Variational Bayesian Adaptive Kalman Filtering Algorithm for Measurement Data with Accuracy Category

An extended variational Bayesian adaptive Kalman filtering (EVB-AKF) algorithm is presented for the filtering of measurement data corrupted by noise with unknown variance but accuracy category information. The proposed algorithm modifies the posterior distribution parameter as follow: when the accuracy level of measurement data does not change or decreases, it is assigned to the weighted sum of the extrapolate approximation of the posterior distribution parameter in VB-AKF and the upper variance bound in the accuracy category information; when accuracy level increases, it is reset to the upper variance bound in the accuracy category information. Thus the problem that the first-order constant coefficient model of the posterior distribution parameter in VB-AKF cannot fully adapt to the measurement noise variance dynamics is solved. Simulation results show that the proposed algorithm can estimate the dynamic varying measurement noise variance effectively and efficiently, thus achieve an effective data filtering.