Issue 55 Uncrewed Systems Technology Apr/May 2024 Sellafield’s UAV equipment l Applied EV Blanc Robot l Battery tech l Robotican’s Goshawk l UGVs l UAVHE RW1 rotary l Roboat UVD l Autopilots l Arkeocean UVD l UMEX 2024 l CycloTech UVD

Robotican’s Goshawk counter-UAV | In operation The packs are kept at 100%, so the LEDs are just for the humans in-the-loop,” Geva says. The Goshawk will tilt its gimbal upwards after a launch sequence has been triggered to confirm that its Nest’s doors have opened successfully (rather than relying on an open-loop of simply being told by the Nest that it has opened its doors, which could present a problem for the UAV if the Nest’s sensors are at fault). “As security is an asymmetric problem these days, and the threat of swarming drone attacks is a looming concern for some, we work closely with customers to assess the smartest way to deploy Nests for each facility’s unique threat level,” Balshai says. “For instance, some might feature multiple attacks at a certain portion of their site, and so maybe more than one Nest and Interceptor at that area is smart. On the other hand, the UAVs don’t always have to return to their Nest, or even to another nest – the Goshawk has carbon-fibre landing struts for touching down anywhere, and technicians can simply drive or walk over at the end of the day to collect them once the area is deemed safe.” In pursuit Such is the level of autonomy in the Goshawk and Smart Nest mission infrastructure that waypoints are not used in the path planning and guidance of the Goshawk, or at least none visible to the monitoring crew other than any pre-set preferential flight corridors which the UAVs opt into when flying through the affected areas. “Although the operator can set such barriers, geofences and preferred path sections, and the Interceptor takes these into account, it’s the UAV that plans its own route to the target, constantly checking that path and if it still leads to the target, based on where the detection system or the UAV’s own camera say it is,” Geva says. “We often expect that a hostile drone won’t be stationary; it’ll be manoeuvring, climbing, accelerating and so on, and so the Interceptor must be able to detect changes in where the target seems to be, relative to where its own GNSS and inertial system says it is, in order to keep homing in on the target until the capture stage can be initiated.” Part of this is achieved by the Goshawk changing its search location in accordance with changes in the target’s location, as communicated by the enduser’s ground-based detection system. To develop the Goshawk’s embedded software, Robotican made use of certain open-source frameworks, including choosing ROS 2 as its middleware. However, its capabilities in the detection, recognition and tracking of objects such as other drones were developed in-house through machine learning. “We used tens of thousands of images of potential enemy drones with an artificial neural network to train the software to detect hostile UAVs, and we developed some very unique detector and tracker software modules that work together with the guidance algorithm to produce the homing intelligence by which the Interceptor closes in on its targets,” Geva says. The four-corded ‘net’ is used for non-destructive tangling and capture of hostile drones by the Goshawk, which snares them from above Download the whitepaper for free! How UAV/AAM leaders can challenge the sky without reinventing the wheel. Aviation experts reveal what makes the ideal key component supplier. VOLZ Servos Reliability for progress