An extensive analysis for the use of back propagation neural networks to perform the calibration of MEMS gyro bias thermal drift

In recent years, the improvement of advanced micro-fabrication techniques has allowed the development of Micro Electro-Mechanical Systems (MEMS) inertial sensors. These sensors have the advantages of small volume, light-weight, high reliability and low-cost, so they result as the most common sensors used to perform the flight attitude calculation for Unmanned Aerial Systems (UASs). Even if they are small size and light-weight sensors, they suffer more than other higher grade gyros for some types of errors such as turn-on to turn-on bias, in-run bias, bias drift, scale factor drift and other environment dependent errors. In particular, the performance of MEMS inertial sensors is greatly affected by temperature variations, due to the sensitivity of silicon's material properties and gyro's packaging and electronics to temperature. This paper investigates the calibration of temperature effects on MEMS gyroscope bias drift, proposing an innovative calibration method, based on the use of Back-Propagation (BP) Neural Networks to model the nonlinear relationship between MEMS gyroscope null-voltage and temperature. BP Neural Networks have the advantages of nonlinear fitting, regardless of the mathematical model of the sensor and various non-linear factors. First of all, the reference model for bias trend vs. temperature is reported. Subsequently, the proposed innovative calibration method is compared with the traditional polynomial fitting technique. Finally, the performance of the proposed technique is discussed.