Unmanned Systems Technology 019 | Navya Autonom Cab | Batteries | UGVs Insight | UAV Factory UAV28-EFI | Swiss Aerobotics Hummel | UMEX 2018 report | Antennas | Oceanology International 2018 report

70 system such as a net thrown over target. Agencies in the US want to be able to defeat different or multiple UAVs with one counter-platform, the idea being to throw the net, fly back, dump the UAV and repeat the process. To achieve that, Weiss started out with a stripped-down sense-and-avoid technology as the aim is to target the other aircraft rather than avoid it. This is less challenging than developing a system to avoid large aircraft, which needs a range of several hundred metres. This system uses optical detection with the OpenCV image detection library running on a Raspberry Pi processor board linked to a simple camera. The next step is to use deep-learning neural network algorithms to identify the shape of the target, but there is currently no dataset for training such systems; Swiss Aerobotics is working with a Swiss company on that data. “For me the important thing is to demonstrate that the detection system works, so we are using a tag on another UAV to detect where it is,” says Weiss. The idea is to have three identification systems: the computer vision system, a radar system and another event-driven visual system to detect the moving propellers from any kind of UAV. The company has therefore developed a frequency modulated continuous wave (FMCW) 25 GHz radar system with a radar reflector. FMCW emits a stable- frequency continuous wave and receives signals back from any reflecting objects. It uses Doppler detection to eliminate signals from large stationary objects and slow-moving clutter, allowing the system to target the rogue UAV. The idea of having radar on the system is so that it works in deserts or snowstorms where visual systems may not operate reliably, says Weiss. The computer vision and event-driven visual system is more suited to urban environments where a counter-UAV system would be used, such as a stadium or other buildings. “The vision system is an important thing to be able to operate a counter- UAV system in populated areas,” says Weiss. “Someone needs to make a decision about whether to go in for the capture and you need feedback from the UAV, and that’s where the vision system comes into play, by sending a single image back to confirm the intercept.” While the counter-UAV is mostly autonomous in targeting and capturing a rogue UAV, the connection to it needs to be secure. Weiss is using a 4G cellular link that would operate well in urban areas, and a huge bandwidth is not required for the single image, which increases the range, but security is still key. So the PixHawk flight controller will be linked to a host controller, which could be a Raspberry Pi or an Nvidia Jetson, on the Hummel. The comms April/May 2018 | Unmanned Systems Technology Testing a prototype of the Hummel UAV It is important to demonstrate that the detection system works, so we are using a tag on another UAV to detect where the craft is