基于高斯过程回归的光纤陀螺温度漂移补偿

摘要

To improve the adaptability of a fiber optic gyro (FOG) working in the full temperature range,its temperature drift characteristic is analyzed based on the full temperature and angular rate test data and its compensation method is studied.A new compensation approach based on Gaussian process regression is proposed to deal with the complex characteristic of temperature drift.It maps the temperature and output of FOG to the angular rate of carrier directly by the optimization of hyperparameters,so traditional methods' two compenation erross caused by separate modeling of zero bias and scale factor can be avoided.The compensation precision is improved and the compensation process is simplified.The simulation results show that the model built by the new approach,in comparison with least square support vector regression,can reflect the temperature drift characteristic of FOG more accurately,and can obtain the better generalization ability and higher compensation precision with the predicted root mean square error less than 0.003(°)/s.Furthermore,it can reduce the influence of temperature on FOG output precision.%基于实测数据分析了全温度全速率下的陀螺漂移特性,提出了一种基于高斯过程回归的补偿新方法.该方法通过超参数训练直接建立了温度、陀螺输出和载体角速率之间的映射关系,弥补了传统方法对零偏和标度因数分别进行建模导致引入两次补偿误差的不足,在提高补偿精度的同时,简化了补偿步骤.仿真结果表明,相较于最小二乘支持向量回归方法,由高斯过程回归方法训练得到的模型能够更加准确地描述温度漂移特性,具有较高的预测补偿精度和良好的泛化能力,预测均方根误差小于0.003(°)/s,有效抑制了温度对光纤陀螺精度的影响.