对光纤陀螺随机噪声的ARMA建模及卡尔曼滤波方法进行了研究。针对ARMA(Auto-Regressive and Moving Average自回归滑动平均)模型的有色噪声在状态方程中不能通过传统的状态扩充法进行白化的问题,提出了新的噪声白化方法:采用增广最小二乘法估计ARMA模型的参数,同时提取出ARMA模型中的驱动白噪声,从而可以把ARMA模型中的有色噪声项作为控制项放入系统的状态方程,通过Sage-Husa的次优无偏MAP(Maximum A Posteriori,极大后验)噪声统计估值器对系统噪声的统计特性进行估计,实现了系统噪声的白化。在此基础上应用自适应卡尔曼滤波,有效消除了误差,得到状态值的准确估计。实验结果表明,对于随机噪声的自相关和互相关特性均呈现拖尾性质的光纤陀螺,采用新方法比传统基于AR模型的Kalman滤波降噪方法滤除噪声的效果提高了10%以上。
Traditional coning algorithms are based on the first-order coning correction reference model.Usually they reduce the algorithm error of coning axis(z)by increasing the sample numbers in one iteration interval.But the increase of sample numbers requires the faster output rates of sensors.Therefore,the algorithms are often limited in practical use.Moreover,the noncommutivity error of rotation usually exists on all three axes and the increase of sample numbers has little positive effect on reducing the algorithm errors of orthogonal axes(x,y).Considering the errors of orthogonal axes cannot be neglected in the high-precision applications,a coning algorithm with an additional second-order coning correction term is developed to further improve the performance of coning algorithm.Compared with the traditional algorithms,the new second-order coning algorithm can effectively reduce the algorithm error without increasing the sample numbers.Theoretical analyses validate that in a coning environment with low frequency,the new algorithm has the better performance than the traditional time-series and frequency-series coning algorithms,while in a maneuver environment the new algorithm has the same order accuracy as the traditional time-series and frequency-series algorithms.Finally,the practical feasibility of the new coning algorithm is demonstrated by digital simulations and practical turntable tests.
