MEMS accelerometer error compensation method based on LMS adaptive filtering and gradient descent

Abstract

The invention claims for protection of an MEMS accelerometer error compensation method based on LMS adaptive filtering and gradient descent. An MEMS accelerometer error compensation model with a measured value as an input and a true value as an output is established; random noises in actually measured data of the MEMS accelerometer at six positions are analyzed and processed by using an Allan variance and an LMS adaptive filtering algorithm; model parameters are trained by using all processed measurement data as samples; parameter prior values are calculated by using least squares to obtain aninitial value of batch gradient descent; training is carried out to obtain optimal fitting of samples to true model parameters; and error compensation is carried out on the MEMS accelerometer by using the model. The experiments show that the mean value error of output values of the accelerometer is reduced by two magnitude orders after error compensation and the standard deviation is reduced by one order of magnitude. The MEMS accelerometer error compensation method can be applied to high-precision calibration of MEMS accelerometers in the micro inertial measurement unit (MIMU), so that the measurement accuracy and output stability of the accelerometer are improved.

Description

technical field [0001] The invention belongs to the field of inertial navigation, in particular to an output error compensation method for a MEMS accelerometer, a core device of a micro-electromechanical system inertial measurement unit (MEMS-IMU). Background technique [0002] The inertial measurement unit (IMU) is the core unit of the inertial positioning system, and the accelerometer is one of the core components of the IMU, and the accuracy of its output value directly affects the positioning accuracy of the positioning system. MEMS accelerometers are widely used due to the advantages of microelectromechanical system (MEMS) sensors in terms of size, weight, cost, and power consumption. The MEMS accelerometer is affected by factors such as installation error, scale factor, zero bias and random noise, and its output value has certain errors. Therefore, compensating the output error of the MEMS accelerometer is a key step in the device-level optimization of the inertial po...

AI Technical Summary

Problems solved by technology

However, the mean value of the actual measurement data cannot fully reflect all the actual measurement data

In addition, the random noise in the actual measurement data is not processed, and the noise error will be left in the calibrated parameters

For the processing of random noise, Wang H et al. ) using wavelet analysis algorithm to denoise, Du Shaolin et al. Kalman filtering algorithm for noise reduction, none of which use the results of noise processing for the calibration of deterministic errors

Yin Hang et al. (Document: Yin Hang, Zhang Wei, Yuan Linfeng. A MEMS accelerometer error analysis and calibration method [J]. Sensor Technology Journal, 2014 (7): 866-869.) on deterministic error and random The noise is calibrated and processed, but it is actually carried out independently. Before the deterministic error is calibrated, the random noise in the actual measurement data is not processed, and the calibrated parameters are not the best fit for the real parameters.

Images