Fiber Optic Gyroscope Random Error Modeling Based on Improved Kalman Filtering

Abstract

A modeling method based on time series autoregressive model (AR) is proposed to address the problem of the existence of random errors in fiber optic gyroscopes, which affects their output accuracy. Based on the time series modeling requirements, the original data is first preprocessed, and then the autoregressive model order is determined according to the AIC criterion, and then the time series autoregressive model is established. On the basis of modeling, the traditional Kalman filtering algorithm and the improved adaptive filtering algorithm proposed in this paper are applied to filter the established signal model, and the filtering results are compared. Finally, the filtered noise coefficients are analyzed by using Allan's variance. The analysis results show that after Kalman filtering, the random error coefficient of the fiber optic gyro has been significantly reduced, which proves the correctness of the random drift model; after the improved adaptive filtering, the random error of the fiber optic gyro has been significantly reduced, which proves the correctness and applicability of the modified filtering method.