硅微陀螺信号的自适应UKF滤波处理

An adaptive UKF algorithm for silicon micro-machined Gyroscopes

针对硅微陀螺零偏重复性差和漂移信号中非线性因素的存在,提出利用AR模型系数实时递推方法对漂移信号进行AR建模,并建立相应的状态方程和量测方程;基于方差膨胀原则的自适应无迹Kalman滤波(AUKF)方法对漂移信号进行处理,采用简化的AUKF滤波过程提高漂移信号滤波的实时性。静态信号和动态测试处理结果表明,简化的AUKF算法效果明显优于UKF滤波和Kalman滤波(KF),滤波后硅微陀螺零偏稳定性提高倍数分别是UKF和KF的3倍和2倍,动态信号滤波后误差减少倍数分别是UKF和KF的1.46和1.34,信号均值不变,但AUKF需要较多的信号处理时间。

Considering that the silicon micro-machined gyroscope bias is often non-repeatable and is easily affected by some nonlinear factors, the paper suggests that the bias is AR modeled by a really-time recursive method, the state equation and measurement equation are built, and an adaptive unscented Kalman filtering （AUKF） is devised to filter the gyroscope bias by means of the variance inflation. Based on the linear measuring procedure, a simplified UKF algorithm is used to really-time improve the procession. The simulation from the gyroscope static bias and the dynamic measurement shows that the simplified AUKF algorithm is evidently superior to the UKF and the Kalman filtering （KF）. The simplified AUKF can increase the bias stability 3 times and 2 times of the KF and the UKF. The decreased error between the filtered and unfiltered dynamic measurement is 1.46 and 1.34 times of the UKF and KF. The mean of the dynamic measurement does not change. Compared with the KF and the UKF, the AUKF occupies the most processing time.