光纤陀螺随机漂移的实时滤波方法研究

Real-Time Filtering Methods of Random Drift of Fiber Optic Gyroscope

在光纤陀螺捷联惯导系统的初始对准过程中,光纤陀螺的随机漂移是影响对准精度的重要因素。通过离线的建模和滤波,能够在一定程度上抑制光纤陀螺随机漂移的影响,但由于受环境因素及光纤陀螺重复启动性能的影响,离线建立的模型通常不具备普适性,无法实现初始对准中随机漂移的在线滤波。为了解决这一问题,论文研究了随机漂移的实时滤波方法,包括基于ARMA模型的Kalman实时滤波方法和基于滑动数据窗的小波实时滤波方法,并对两种方法进行了改进。最后,进行了光纤陀螺捷联惯导系统的初始对准试验,研究了两种滤波方法对对准精度的影响,试验结果表明两种在线滤波方法均能够在较大程度上提高初始对准的精度,而且小波实时滤波方法的精度和实时性均优于基于ARMA模型的Kalman实时滤波方法。

During the initial alignment of Fiber Optic Gyroscope Strapdown Inertial Navigation System（FSINS）,the random drift of FOG has a great influence on the alignment accuracy of FOG.Through off-line modeling and filtering,the random drift of FOG can be depressed partly.However,the off-line models are usually not universal for environmental change and FOG＇s instability,and the models can not be used for real-time filtering in the initial alignment of FSINS.To cope with this problem,several real-time filtering methods,including the kalman real-time filtering method based on ARMA modeling,the wavelet real-time filtering method based on wavelet thresholding and their modification are studied in the paper.In order to verify the performance of real-time filtering methods,experiments are carried out for the initial alignment of FSINS,it is proved that both proposed filtering methods can improve the alignment accuracy obviously,and the accuracy and the real-time performance of the wavelet real-time filtering method are better than the kalman real-time filtering.