Sensor Fusion Using Extended Kalman Filter for 9-DOF IMU

This paper presents a sensor fusion algorithm using an extended Kalman filter (EKF) for 9-DOF inertial measurement unit (IMU) including three gyroscopes, three accelerometers and three magnetometers for attitude tracking system. The quaternion-based orientation is tracked by four states Kalman filter. In the predict phase of Kalman filter, the gyroscopes provide angular rate to calculate orientation in quaternion. The measurement update phase of Kalman filter does not use orientation calculated from gravity and magnetic field as measurement vector. Instead, the gravity and Earth's magnetic vector are used directly to constrain the orientation computed previously. Experimental measurements show the performance of sensor fusion with discussion about its advantages and disadvantages.