Abstract:

In order to improve the operational efficiency of UKF (Unscented Kalman Filter), this paper analyzes the effect affected by the parameters in UKF, and then  proposes an optimization algorithm of UKF while the state transition matrix of the system is linear, and shows the correctness of the algorithm via the proof. Concerning the defect that the outliers affect the accuracy of UKF, this paper proposes a test method by using the innovation to judge whether the outliers exist or not. When there are  some outliers, the algorithm in this paper firstly gets rid of them, and then uses the least square method to estimate the current states based on the states obtained. However, this algorithm directly uses UKF when there is not any outlier, and finally it deduces the rationality of using the least square method to fit the estimates when there are  some outliers, so this paper proves that the method improves the ability of UKF to resist the outliers greatly. Finally, this paper presents a concrete simulative example to show the validity of the algorithm, which combines the least square method and UKF.

Key words: optimization algorithm;outliers;innovation;least square method