The Kalman filtering is easily affected by the gross error and it will cause larger distortion of the result. To overcome this problem, an adaptive robust Kalman filtering was proposed. It combines the advantages of adaptive Kalman filtering and robust Kalman filtering by using the adaptive factor and the robust factor. And the improved two segments Huber function was designed as the robust factor, the two to three times observation noise error was designed as the gross error determination standard respectively. Compared the result of Kalman filtering and adaptive robust Kalman filtering, the latter could effectively resist the influence of abnormal observation to the state estimation, and can improve the stability and availability of the filtering estimation significantly.