Sage-Husa随机加权无迹卡尔曼滤波及其在导航中的应用

Sage-Husa Random Weighting UKF and Its Application in Navigation

在吸收Sage-Husa滤波和无迹卡尔曼滤波优点的基础上,利用随机加权估计算法将传统的定义在线性系统上的Sage-Husa噪声估计器推广到非线性系统中,提出一种非线性Sage-Husa随机加权无迹卡尔曼滤波算法。该算法首先利用Sage滤波的开窗平滑方法求得观测残差向量和新息(预测残差)向量的协方差阵;然后用随机加权自适应因子对观测残差和预测残差进行调节;最后对状态预报向量的协方差矩阵进行自适应随机加权估计,以控制观测残差和预测残差对导航精度的影响。计算结果表明,提出的非线性Sage-Husa随机加权无迹卡尔曼滤波算法,滤波精度明显优于无迹卡尔曼滤波和自适应无迹卡尔曼滤波算法,能够提高组合导航的解算精度。

On the basis of absorbing the advantage of Sage-Husa filtering and unscented Kalman filtering, this paper uses random weighting estimation to extend the traditional Sage-Husa noise estimation defined in linear system to nonlinear system and a nonlinear Sage-Husa random weighting unscented Kalman filtering method is presented. This method adopts Sage filtering window smoothing method to obtain the eovariance matrix of observation residual vector and innovation （prediction residual） vector. Then, observation residual and prediction residual are adjusted by using random weighting adaptive factor. Finally, the covariance matrix of state prediction vector is estimated by using adaptive random weighting estimation to control the effect of observation residual and prediction residual on navigation accuracy. Experimental results demonstrate that the proposed Sage-Husa random weighting UKF is obviously better than unscented Kalman filtering and adaptive unscented Kalman filtering and the proposed method can improve the accuracy of integrated navigation.