UST030

51 a 10 m wingspan, and when dismantled will fit into a container 6 m across so that it and its ancillary systems can be transported by cargo aircraft.” Opteran is a spin-out from the University of Sheffield that has developed a SLAM navigation system for UAVs. It is based on how honey bees travel several kilometres per day between their hives and flower patches using only visual memory. “The brain of a honey bee only has about a million neurons [compared to 100 billion in the human brain], which makes them incredibly computationally efficient,” explained Alex Cope. “We reverse engineered the principles of how they measure things moving across their visual field, using existing research on the bees’ neural underpinnings and behaviours to reconstruct nature’s algorithms from scratch,” he said. “That meant we knew our algorithms had to include insensitivity to contrast and spatial frequency, and operate over three orders of magnitude of angular velocity. It also had to include a log-linear response function, using the kinds of correlation detectors that a lot of insect species use to view the world. “Our testbed UAV uses the resulting algorithms for flight control and navigation, and our tests so far show that the algorithms outperform the deep neural network-based approaches while using several orders of magnitude less processing power, with no training required.” As a result, Opteran’s system can be run on a 1 W processor, compared with the much larger, more expensive and power-hungry GPUs and CPUs needed to run conventional SLAM systems. The testbed UAV weighs about 250 g (including the required processor, vision camera and a panoramic mirror system to give a wide field of view), and the company sees applications for its software across a range of UAVs, driverless vehicles and other autonomous systems. “From those base algorithms we can develop further intelligent functionalities for autonomous mapping and navigation, while remaining as computationally efficient as before – or becoming even more so,” Cope added. Aichi Europe, the European subsidiary of Japan-based Aichi Steel Corporation, attended the show to display the electric lift motors it has developed in partnership with Sawafuji Electric, which are installed on its hybrid-electric hexacopter UAV (see page 58). As Aichi Europe’s Sven Meise told us, “These 1.5 kW motors use magnets made from Magfine, a plastic-bonded anisotropic neodymium, instead of sintered neodymium and dysprosium. That means they can be manufactured much faster and cheaper than typical hand-made or machine-produced motors.” The company’s UAV motors are assembled using injection moulding to install a large-bore Magfine magnet ring into a magnesium rotor housing, which enables weight reductions of around 30% compared with conventional assembling and bonding techniques. The anisotropic magnet material can also be moulded to optimise the direction of the magnetic field, which reduces torque ripple by roughly 20%, significantly reducing a large source of motor vibration. The motors also use aluminium coils rather than copper windings, to reduce their weight and manufacturing cost, as well as improve their lifespan and insulation stability. Aichi Steel is planning further r&d into magnetic powders and advanced machinery for injection moulding and aluminium winding, in order to extend its capabilities towards manufacturing complex customised versions of its motors. Unmanned Systems Technology | February/March 2020 Schematic of Opteran’s neural algorithm Rotor element of Aichi Steel Corp’s UAV lift motor