Issue 40 Unmanned Systems Technology October/November 2021 ANYbotics ANYmal D l AI systems focus l Aquatic Drones Phoenix 5 l Space vehicles insight l Sky Eye Rapier X-25 l FlyingBasket FB3 l GCS focus l AUVSI Xponential 2021

81 Ground control stations | Focus power and avoid aerial detection while the latter transports it to a new launch location. Alternatively, multiple UAVs could navigate to a new area in a morphing swarm formation, bunching closely together to avoid topographic obstructions or unwanted surveillance, before spreading out again to engage in wide-area gathering of data. However, making such a solution work is easier said than done. Accurately recognising and gauging data inputs, then correctly selecting and smoothly transitioning into the right sequence of response behaviours, relies on a considerable range of AI technologies. These include developing semantic reasoning engines that can use sensor inputs to correctly and logically infer the consequences of sensor data, as well as training the sequences of behaviours through various forms of machine learning such as reinforcement learning, unsupervised learning and supervised learning. However, following this path is key to GCSs unlocking advanced capabilities such as dynamic geo-fencing, seamless handing over the control of more than one or several unmanned vehicles from one operator to another, or even enabling decentralised ad hoc swarming behaviours that have long been touted in academic literature. AI systems are also key to effectively tracking the hours logged by each component on an unmanned system. Automated alerts and reports can then be supplied to an end-user when something like a muffler, fuel pump or battery is due for replacement. Tracking component usage can also be used to view the speeds, power outputs or hours of use on vehicles, engines and other subsystems, and thereby visualise trends in how hard or how long operators are running their unmanned systems. By incorporating AI-based maintenance logging into their GCS packages, OEMs can not only ensure operators comply with airworthiness standards on maintenance, they can also understand better how to tailor their next-generation systems to how their customers need them to perform. Computing advances Controlling swarms of different unmanned vehicles from a single GCS, and performing other similarly advanced capabilities, requires considerable computer processing power, and UAVs tend to hit their limits in this respect before USVs and UGVs, simply by being more weight-constrained. A large UAV or UGV equipped with a fuel tank and hybrid-electric powertrain should have little trouble running the latest Nvidia or Qualcomm SoCs, which were developed with computer vision, machine learning and other AI technologies in mind. By contrast, an electric quadcopter small enough to be handheld is far less likely to be able to carry such a powerful computing system, but there are workarounds for this. Unmanned Systems Technology | October/November 2021 High-end networking protocols and graphics engines are a must-have for professional operators across defence and commerce (Courtesy of Robot Aviation) A robust architecture of edge and cloud computing is key to remotely running mission profiles, AI behaviours and analytics (Courtesy of Aquiline Drones)