Abstract: In the adverse application environment of micro air vehicle (MAV) with high dynamics and poor temperatures, a micro-electro-mechanical system (MEMS) gyro may exhibit very serious nonlinear scale factor error due to the coupling effect of temperature and input rotation, for which compensation by conventional method is no longer sufficient. By analyzing the mechanism of the scale factor error of MEMS gyro, a model with temperature and rotation factors is established and a compensation method is proposed for the coupled nonlinear scale factor error based on radial basis function (RBF) neural network. This method can deal successfully with the large error problem of the conventional method. The calibration and compensation test results show that using the new compensation method, in the temperature range of -10-+55 ℃ and the rotation rate range of -150-+150 (°)/s, the accuracy of MEMS gyro is increased by 7 times as compared with the conventional compensation method. And in the strong nonlinear error rotation rate range of -20-+20 (°)/s, the accuracy is increased by nearly 20 times. Through the tests, the effectiveness and superiority of this method are proved.

Key words: instrument and device, error compensation, neural network, scale factor, MEMS, micro air vehicle