Unmanned Systems Technology Dec/Jan 2020 | Phoenix UAS | Sonar focus | Construction insight | InterGeo 2019 | Supacat ATMP | Adelan fuel cell | Oregon tour | DSEI 2019 | Copperstone Helix | Power management focus

57 to develop it, but they didn’t feel that would enable them to develop sufficient knowledge and expertise, while the alternative of developing the autonomy from scratch would be prohibitively expensive and time-consuming. They have found a balance between the two approaches in a UK government- supported Knowledge Transfer Partnership (KTP) with Exeter University, a cooperative mechanism that Supacat was already using for development of the electric/hybrid drive. The company sees electric drive as a way of future-proofing its vehicles and as a smoother path towards autonomy. As Austen explains, “Controlling electric motors is much easier than controlling everything associated with a more traditional mechanical driveline.” In a KTP, the university provides an engineer whose salary is paid jointly by Supacat, the government via Innovate UK and the university. Known as a KTP associate, the engineer works at Supacat’s premises to learn about the company’s vehicles and the defence market, while the university provides expertise in autonomy and its applications. Under the arrangement, the KTP associate develops knowledge that becomes the company’s IP, including the software that will run on the vehicle. Autonomous in parts The headline goal of this optionally manned project is to develop and integrate an autonomous control system for the ATMP that can operate in harsh off-road environments and fulfil manned roles in military and civil applications if required. “Over the past three or four years, there has been a big splash on autonomy, with all the car manufacturers saying they will have autonomous vehicles available soon,” says Mark Field, Supacat’s principal electrical engineer. “They seem to have stepped back from that a little now – it’s less about being fully autonomous and more about having autonomous features.” That, he adds, is because developing systems that can make reliable high-level decisions is very difficult. “We are considering operations in GPS- and comms-denied environments, and working to make sure the vehicle knows how to behave when it loses those signals,” he says. “Does it go back to a predefined position, wait where it is or carry on with its mission? Those things will depend on what it is doing at the time and what it can sense locally.” Among the key goals is to demonstrate variable levels of collaboration and control from human to vehicle and between vehicles. “With an optionally manned vehicle, levels of collaboration will change depending on the mission, so it is vital to have a really good human-vehicle interface,” Austen says. “It needs to be an intuitive interface that doesn’t overload the operator or controller, because they may well be supervising a number of different autonomous vehicles.” The demonstrator will also include terrain detection and response, to enhance vehicle mobility and optimise endurance, plus the ability to categorise objects as obstacles – or non-obstacles – and to tailor the vehicle’s response in terms of avoidance or clearance. “You might ask it to go from A to B to C based on a series of waypoints, but what it sees on the ground could be different from what was there when the mission was planned,” Austen says. “On our own test track, for example, after 24 hours of rain a gulley had washed out across the track, and what had been easily passable the day before became an obstacle,” he explains. “So having a terrain detection system allows the system to compare what was predicted with what is actually there, and then refer that to a decision support tool.” If the system determines that the vehicle cannot continue, it can ask for support from a human driver aboard when operating manned, a remote operator or the mission planning software. “With our AI on board, we also expect that the machine will learn and improve its response to the terrain it encounters,” Austen says. Its path planning and motion behaviour will rely on simultaneous Supacat hybrid ATMP | Digest Unmanned Systems Technology | December/January 2020 The electrified driveline retains drive to individual wheels but not via in-wheel motors, a decision that minimised changes to the vehicle’s layout