A comparison between extended and unscented Kalman filter in a realistic sensor model

Purpose This study aims to compare the performances of extended and unscented Kalman filters in the presence of noise and errors from the inertial measurement unit (IMU). Design/methodology/approach The parameters of real gyroscopes and accelerometers were mathematically modeled to simulate their performance. The effectiveness of extended and unscented Kalman filters was analyzed for both calibrated and uncalibrated sensor dynamics. The results of the sensor fusion algorithms for both pitch and roll angles were presented. Findings The extended Kalman filter (EKF) and the unscented Kalman filter (UKF) were used to combine data from the IMU sensor to derive pitch and roll angles. Results from calibrated and uncalibrated sensors were analyzed in MATLAB, showing that the UKF outperformed the EKF. Analysis in MATLAB showed that the UKF outperformed the EKF, with lower root mean square error (RMSE) and mean absolute error (MAE). For pitch, the UKF had an RMSE of 1.97° compared to 2.12° for the EKF. For roll, the UKF’s RMSE was 2.70°, while the EKF’s was 2.84°. Originality/value The purpose of this paper was to compare the extended and unscented Kalman filters using a realistic IMU sensor model in a MATLAB/Simulink environment.