appear, three filters are presented: the Schmidt-Kalman Filter (SKF), the Observability-Constrained EKF (OCEKF) and the Invariant Filter (IF). These techniques, designed to preserve as much as possible true unobservable directions in the estimated system, are supposed to bring more robustness in the estimation. These three filters are operational (SKF and OCEKF) or being implemented (IF) in different Safran INS. Actual results are presented to confirm the theoretical benefits of these more advanced techniques in practice.