Fusion Kalman and Weighted UFIR State Estimator With Improved Accuracy

Abstract

In this article, estimates of the Kalman filter (KF) and weighted unbiased finite impulse response (UFIR) filter are fused in discrete time-varying state-space to improve the robustness in uncertain environments associated with industrial applications. The weighted UFIR filter is derived using the Frobenius norm and termed as Frobenius finite impulse response (FFIR) filter. It is confirmed that the FFIR filter has better performance under the uncertainties and errors in the noise statistics, while the KF filter is best when the model and noise are exactly known. Based on a numerical example of a hover system, we show that the FFIR filter is able to outperform the UFIR filter and that the fusion KF/FFIR filter is able to outperform both of them. An experimental verification provided for the drone velocity estimation under the hover operation conditions has proved a better accuracy and robustness of the proposed fusion KF/FFIR filter.