Optimized Kalman Filter Approach with Innovation-based Outlier Diagnosis

Due to the statistical property of measurement noise varying from time and outliers in engineering applications, the standard Kalman filter is oscillating or even divergent. To solve this problem, a new optimal method is proposed. The measurement covariance is estimated more precisely in time by a replacement of a posteriori covariance at last step with a priori covariance which contains more current information. A novel three-segment function allowing to simultaneously restrain the outliers and tune the a posteriori covariance is presented. The experimental results show that the proposed method outperforms the common robust adaptive filter.