Sensor Fusion Design by Extended and Unscented Kalman Filter Approaches for Position and Attitude Estimation

For a stable autonomous flight for small unmanned aerial vehicles (UAV), high-precision position and attitude information is required without using heavy and expensive sensors. For this purpose, position and attitude estimation of UAVs can be performed using sensor fusion algorithms based on different approaches. Although there are many studies about the subject, it is difficult to theoretically evaluate the effectiveness of the preferred approaches. This study covers the sensor fusion design and implementation in MATLAB simulation environment for position and attitude estimation by Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF) approaches, which are the most preferred filters in navigation systems. In the study, sensor fusion design is using sensor data of an inertial measurement unit (IMU), global positioning system (GPS), a barometer and a magnetometer on the UAV, without using the dynamic model of the UAV. Quaternions were used as the state variables which indicate the attitude with respect to local navigation frame. The system consists of a total of 22 state variables: quaternions, velocity, position, gyroscope and accelerometer bias, magnetic field, and magnetic field bias. Two different sensor fusion models designed by EKF and UKF approaches which has sensor models that use real (sensor error-free) flight data provided by MATLAB, were run in MATLAB simulation environment. The estimator performances of these approaches were compared with real flight data and results were evaluated.