Compensation of MEMS Gyroscope Random Error Based on Square- Root Risk-Sensitive Unscented Kalman Filter

Abstract

In order to compensate MEMS gyroscope random error, a new method employing the squareroot risk-sensitive unscented Kalman filter (SR-RSUKF) and a nonlinear model is proposed. The nonlinear model based on ARIMA takes model parameters as states, and thus realizes the online model estimation. The SR-RSUKF deals with non-additive noise items through augmented state vector, and employs a square root algorithm to get well numerical stability, and improves the flexibility by extending scalar risk parameters to a risk sensitive matrix. In experiments, the raw sample data is processed with three methods using different models and filters. And the results show that the SR-RSUKF together with the nonlinear model provide a competent solution to compensate MEMS gyroscope random error.