Self-adaptive prediction and compensation model for high-accuracy MEMS gyro's startup drift

Abstract

Inertial navigation systems for most actual applications require rapid reaction from cold startup to “Navigation Ready” status, especially for MEMS gyroscope with bias drift less than 5°/h. During the startup phase, angular velocity measurement precision is greatly affected by startup errors. To compensate the startup drift curves of MEMS gyroscopes to satisfy the fast startup situation, it is necessary to forecast the trend of them. This paper presents a new automatic prediction and compensation frame based on classification and recognition algorithms of curves with knowledge of Support Vector Machine and χ2 statistic. First, using this frame, startup drift of high-accuracy MEMS gyroscope is automatically classified and recognized. Then corresponding exponential or polynomial prediction model is chosen for the classified startup drift. Experiments proved that with this self-adaptive frame, MEMS gyroscopes can enter the high precision working state in about 100 seconds after cold startup.