A Signal Processing Technique for Improving the Accuracy of MEMS Inertial Sensors

Navigation, guidance and control for small space vehicles require inertial measurement sensors which are small, inexpensive, low power, reliable and accurate. Micro inertial sensors, such as MEMS gyroscopes, can provide small, inexpensive, low power devices; however, the accuracy of these devices is insufficient for many space applications. Signal processing methods can be used to provide the necessary accuracy. The individual outputs of many nominally identical micro sensors can be combined to generate a single accurate measurement. An extended Kalman filter (EKF) (A.R. Stubberud and X.-H. Yu, 2000) which includes the dynamics of every sensor can be used for such a combination; however, the 'curse of dimensionality' limits the number of sensors which can be used. In this paper, a new EKF technique for combining many sensors is proposed which, using a common nominal model for the micro sensors and a single EKF with the state dimension of a single sensor, has accuracy comparable to the high dimensional EKF and is significantly more accurate than a single sensor.