Modeling the Random Drift of Micro-Machined Gyroscope with Neural Network

Abstract

In this paper a new combined method was applied to model the random drift of a micro-electro-mechanical system (MEMS) gyro to enhance its performance. The gyro is used to set up a micro-inertial -measurement unit (MIMU) for its low cost, low power consumption and small dimensions. To improve the MIMU’s performance, we model the gyro’s random drift by a statistic method. Given the paucity of the knowledge of fabrication of the gyro, we select a neural network model instead of making a delicate physical-mathematical model. Since the gyro we used is a tuning fork micro-machined sensor with large random drift, the modeling performance is affected by the randomness inherent in the output data when neural network approach is applied. Therefore, radial basis network structure, which was successfully applied to model temperature drift of fiber optical gyros, was chosen to build the model and the grey neural network. Compared with autoregressive model, the standard error of the gyro’s random drift is reduced dramatically by radial basis model and grey radial basis model.