Adaptive Kalman filter based on improved second order mutual difference estimation

Abstract

In this paper, a method to compute noise variance and adapt measurement noise covariance matrix R in Kalman filter is proposed. We construct a virtual redundant measurement using α-β-γ filter to apply the second order mutual difference estimation method, which estimate noise variance effectively, in single measurement to calculate noise variance. And statistical data selection algorithm is proposed to avoid inaccuracy caused by lag in the α-β-γ filter. Simulations indicate this method is effective in R adaption with relatively low computation.