Unmanned Systems Technology 020 | Alpha 800 I Additive Manufacturing focus I USVs insight I Pegasus GE70 I GuardBot I AUVSI Xponential 2018 show report I Solar Power focus I CUAV Expo Europe 2018 show report

12 June/July 2018 | Unmanned Systems Technology Two Italian engineers have invented a tele-operated underwater robot that resembles a self-propelled stretcher to help save people from drowning (writes Peter Donaldson). Angelo Bonfitto and Nicola Amati said the vehicle, which can also be equipped for autonomous operation, consists basically of two cylindrical hulls linked by a pair of transverse members and a net between them. To effect a rescue, the robot is thrown into the water and driven under the victim, who can be either conscious or unconscious, then scoop them up Chip maker Ambarella has developed its own driverless car to test out a range of visual sensor technologies without using Lidar lasers. The Embedded Vehicle Autonomy (EVA) is a Lincoln MKZ fitted with Ambarella’s CV1 camera system. This has two stereoscopic systems for detecting objects, one of which is optimised for long range up to 150 m and the other for a shorter range within that. It also has a monocular system with a range of up to 180 m for object identification and mapping. The long-range system uses two 4K sensors (8 MP resolution) with a 30 cm baseline and a 75 º horizontal field of view (FoV). It is mounted on top of the car to provide 360 º coverage. The shorter range system consists of four stereo cameras using 2 MP sensors with a 10 cm baseline and fisheye lenses mounted on the front and rear bumper, and on the two sides of the car. With a 180 º horizontal FoV, the cameras can into the net. It is designed to operate at depths of up to 1.5 m. Bonfitto and Amati said the robot is effective in rough water, at a maximum speed of 10 knots (5.15 m/s) and with an endurance of at least 15 minutes. The hulls are made from PVC but with an aluminium head at each end and welded aluminium mountings to support the transverse members and other appendages. These include the shafts for the front and rear hydroplanes that provide depth control in dynamic diving mode, a podded propeller under the rear of each hull and a buoyancy change pump under perceive obstacles in all areas around the vehicle. Radar is also installed on the front bumper for additional frontal perception. All the cameras are powered by Ambarella’s ultra-HD CV1 chip, with each long-range camera using two of them. The results from all the cameras, along with the radar and other sensors, are sent to a central PC running data fusion and path-planning algorithms. Ambarella has also launched a next- generation chip, the CV2, which provides the front of each hull to pump water in and out of ballast tanks in static diving mode. Batteries, motors and control electronics are contained in each hull. A searchlight and a camera to help locate the victim are mounted on the front transverse member. An operator on the shore, aboard a vessel or in a helicopter, controls the vehicle over a radio link to a buoy that is connected to one of the hulls by a short tether over ranges of up to 400 m. The inventors used Abaqus software to validate the design of the hydroplane and propulsion systems. 20 times the deep neural network processing performance of the CV1. Four 1.2 GHz ARM Cortex A53 processor cores with Neon digital signal processing extensions and a floating- point unit allow the chip to run multiple algorithms simultaneously, for delivering higher perception accuracy and to reduce the total number of chips required. It also includes security features to prevent hacking, including secure boot, ARM’s TrustZone hardware and I/O virtualisation. ‘Lifeguard’ to the rescue No-Lidar autonomous car Underwater vehicles Driverless vehicles Cameras around the Ambarella EVA feed data to a PC for data fusion and path planning