Kalman Filter Outlier Detection Methods Based on M-estimation

Aiming at the outlier influencing the data processing and analyzing in the observation sequence of measurement and control fields, the precision of data processing will be decline or radiation by using the standard Kalman filter. So a method on Kalman Filter Outlier Detection Methods Based on M -estimation was proposed. Firstly, the prediction data of next time was estimated by using the adaptive recursive M-estimation; and the real measured data and the estimation data; then the remaining error was into the Kalman equation. Thirdly, whether the outlier was existed was judged by using the measured data. The results and simulation shows that the methods was robustness, and had the immunity to not only the single outlier but also a series of outliers no more than 15; and the threshold δ was not initialized; the precision of outlier detection could be enhanced.