An Improved Robust Adaptive Kalman Filtering Algorithm

Abstract

Robust adaptive Kalman filter (RAKF) is a method that can not only compensate for the model errors but also resist the measurement outliers. The method reduces the influence of the measurement outliers on the filtering system by reducing the weight of the abnormal observation. After the processing of robust estimation, all the errors are attributed to the inaccuracy of the model. Since the robust estimation is not theoretically quantitative that it can only reduce the observation error to a certain extent, the remaining observation errors are compensated as the model errors, resulting in new estimation errors. To address this problem, a method for detecting the model error is proposed as a trigger condition for RAKF to compensate for the model error, and this constitutes an improved robust adaptive Kalman filter (IRAKF) algorithm. The new algorithm performs the model errors compensation struand then adjust the adaptive factor cter than the original method. The efficacy of the improved algorithm is demonstrated via a traffic running testing with GPS/INS tight integration navigation system. Simulation results indicate that the improved algorithm can effectively reduce the inaccurate adaptive procedure and obtain relatively accurate and stable filtering results.