An Improved Kalman Filter Based on Sequentially Estimating Measurement Noise Covariance

Abstract

It is well-known that the covariance matrix of measurement noise plays an important role in Kalman filter design. Based on recursive covariance estimation, an improved Kalman filter, termed as IKF-RCE, is proposed in this paper to solve the challenging problem of state estimation when the covariance matrix of measurement noise is completely unknown. Finally, the stability of IKF-RCE algorithm is verified by simulation studies to demonstrate that the state estimation obtained from the new algorithm is asymptotically consistent with the optimal estimation obtained from the ideal Kalman filter with exact measurement noise covariance matrix.