Modeling of MIMU sensors random error using wavelet transform

MEMS inertial sensors have large random noises and errors. In order to compensate them, a proper random error model must be established. In this paper, the random error of Analog Devices Incorporation's MIMU--ADIS16355 is analyzed using an Allan variance method. The analyses show that MEMS inertial sensor's random errors can be divided to two parts: white noise and colored noise. Then a compound modeling method for random errors of MEMS inertial sensor is put forward which adopts both the wavelet transform-based noise separation preprocessing method and AR time series modeling algorithm. The random errors of the ADIS16355 are modeled using the proposed compound modeling method. The simulation shows that, compared with traditional approximate one-order Gauss-Markov random error model, the compound modeling method can effectively filter out the effects of white noise and obtain more accurate random error model for MEMS inertial sensor.