Research Of High-precision Filtering Algorithm For MEMS Gyroscope

With the fast development of inertial navigation technology, MEMS gyroscope applications are increasingly widespread. The issue of how to improve the precision of MEMS gyro has become a hot research scholars at home and abroad. To solve this problem, based on the signal acquisition and processing system,this paper collected the output raw data of gyro and give error analysis for them, then built a mathematical model of gyro random drift using time series analysis methods. Finally, this paper studied and simulated the zero-drift error and dynamic error of gyroscope applying different filtering techniques, determined the final signal processing methods.First, this paper introduced the basic principle of signal acquisition system, and further analyzed its hardware structure and software composition, and then designed a signal acquisition system with STM32 as the core controller. Then this paper illustrated in detail the working principle of MEMS gyroscope, and analysed the zero adjustment mechanism. In addition, by determining the sampling frequency of gyro original signal collecting the original output data of the gyroscope, this paper identified the error term using the Allan variance analysis under MATLAB environment, and determined the coefficient of the error term.Secondly, based on the deep understanding of stationary random process, by preprocessing the output of the gyroscope data, this paper established the error model for gyroscope random drift adopting time series analysis method. At the same time, set order and estimated the parameter for its mathematical model.Thirdly, the paper analyzed the design, working principle and mathematical model of both traditional kalman filter and second-order low-pass filter, and then tested the statistical properties of the gyro output data, and determined the order number and parameters of error model after the experiments, chose the AR(1) model as its error model. Finally, this paper analyzed the static data of gyroscope using kalman filter and second-order low-pass filter respectively with MATLAB, identified stable zero of the traditional kalman filter by analyzing and comparing the filtering results.Finally, this paper analyzed the working principle and mathematical model of kalman filter with deterministic inputs, and then treated the gyro dynamic data with second-order low-pass filter, the analysis of the experimental results showed that although this method could reduce the errors to some extent, but there existed lag problems. So this paper designed a new algorithm, namely the extended kalman filter and second-order low-pass filter fusion algorithm, by putting the difference of the second order filter processing results as deterministic inputs of kalman filter, the experiment results showed that the algorithm could handle dynamic error of gyroscope.