中文摘要：

光纤陀螺随机漂移误差是影响航空矢量重力测量系统姿态解算精度的关键因素。建立模型并在输出中对其补偿是抑制该项误差的有效方法。针对传统ARMA模型只能对平稳随机漂移误差建模，且模型无法满足实时滤波需求的问题，本文引入适用于非平稳随机漂移误差的ARIMA模型，同时给出详细的建模过程，并提出采用实时平均算法消除原始采样序列中常值分量的思路，实现了随机漂移误差的实时Kalman滤波估计。基于本文所提出的模型和实时滤波算法，对光纤陀螺实测数据进行分析，结果表明处理后信号中随机漂移误差的方差减小了46.5%。Allan方差分析结果表明，滤波后角度随机游走系数和角速率随机游走系数分别降低了约50%和40%。本文的结果说明ARIMA模型能够准确描述陀螺的非平稳随机漂移误差。基于实时平均算法的Kalman滤波可实现随机漂移误差的在线估计，有望提高航空矢量重力测量系统的姿态解算精度。

英文摘要：

Random drift error of fiber optic gyroscope is the crucial factor that influences the calculation accuracy of the attitude of airborne vector gravimetry. Modeling and compensating it can restrain this type of error significantly. Given the problem that traditional ARMA model can be only applied in the case of stable random drift, which cannot meet the need of real-time filtering, an ARIMA model （autoregressive integrated moving average） which is suitable for non-stable random drift is introduced along with the detailed procedure in this paper. The algorithm that can eliminate the constant component of original sampling sequence with real-time averaging method is also proposed as well as the real-time Kalman filtering estimation of the random drift. With the methods proposed above, the variance of random drift can be reduced by 46.5%. The analysis of Allan variance suggests that the coefficients of random drift for angle and angular speed have decreased about 50% and 40%, respectively. The results showed that non-stable random drift can be accurately characterized by ARIMA model and that online estimation of random drift can be realized by real-time average algorithm, indicating the potential to improve the calculation accuracy of the attitude of airborne vector gravimetry.