Combining numerous gyroscopes for accuracy improvement using autoregressive process for rate signal modeling

A signal processing technique for gyro array is presented in this paper to reduce noise and improve the accuracy of micro-electro-mechanical system (MEMS) gyroscope. To improve the dynamic performance, the true rate is modeled by an autoregressive (AR) model instead of a random walk. According to the analysis results of the Allan variance, the measurement model is simplified and a mathematical model is established. Based on this model, a novel Kalman filter (KF) with modified estimation process for combining outputs of a gyroscope array is designed. The performance and the affect factors of the gyro array are analyzed by using a steady-state covariance. The experimental results indicate that the RMSE of the gyroscopes can be reduced from 0.4925 deg/s to 0.0034 deg/s by an array composed of six gyroscopes in static test. The dynamic test is also discussed and the validity of the proposed modeling and fusion method is proved.