Allan variance method for gyro noise analysis using weighted least square algorithm

The Allan variance method is an effective way of analyzing gyro's stochastic noises. In the traditional implementation, the ordinary least square algorithm is utilized to estimate the coefficients of gyro noises. However, the different accuracy of Allan variance values violates the prerequisite of the ordinary least square algorithm. In this study, a weighted least square algorithm is proposed to address this issue. The new algorithm normalizes the accuracy of the Allan variance values by weighting them according to their relative quantitative relationship. As a result, the problem associated with the traditional implementation can be solved. In order to demonstrate the effectiveness of the proposed algorithm, gyro simulations are carried out based on the various stochastic characteristics of SRS2000, VG951 and CRG20, which are three different-grade gyros. Different least square algorithms (traditional and this proposed method) are applied to estimate the coefficients of gyro noises. The estimation results demonstrate that the proposed algorithm outperforms the traditional algorithm, in terms of the accuracy and stability.