Unmanned and manned aerial vehicles have been used extensively for civilian purposes over the past few decades, thanks to unprecedented advances in micro-electro-mechanical systems (MEMS) sensors and microprocessors. One can find various sensors in aerial vehicles, such as triaxial accelerometers, triaxial gyroscopes, triaxial magnetometers, GPS modules, and pressure sensors. The digital sensor signals are processed in microprocessors using sensor fusion techniques, which provide navigation capabilities for aerial vehicles. In this thesis, we present an approach of design and implementation of a sensor fusion algorithm for fixed-wing aerial vehicles. The flight state variables of the sensor fusion algorithm are the position, velocity, orientation of the aerial vehicle, biases of an inertial measurement unit (IMU), and wind speed. The state dynamics are nonlinear due to the orientation. Therefore, we propose to apply an extended Kalman filter (EKF) using inertial measurements (acceleration and angular velocity), GPS position, static pressure, dynamic pressure, and air temperature measurements. In addition to these measurements, two aerodynamic constraints (side force and sink rate polar) of a fixed-wing airplane are used for the wind estimation using an assumption of a wind triangle. We propose not to use magnetic measurements since they are easily distorted by un- known magnetic fields of other electronic devices in the vicinity. The distortion is not well compensated, if at all. Without heading information from the magnetic measure- ments, the horizontal wind cannot be uniquely determined in a single wind triangle. Therefore, we investigate the wind estimation in the proposed EKF without a magnetic sensor. Using an analytical observability analysis, we prove that the wind is observable in the case of a time-varying true airspeed (TAS) direction, which is commonly fulfilled in practical flights. To this end, a numerical study of the observability of the EKF using measured flight data of a manned glider shows that the state vector is effectively observable independent of flight maneuvers. We show that the EKF works without a magnetometer. Furthermore, in order to study the tracking behavior of the individ- ual state variables, we present a method based on the triangularization of the system transition matrix. We show that the wind estimation error can converge in flights with dynamically changing TAS direction. The TAS direction determines which direction of the wind estimate converges faster. A further key contribution of this thesis is the experimental evaluation of the proposed EKF using recorded flight measurements in different manned gliders under realistic envi- ronmental conditions, e.g., smooth air and turbulent atmospheres. The pilot-in-the-loop strategy allows us to collect and label various flight maneuvers, including gliding, soaring in thermals, uncoordinated turning, stall, and free-fall flight. The results show that the EKF can accurately estimate the position, ground speed, orientation, IMU biases, and wind speed in real-time. The horizontal wind estimate is verified by circle shifting in a thermal soaring of a glider. The vertical wind estimate is instantaneous and accurate and can be used to indicate a strong updraft. In addition, we determine the side force coefficient of the side force model using uncoordinated turnings. As a byproduct, the EKF can estimate the angle of attack using a three-dimensional TAS vector. The AoA estimates are evaluated using two specific flight maneuvers, i.e., stall and free-fall flight.weiterlesen