Obtaining ground truth data in C-UAS trials

Performance evaluation of C-UAS sensor systems and the development of detection algorithms require ground truth knowledge of the position of the target drone to be available during the measurement. In practice this is done by retrieving flight logs from the UAS, and comparing the obtained GNSS positions with the known position and orientation of the sensor at the corresponding timestamps in the recorded sensor data. Alternatively, video or sensor data can be manually annotated by an operator to indicate whether a drone is seen in each frame, allowing sensor performance to be evaluated. These methods are time consuming and can be prone to introduce errors into the process. Also, not all UAV platforms allow for these data to be readily accessible. In our approach we use the open source MAVLink protocol used in Ardupilot-based UAVs to implement a trial range framework, which communicates between drones in flight, the ground control station, sensor platform trackers and optionally GNSS tracker modules to build a real time representation of the full trial setup, in order to provide unambiguous ground truth with known accuracy at all times. We demonstrate this setup, which will allow us to perform consistent evaluations of sensor systems in future trials.