Application of Akaike information criterion in selecting random error model for inertial measurement unit
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
万方数据
针对当前基于目视 Allan 方差曲线进行惯性元件随机误差建模中存在随机误差类别识别困难,人工干预较多的问题,提出了一种应用赤池信息量准则(AIC)优选惯性元件随机误差模型的方法.首先,采用加权最小二乘拟合法对不同间隔的 Allan 方差数据进行合理的加权,并将随机误差系数用指数形式表示作为待估系数,避免了随机误差系数估计值为负的问题.其次,采用AIC对模型中包含的随机误差类型进行优选.最后,基于光纤陀螺仪测量数据对所提方法进行了实验验证,实验结果表明,所提方法能根据AIC的值自动识别出被测光纤陀螺的随机误差类别为:角度随机游走、零偏不稳定性和角速度随机游走.所选模型模拟的功率谱密度与实测数据的功率谱密度吻合度较好.验证了所提方法的有效性.
Aiming at the problems of incorrect recognition of noise class and more manual intervention in random error modeling of inertial components based on visual Allan variance curve,a method to select the random error models of inertial components using Akaike information criterion(AIC)is proposed.Firstly,the Allan variance data at different intervals are reasonably weighted by the weighted least squares fitting method,and the logarithm of the random error coefficient is used as the coefficient to be estimated,which avoids the problem that the estimated value of the random error coefficient is negative.Secondly,the AIC is used to select the random error types contained in the model.Finally,the proposed method is verified by the measurement data of fiber optic gyroscope.The experimental results show that the proposed method can automatically identify the random error categories of the measured fiber optic gyro according to the AIC value:angle random walk,bias instability and angular velocity random walk.The power spectral density simulated by the selected model matches well with the power spectral density of the measured data.The effectiveness of the proposed method is verified.
inertial measurement unitstochastic error modelAkaike information criterionparameter fitting
NETL
NSTL
万方数据
