基于交叉验证的陀螺仪温度漂移建模方法

根据k次交叉验证思想,分别在测量方差已知与未知的情况下引入了一种模型选择准则,该准则充分考虑了模型的拟合能力与预测能力,在获得极大似然估计的同时,避免了参数估计中的过拟合问题。利用这一准则,对陀螺仪误差系数温度漂移建模进行了分析,获得了温度漂移的最优多项式模型。并与传统的AIC准则、MDL准则进行对比,通过Monte Carlo仿真说明了该选择准则在样本较小的情况下具有更好的性质。

Temperature Drift Modeling of Gyroscope Using Cross-Validation

The paper introduces a new model selection criterion which based on the k-fold cross-validation under the unknown variance and known variance respectively.The new criterion can obtain maximum likelihood estimate of optimal model and at the same time avoid the large prediction error.Based on the criterion,the optimal polynomial model of gyroscope temperature drift is set up.The performance of the new criterion is compared with AIC and MDL through Monte Carlo numerical simulations.The result shows that the criterion has a more effective performance than AIC and MDL in small sample.