多准则MEMS陀螺随机误差在线建模与实时滤波

针对随机误差相关性较弱的MEMS陀螺仪,提出采用多准则曲线方法辨识其带有截距项的随机误差时间序列模型.采用该模型可直接对MEMS陀螺仪的实测量数据进行在线建模,而无需零均值化离线处理.基于该模型并采用状态扩增的方法设计卡尔曼滤波器,实现了MEMS陀螺仪随机误差的实时滤波.实验结果表明,针对某MEMS陀螺仪带有截距项的AR(2)模型可以作为其随机误差模型,经过在线建模和实时滤波后,MEMS陀螺仪随机误差的标准差降低了50%,有效抑制了MEMS陀螺仪的随机误差.

The Multi-criteria on-line modeling and filter for random error of MEMS gyroscope

As the random error of some MEMS gyroscope had weak correlation,the Multi-criteria curve method was introduced to identify the time-sequence model with a intercept. Using this model,the random error model of the MEMS gyroscope could be established on line without zero mean offline processing. Based on this model,an aug-menting state vector was used to design the Kalman filter which had been used to filter real-timely. The results of the experiment showed that AR(2)model with a intercept could be used as a good MEMS gyroscope random error model. After the on-line modeling and real-time filtering,50% of the standard deviation of the random error had been reduced and the random error of MEMS gyroscope was effectively restrained.