基于RBF神经网络的机抖激光陀螺零偏的温度补偿的研究

Research on temperature compensation for bias of mechanically dithered RLG based on RBF neural network

温度偏置漂移是存在于激光陀螺中使得输出信号产生较大偏置误差的一种不可忽略因素,准确地辨识漂移并有效地对其进行补偿直接关系到激光陀螺的测量精度。通过温度实验研究了机抖激光陀螺的温度特性,采用RBF网络进行温漂辨识。温漂辨识可以通过离线事先学习,因而在多种学习方法中选择了简单易行、精度高且运算速度快的正交最小二乘(OLS)法。通过实验验证,采用RBF网络及其OLS学习算法可以快速、有效、高精度地辨识并补偿温漂。

Temperature drift is a nonneglig ible factor causing big biasing error in RLG. How to identify and compensate this error relates directly to the measuremont accuracy. Based on temperature experiment, the temperature characteristic of mechanically dithered RLG has been studied. The RBF networks was applied to identify the temperature drift as it can achieve the global optimum evaluation and has the linear weight - combiner and fast learning. Of the different approaches of parameter learning, the OLS algorithm is prior by its simplicity, high accuracy and fast speed. The results show that the RBF network based method with the OLS learning offers a powerful and successful procedure for fitting and compensating the temperature drift.