An accurate modeling method for random drift of MEMS gyro

To compensate random drift in MEMS gyros, an auto-regressive moving average (ARMA) model for measured data drift was developed using time series analysis, and a new estimation method was proposed for moving-average (MA) models. The gyro noise was modeled as an ARMA with the observation noise. After the auto-regressive (AR) parameters were estimated, a more accurate estimation with a smaller variance of the MA autocovariance sequence was deduced for the residual noise by the AR filtering. The statistics were used as the input of Gevers-Wouters (GW) method to estimate MA parameters. The results of simulation prove that both the accuracy and reliability of parameter estimation are improved. The compensation experiment of MEMS gyros random drift further verifies that the proposed method is more accurate than the traditional one.