Issue 39 Unmanned Systems Technology August/September 2021 Maritime Robotics Mariner l Simulation tools focus l MRS MR-10 and MR-20 l UAVs insight l HFE International GenPod l Exotec Skypod l Autopilots focus l Aquaai Mazu

56 Insight | UAVs days, but the way Skydio have used it makes it more advanced in many ways than Tesla’s autonomous driving system for example,” Smith observes. “That’s because it has to be self-aware in the vertical axis as well as watching out for what’s in front, behind, to the left and right. To use the system in a warehouse, Ware first creates a 3D digital twin of the facility. That generates the most accurate cartographic map available for the Skydio 2s – Ware routinely does this remotely – to ensure the UAVs know where each rack and pallet is. The map is used to generate flight paths for the UAVs, so they can launch, take several hundred pictures (encompassing every pallet location) and land back to where they can be recharged, while the data they have collected is uploaded to Ware’s cloud platform for processing. Using Ware’s back-end image analysis, stock barcodes are recognised and cross-referenced with localisation data to ascertain whether or where the end-user’s WMS contains errors or needs updating. This information is then shared with the client via reports in the front end of Ware’s cloud platform as well as direct API data transfer into the customer’s system for them to take the appropriate action in their WMS. Front-line law enforcement As well as a highly robust visual navigation system, having some manner of protective cage encircling a UAV is ideal for keeping it protected while flying indoors, as companies such as Flybotix and Flyability have shown. The Dronut UAV from Cleo Robotics is the latest system to follow this trend. Rather than being content with a mere cage, however, CEO Omar Eleryan and his co-founder Simon Czarnota have designed it to form essentially a small ducted fan (with the duct also serving as the fuselage for housing the avionics) with two coaxial motor-propellers, covered on top and bottom by vents. That makes it near-impossible for anything to get caught in the prop blades. “After studying mechanical engineering at college I worked for a multinational oil firm, and saw first-hand how dangerous visual inspections of tanks and pressure vessels can be,” Eleryan recounts. “Humans need to don so much protective gear and move so cautiously, so it makes far more sense to use a robot; however, existing robots weren’t ideal for that purpose. Even now, most small quadrotors – caged or otherwise – are too big; we wanted to create something that can get into really tight spaces.” To that end, the Dronut is just 155 mm in diameter and 100 mm in height. In addition to integrating an array of different sensors for autonomous navigation and inspection, it carries 2800 mAh of battery capacity, which typically enables 15-16 minutes of flight, for an overall weight of 370 g. “We couldn’t just take a standard quadrotor design and shrink it – we’d lose a ton of efficiency, and get maybe a few minutes of flight at best,” Eleryan explains. “I’d previously looked into ducted fan technology, which has existed for around 70 years but has never really been commercialised. We knew though that compared with a quadcopter it would be ideal for carrying a much heavier payload, have a much smaller form factor, and no exposed propellers that could be damaged from bumping into objects.” Eleryan and his team therefore set themselves to solving key issues previously encountered with ducted fans, such as controllability and manoeuvrability, to ensure flight stability. Pitch and roll are achieved via a series of control flaps on the underside, which direct the downwash of air as needed to bank and rotate the Dronut through corridors and pipes. Mounted at the front of the duct enclosure (which is made from reinforced plastic) are the sensors for navigation and inspection. A compact Lidar sits at the front and is the main tool for object sensing and optical flow measurement. Several IMUs are also mounted for accurate dead reckoning, giving an accuracy in hover of ±5 cm and a horizontal accuracy of ±10 cm. Also, two cameras are mounted at the front, facing upwards and downwards to fuse together a 100 º FoV in front of the August/September 2021 | Unmanned Systems Technology Cleo Robotics’ Dronut flies using a ducted coaxial fan, and navigates indoors using a combination of Lidar and vision (Courtesy of Cleo Robotics)