Abstract:

In view of poor accuracy and divergent of the traditional Kalman filter (KF) algorithm when there exits model errors and noise, a new multiple fading factors KF is proposed, which is based on the innovation covariance estimator of the fading memory index weighting to calculate the innovation covariance estimator. The new algorithm can adjust prediction error covariance matrix by multiple fading factors, so that each filter channel has different regulatory capacity, which improves the accuracy and robustness of the filter algorithm even if there exits poor tracking between the single fading factor and multivariate. The simulation and experiment results indicate that the new algorithm can effectively suppress the filter divergence, compared with the conventional KF and single fading factor KF, and the filter accuracy and robustness are improved, which can better meet the requirements of engineering applications.