Temperature Compensation for Fiber Optic Gyroscope Based on Bayesian Optimized : Support Vector Regression

In this paper, the temperature compensation method of Fiber Optic Gyroscope (FOG) based on Bayesian Optimization Support Vector Regression (BO-SVR) is proposed, which can improve the working accuracy of FOG in the temperature changing environment. According to the nonlinear characteristics of temperature drift with zero bias and hysteresis of temperature rise and fall, a piecewise Support Vector Regression (SVR) model of temperature rise and fall is constructed. In order to improve the modeling accuracy, the penalty parameter C and kernel function parameters of BO-SVR are introduced for optimization. The experiment of temperature compensation in the range of –400C to 600C is verified by medium precision FOG, and compared with the results of neural network compensation method, the results show that Bayesian optimization algorithm can reduce the time delay of temperature change for BO-SVR model.