Analysis and Improvement of Multi-Scaled Unscented Kalman-filters to Increase Accuracy
József Kuti, Péter Galambos · 2022
The paper analyses the so-called Unscented Kalman-filter with a set of scaling parameters and points out its limitations. Furthermore, it defines additional parameters that can highly improve the efficiency of the method. A comprehensive numerical simulation shows the effect of these parameters in computational time and accuracy. With appropriately chosen parameters the RMSE is reduced to 15% of the original value.