Joint Virtual and Physical Prototype Design and Testing of a Sensor Fusion Workbench for Fixed-Wing UAVs

Abstract

Unmanned aerial vehicles (UAVs) require cost-efficient on board multi-sensor fusion to achieve accurate and reliable flight state estimation. The challenges behind the implementation of sensor fusion algorithms from scratch towards in-flight testing on microcontroller-based hardware are, however, demanding, since it requires not only understanding and implementing complex sensor fusion algorithms but also developing embedded software in a microcontroller. Those challenges make the process of prototyping onboard sensor fusion algorithms time-consuming, expensive, and inefficient. We present fast prototyping of a sensor fusion workbench based on extended Kalman filtering (EKF) for fixed-wing UAVs. The workbench incorporates multiple sensors, including an inertial measurement unit, a GPS receiver, and static and dynamic pressure sensors. The multi-sensor fusion algorithm has been tested under virtual flight data, as well as measured flight data including challenging environments with thermals and gusts. After performance evaluation, we transferred the algorithm from MATLAB code to C code and integrated it into an avionics device LXNAV S10. The implemented sensor fusion algorithm has been successfully tested on a manned glider under windy and turbulent environment.