Robust Kalman-type Filtering in Positioning Applications
Tommi A. Perälä · InTech eBooks · 2010
In this chapter robust filtering techniques for positioning using satellite measurements were presented. The Extended Kalman Filter was chosen as basis for robust filter design. Six filters were presented and tested in the simulations and using real GPS data. Based on the simulations the proposed filters seem to outperform EKF when blunder measurements occur, and do almost as well in normal cases. However, ABEKF does not seem to work well when using the Damped Hampel M-estimator, but performs better than EKF when using the Huber's M-estimator or the p-point M-estimator. The best performance was obtained using REKF with the Damped Hampel M-estimator, but other M-estimators also seemed to work almost as well. The tests using real GPS data showed similar trends in the results except that EKF performed very poorly. REKF with the Damped Hampel M-estimator performed best also when using real GPS data. Therefore, REKF should be taken into consideration when implementing positioning algorithms in mobile positioning devices due to its light computational and memory requirements and relatively high accuracy. The most promising approach was to use the Damped Hampel weight functions, but the performance degrades only a little when using any other influence and weight function proposed in this chapter.