A Sliding Window Variational Outlier-Robust Kalman Filter based on Student's t Noise Modelling

Existing robust state estimation methods are generally unable to distinguish model uncertainties (state outliers) from measurement outliers as they only exploit the current measurement. In this paper, the measurements in a sliding window are therefore utilized to better distinguish them, and an adaptive method is embedded, leading to a sliding window variational outlier-robust Kalman filter based on Student's t noise modelling. Target tracking simulations and experiments show that the tracking accuracy and consistency of the proposed filter are superior to those of the existing state-of-the-art outlier-robust methods thanks to the improved ability to identify the outliers but at a cost of greater computational burden.