基于岭回归径向基神经网络的MIMU误差建模

对MEMS（Micro Electro Mechanical systems微电子机械系统）IMU（Inertial Measurement Unit惯性测量单元）的随机漂移和误差进行建模和补偿是提高捷联惯导系统性能的主要方法之一。为了避免径向基（RBF）神经网络输出之间可能存在的复共线性,更加准确有效的对MIMU的输出误差建模,提出将岭回归径向基神经网络用于MIMU的误差建模中。通过与时间序列建模方法的仿真比较表明,在对MIMU误差补偿上所建立的岭回归RBF网络与四阶AR模型方法精度相当,比一阶AR模型精度高,而且无需对数据平稳性处理。

Abstract：

It is one of the main methods to improve the performance of Strap-down Inertial Navigation System for compensating the random drift and bias of MEMS （Micro Electro Mechanical systems） IMU （Inertial Measurement Unit）. In order to eliminate latent multicollinearity of radial basis function neuron network output layer and model the drift and bias of MIMU accurately,the radial basis function neuron network based on ridge regression method was proposed which was applied in modeling and compensating MIMU errors. The simulation shows,compared to the AR model,the precision of compensation of MIMU error using radial basis function neuron network based on ridge regression method is equal to the fourth order AR model,better than first order AR model,and no data stabilization processing.

Research on MIMU Error Modeling Based on Ridge Regression Radial Basis Function Neuron Network

It is one of the main methods to improve the performance of Strap-down Inertial Navigation System for compensating the random drift and bias of MEMS （Micro Electro Mechanical systems） IMU （Inertial Measurement Unit）. In order to eliminate latent multicollinearity of radial basis function neuron network output layer and model the drift and bias of MIMU accurately,the radial basis function neuron network based on ridge regression method was proposed which was applied in modeling and compensating MIMU errors. The simulation shows,compared to the AR model,the precision of compensation of MIMU error using radial basis function neuron network based on ridge regression method is equal to the fourth order AR model,better than first order AR model,and no data stabilization processing.