Application of time series modeling in initial alignment of strapdown fiber optic Gyro (FOG) compass system

Fiber Optic Gyro (FOG) random errors seriously affect the initial alignment accuracy of strapdown FOG compass system. In order to minimize FOG random error, a method for time sequence modeling of FOG random drift was presented, and Kalman filter was built. In the process of modeling, statistical test and corresponding pretreatment of FOG drift was essential, as the estimation procedures were available only for stable, normal and zero-mean series. Model was selected by judging the characteristic of "tail off" and "cut off", and the order was determined by final prediction error (FPE) criterion. Burg algorithm based on Levinson constraints was adopted to solve the model parameters. Based on Auto Regressive Moving Average (ARMA) model, system equations and observations were established, and Kalman Filter was carried out. The experimental result shows that Kalman Filter, which bases on AR (3) model, can effectively eliminate random errors. Not only the method improves the alignment accuracy, but also shortened the alignment time.