抗差与自适应组合的卡尔曼滤波算法在动态导航中的研究

针对观测信息不充足时,无法使用现有的一些抗差自适应滤波的问题,提出一种组合抗差滤波和自适应滤波的方法.该方法利用基于m估计实现的抗差滤波和基于新息向量马氏距离平方服从卡方分布而构造的自适应滤波,同时采用2次对检验统计量进行判别的方法,可以在单个历元实现在标准卡尔曼滤波、自适应卡尔曼滤波和抗差卡尔曼滤波之间选择一种当前时刻的最优滤波,因此,采用该方法也能构成抗差自适应卡尔曼滤波.仿真结果表明,在观测信息不足且滤波模型出现异常时,该方法能有效控制动力学模型误差和观测异常对导航解的影响,使导航解更能反映导航系统的真实情况.

Robust adaptive Kalman filter in kinematic positioning

The existing robust adaptive Kalman filter (RAKF) cannot be applied under the condition that observation information is insufficient. To solve this problem, a method which combines robust Kalman filter (RKF) based on M estimated and adaptive Kalman filter (AKF) based on the chi-square distribution of the square of the Mahalanobis distance of the innovation is proposed. A robust adaptive Kalman filter can be obtained by choosing among the adaptive, the robust, and the conventional Kalman filter (CKF) approaches through judging the test statistics twice at one epoch. The simulation results show that under insufficient observation information and abnormal innovations circumstances, the algorithm can also effectively control the disturbances of the kinematic model and gross error in observation during the navigation positioning and improve the performance of the filter.