An estimation method of 3D posture using quaternion-based Unscented Kalman filter

This paper proposes an estimation method of 3D posture (quaternion) using inertial sensors. The gyro sensor output includes the drift error. The calculation error of 3D posture by gyro sensor increases with the passage of time due to the calculation of time integration. This method can estimate the 3D posture compensating the drift error by the Unscented Kalman filter, and the method can avoid the effect of dynamic acceleration and magnetic disturbance. The 3D posture is represented by quaternion in this method. Therefore, the 3D posture estimated by this method can avoid the effect of singular value problem. In order to indicate the effectiveness of the proposed method, we conducted the experiment using the experimental setup installing rotary encoders. The results by proposed method approximately corresponded with the results by rotary encoders. The precision of results by proposed method was higher than that of results by gyro sensor, and we estimated the 3D posture in the motion generating singular value problem. The proposed method is used to the 3D posture measurement.