顾及有色噪声的光纤陀螺随机噪声自适应滤波方法

作为光纤陀螺误差的重要组成部分，随机噪声严重影响着光纤陀螺的精度，对光纤陀螺随机噪声进行准确建模和补偿是提升陀螺精度的有效方式。本文针对光纤陀螺随机噪声的复杂性，难以对其进行精确分析，ARIMA （auto-regressive moving average）模型Kalman滤波中有色噪声不能使用状态扩充法建模的问题，扩展了Harvey方程，实现有色噪声白化。同时，考虑先验噪声的不确定性以及模型参数在线更新导致的参数与状态噪声相互耦合，分析了动态Allan方差估计量测噪声的不足，使用VBAKF （variational Bayesian adaptive Kalman filter）实时修正滤波状态噪声与量测噪声。试验表明，Harvey法较传统滤波建模方式，随机噪声序列方差降低40%，Harvey法结合VBAKF使序列方差降低了54%；VBAKF较动态Allan方差，可以更好地估计量测噪声。结果表明，此方法可有效抑制随机噪声Kalman滤波中有色噪声和随机模型不准确的影响，提高随机误差补偿精度。

Adaptive Kalman filter method with colored noise for fiber optic gyroscope random drift

Random noise reduces the accuracy of output seriously as an important part of fiber optic gyroscope(FOG)error.Accurate modeling and compensation of random noise is an effective way to improve the accuracy of FOG.To solve the problem that FOG random noise is complicated and to accurately analyze difficultly,and the colored noise in the ARIMA model is modeled as the state equation by using the state expansion method,the Harvey algorithm is reconstruct to whiten colored noise.At the same time,considering the uncertainty of priori noise and the coupling between states and noise caused by online update of ARIMA model,variational Bayesian adaptive filter(VBAKF)is used to correct the state and measurement noise.Experiments show that the Harvey method reduces the random noise sequence variance by 40%compared with the traditional filtering modeling method.The Harvey method combined with VBAKF reduces the sequence variance by 54%;VBAKF can better estimate the measurement noise than the dynamic Allan variance.Method in this paper can effectively suppress the effects of colored noise and random model inaccuracy in the random noise Kalman filter,and improve the accuracy of random error compensation.