Temperature Drift Error Compensation for MEMS Gyroscopes Based on SVR

Due to the susceptibility of MEMS devices to temperature variations owing to their micro-mechanical structures, it becomes imperative to perform temperature drift correction to enhance the accuracy of measurement outcomes. Translate to English: When analyzing real experimental data, by contrasting the zero-offset outputs of MEMS at the same temperature and under different temperature variations, it is deduced that the zero-offset output of MEMS is associated with the temperature slope. This study proposes the utilization of input units that encompass temperature variation rate information in lieu of single-valued inputs, and employs the Support Vector Regression (SVR) method to model and forecast the static zero-offset output as a function of temperature variations. Experimental results demonstrate a substantial reduction in the mean squared error (MSE) on the testing set when compared to scenarios where temperature variation rate is not considered, with the MSE decreasing by orders of magnitude.