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

60 machine learning and convolutional neural networks. It will use dataset extraction and response charts, predefining reactions to objects in the vehicle’s path. The high-level reaction model is predicated on continuing the mission by any practical means or retreating to a predefined fall-back position. To help the perceptive autonomy architecture understand what the Lidar and stereo cameras feed it, it will have a sensor model and supplementary light and temperature sensors operating in real time to identify environmental factors. It will also run a priority model for selecting data and data types from the Lidar and cameras. The model bases the confidence level it assigns to its solutions on factors including environmental conditions, location certainty and comms assurance. Within the SLAM system, the GNSS- independent localisation will depend on proximity sensors and dead reckoning with INUs, while the mapping relies on the vision suite, environment sensors and models that support interpretation of the Lidar data. Going electric While the electric driveline retains the original diesel-engined vehicle’s all- wheel-drive, three-axle, 6 x 6 layout, the way the power reaches the wheels is very different. Each wheel is bolted to a flange on the output shaft of its own Dana Brevini planetary reduction gearbox that is driven via a Gates Polychain belt that provides a 2:1 reduction from the Ashwoods interior permanent magnet electric motor. That is controlled by an inverter/controller from Curtis Instruments. This layout was chosen instead of in- wheel motors because it minimises the amount of change required to the original ATMP chassis – reducing development risk – and because the belt drive from the motors to the gearboxes allowed the company to change the ratios during development more easily. Precise control of the motors via a master controller, two sub-master controllers and three slave controllers enables them to provide braking for skid steering, removing the need for a more complex system, and allows them to work as generators to recharge the battery. This interim control system will be replaced by a new configuration though, using a single master and six slave controllers, between now and 2021. The steering geometry is unchanged from the original vehicle, but the engine- driven pump has been replaced with an electro-hydraulic unit that provides pressurised fluid to the steering ram. The electric vehicle has an ‘e-differential’ function that enables both skid steering and counter-rotation with all the wheels, the latter enabling turns on the spot. Ongoing software development is focused on creating a proportional integral derivative-optimised steering system governed by the autonomy software’s path-planning model. In the current configuration of the electric/hybrid vehicle, the original’s hydraulic braking system is supported by sensor-assisted and controller-configured regenerative braking, with support from the terrain-understanding software. Battery, generator options To keep development costs down, electrical power in the demonstrator is provided by a pack of 14, 12 V lead- acid car batteries connected seven in series, two in parallel, to provide the required 84 V. “Initial testing has been done with lead-acid batteries, because they are cost-effective,” Field says. The same cannot be said though of the modular, high-capacity lithium-ion pack with which they will soon be replaced, which at the moment is priced at about £30,000. Austen remarks that Supacat has a couple of suppliers for this in the frame, and that prices have come down substantially over the past year. Plugging a small engine-driven or fuel cell generator into the system as a range extender will provide the option to run the ATMP as a series hybrid. However, the situation with military generators at the moment is awkward, because fuels other than diesel for ground vehicles don’t December/January 2020 | Unmanned Systems Technology Each wheel is driven by its own reduction gearbox. The drive belts are robust and made changes of gearing relatively simple during the driveline’s development process