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

96 Show report | CUAV Expo Europe 2018 BIT Center displayed its machine learning and computer vision capabilities for unmanned systems. The technology is aimed at providing solutions for areas such as SLAM, reconstruction of 3D maps of streets and corridors, infrastructure defect detection and change detection, as well as collision avoidance for self-driving vehicles. “We use a stereo camera set-up, as it allows us to identify the scale of objects and distances as the scene changes,” explained Gleb Panchishen. “The polarisation from relying on three or more cameras gives enough different angles to collect sufficient data to reconstruct 3D scenes from one shot at 10-15 m to reduce the parallax effect.” BIT Center’s AI learning for collision avoidance is powered by convolutional neural networks (CNNs). The company also plans to apply CNNs to its polarisation techniques for 3D filming in the future. Septentrio launched the new AsteRx-i, which combines an IMU with an RTK-GNSS receiver to provide positioning accurate to 1 cm for UAVs and autonomous road vehicles. “GNSS can do a lot by itself, but having an integrated IMU is only natural as unmanned vehicles evolve to require higher levels of accuracy and reliability,” said Francesca Clemente. “You need redundancies, and to provide not just accurate positioning but attitude readings as well.” The new system also comes with Septentrio’s UAS-tailored carrier board, designed with embedded software and connectivity for seamlessly integrating autopilots and payloads, and a 6-30 V power supply input. “Autopilots typically have specific signal protocols and electrical requirements. The AsteRx-i V UAS carrier board is designed for reducing the complexity of integrating these and other systems together into a UAV or other unmanned systems,” Clemente added. The board uses a micro SD card (up to 32 Gbytes) to log flight data. It is designed for compatibility with Pixhawk and Ardupilot systems, and features three LVTTL serial ports and a micro USB device port. Auterion was walking the aisles of the show to discuss PX4, one of the most widely adopted open-source autopilot software for unmanned aerial and other autonomous systems. PX4 is a complete software platform that includes flight controls, computer vision, comms and GCS, and is part of Dronecode and the Linux Foundation. The software is Berkeley Software Distribution-licenced, which owing to its permissibility allows for mixing and matching of open-source components with proprietary components without exposing the user to legal risk, enabling manufacturers to keep their differentiation and IP. “Another choice we made when developing PX4 was making sure to use a modern messaging architecture, in order to keep it very scalable and modular,” said Kevin Sartori. “That way, the architecture ‘stack’ can be used for very different purposes – for UAS or different CPU platforms. “Our SDK enables integration of different components and apps, and we’re integrating new chipsets from NXP or Qualcomm and AirMap’s UTM software into our GCS, for example.” This choice of messaging architecture means the autopilot is designed in a similar way to modern web services, with a modularity that allows a user to pick, place and remove different blocks within the overall platform. Users can then adopt as little or as much of the programming stack as necessary for their vehicle, such as for integrating comms, computer vision, a GCS or other parts. “Adding a new payload or other system just connects to the message bus, so you don’t need to touch the rest of the architecture,” Sartori added. Ouster attended the show to display its OS-1 Lidar system, which is designed as a lightweight 360 º Lidar sensor for autonomous vehicles, UAVs and other markets. “Historically, Lidar sensors have had many hundreds or sometimes more than 1000 laser diodes, PCBs, receivers and other discrete components, resulting in large, expensive, difficult to manufacture, power-hungry and unreliable products,” explained Cyrille Jacquemet. June/July 2018 | Unmanned Systems Technology The QGroundControl GCS portion of Auterion’s PX4 autopilot software