Uncrewed Systems Technology 044 l Xer Technolgies X12 and X8 l Lidar sensors l Stan UGV l USVs insight l AUVSI Xponential 2022 l Cobra Aero A99H l Accession Class USV l Connectors I Oceanology International 2022

52 extends pneumatically until all four pliers touch the fronts of the car’s tyres. Then a second group of pliers is deployed, one behind each tyre, spaced so as to ‘pinch’ all four tyres to secure a tight grip on the car. “In the first prototype, the Stan used a system of cylindrical actuated rolling wheels to ‘scoop’ up and under the car’s wheels,” Boussard recounts. “We realised during trials though that this approach expended too much energy and needed too much ground clearance. “But we could still lift 80% of the cars, and proved that the concept worked and was profitable for airport owners. So we continued our r&d until we arrived at the optimised solution we have now.” Once held securely, the car is lifted into the air using two mechanisms. One is a pair of electrically powered jacks at the head that lift the front of the platform; the other is a pneumatic cushion in the mobile platform that is inflated to lift the back of the flatbed (the required thinness of the platform made using a jack impossible at the rear). The cushion sits atop a pair of roller wheels that keep the platform from dragging along the road from the weight of the car. SLAM and sensors for parking A considerable amount of sensor data and software intelligence is required to ensure the various mechanical systems can be deployed and used effectively. The primary perception tool here is Lidar. Velodyne Puck Lidars are used for real-time world mapping and obstacle detection around the Stan, although Stanley Robotics tests and benchmarks new Lidars as they come out, and the company says it maintains ongoing discussions with 10-20 Lidar suppliers to enable such testing. “Most often we’ll ask them to send Lidar data recordings, then we analyse them and check them with our own algorithms, and where possible qualify the quality of the data signal,” Trouble explains. “If we think there’s potential, we borrow a Lidar, install it on a test version of the Stan and trial it for a few weeks to decide if we want to go further.” As lengthy as the process is, the crucial importance of high-quality Lidar to the Stan’s operations make it essential, and as solid-state Lidars hit the market and become more affordable, Stanley Robotics anticipates doing more and more such testing in the future to select the right model. Since the Stans have to operate in dynamic and constrained environments, developing SLAM capabilities was vital. Early versions of the company’s SLAM were based on the Hector SLAM approach, but these algorithms produced errors such as the ‘corridor effect’ in which robots cannot determine their longitudinal position in lengthy corridors or streets, or give 90 º false headings in square-shaped rooms. Odometry sensors were installed to mitigate those, but Stanley’s engineers nonetheless developed an optimal algorithm to prevent errors occurring in the first place. “We have now designed a brand new SLAM module based on the Cartographer algorithm,” Trouble says. “It required a lot of adaptation for our use-case though, because the availability of landmarks change a lot throughout each of the Stan’s routes, and landmarks are key for the localisation aspect of SLAM. We’ve run lengthy tests of it, shadowing it with our GNSS readings and our previous SLAM 3D models from older SLAM algorithms, and the results have been great.” Both the front and rear Lidars are critical for obstacle detection, but the forward one is more important for forming 3D point clouds for SLAM processing. “The rear one cannot do that as well, because we don’t spend enough time moving backwards to map in that direction,” Trouble notes. “But it can verify vehicles and other objects recognised by the front Lidar, especially since the Stan reverses into vehicles to slide the platform under them afterwards. “That’s when the rear Lidar becomes indispensable. Sometimes the clearance between a car’s left and right tyres is barely more than 1 m, so to fit our 1 m- wide platform beneath it, we need a millimetre-accurate point cloud that is updating in real-time to hold the central axis as we slide in.” Both Lidars are key for lining the Stan up with the central axis between a car’s left and right wheels, with a wide field of view and high point density being critical performance values. To ensure the timely deployment of the pliers, contact sensors June/July 2022 | Uncrewed Systems Technology The platform uses four pairs of extendable rods, or ‘pliers’, to grip and lift cars firmly by their tyres