SINS/GNSS自适应反馈校正滤波

SINS/GNSS组合导航中，反馈闭环校正比开环输出具有更高的短时精度；但反馈校正的输出会有长时间缓慢发散迹象。传统工程上应用最优奇异值可观测度对反馈向量进行自适应调整，但最优奇异值计算量很大，且不同机动情形下的最优奇异值也不相同，很难确定。通过分析卡尔曼滤波的方差矩阵，提出一种新的可观测度定义，并利用滤波协方差矩阵确定新的可观测度，并结合人工神经网络确定反馈校正的自适应调整系数，仿真表明新方法抑制了反馈校正输出误差的长时缓慢发散，提升了导航精度。

The Research of Adaptive Feedback Kalman Filter for SINS/GNSS

Feedback Kalman filter can increase the short time accuracy of navigation system, although result in obvious error for long time. Traditionally, we can adjust the feedback vector by using the best singular value. Although , the computation, of computing the best singular value, is too large. And the best singular value changes while flight path changes. This article raises a new definition of observability of navigation system via analysising variance array of Kalman filter. Accompany with neural net ,we can ensure the adaptive coefficient for feedback Kalman filter, and more, decrease the error, at the same time, increase the accuracy of navigation system.