Unmanned Systems Technology 020 | Alpha 800 I Additive Manufacturing focus I USVs insight I Pegasus GE70 I GuardBot I AUVSI Xponential 2018 show report I Solar Power focus I CUAV Expo Europe 2018 show report

7 Platform one Lockheed Martin has applied for a patent for a low-cost system to prevent collisions between UAVs and manned aircraft (writes Peter Donaldson). At the heart of the system is a radio beacon installed on a UAV that continuously transmits signals of very low power – about 10 mW for example – on standard training or emergency frequencies such as 121.5, 243 and 406 MHz that are commonly monitored by all manned aircraft on their standard radio systems. Intended for civil as well as military use, the beacon can transmit either simple alert audio or a text message with altitude and/ or location information, allowing pilots of manned aircraft to take evasive action. As well as helping to minimise the size, weight and power requirements for the UAV installation, this low-power transmission ensures that only manned aircraft receive the signal when the UAV is close enough to be of immediate concern. That is because a minimum received power in a signal on a desired frequency is necessary to break through an FM receiver’s ‘squelch’ noise suppression function. As ground-based air traffic management services also monitor the frequencies concerned, they can receive the signal as well, enabling them to warn manned aircraft to keep away from the UAV. This is particularly valuable for aircraft heading towards it but still too far away to receive the signal, said the company. In addition to the transmitter, the beacon includes a small modular antenna designed to be easy to install even on Group 1 UAVs – those weighing 20 lb (7.8 kg) or less. The invention also covers a controller, implemented in hardware or software or a combination of the two, to process signals and display information from the beacon either aboard manned aircraft, in ground- based systems or both. Airborne vehicles Low-cost flight safety A US start-up has used machine learning to develop an autonomous ‘flying camera’ that will follow a user (writes Nick Flaherty). The R1, developed by Skydio and built from aluminium and carbon fibre, has a 13-camera system that sees in every direction. It feeds what it sees into a 256- core TX1 graphics processor from Nvidia to handle the machine learning, which is called the Skydio Autonomy Engine vision processing system. The TX1 examines the images to identify landmarks in the environment, tracking them over time to estimate its own motion with centimetre-level precision. Six pairs of cameras allow the TX1 to build a detailed stereoscopic depth map around the UAV, which are combined into a 3D model. The R1 then uses a deep neural network to identify a subject and other people seen by the forward-facing cameras. As it tracks people, it learns what they look like so that it can distinguish between them and stay locked on to a particular person. The R1 continually predicts its likely position as far as 4 s into the future and maintains a balance between obstacle avoidance, keeping subjects in view and moving smoothly to capture video from the 13th camera. These requirements are converted into rotor speeds and gimbal angles. Learning to stay on track Airborne vehicles The R1 uses cameras and machine learning for navigation and to lock on to a subject Unmanned Systems Technology | June/July 2018