88 Over the last five years, autopilots for uncrewed systems have evolved to become much more than ‘black boxes’ responsible only for automated management of aerial, ground or ocean crossings in executing keyed-in GNSS waypoints. For professional and highly autonomous vehicles today, high-end flight controllers are expected to embed or interface with computer-vision AI towards advanced functions such as object detection and classification, collision avoidance, gimbal target tracking and GNSS-denied navigation. Inflows and outflows of data regarding an autopilot’s central processors can now pass through so many sensor connections, edge computers, bridging hubs and signal converters that it is not incorrect to think of the ‘autopilot’ as a cluster or ecosystem of different devices surrounding the central flight controller. Competing autopilot products will vary in how much or little extra they carry in terms of GPUs, IMUs, GNSS receivers and so on. Depending on the solution, some parts may be omitted; others may be installed in triplicate for redundancy. Either way, there is a growing need for AI, along with the processing power to enable it, in today’s autopilot systems, as well as copious connectivity for linking sensors, payloads, secondary computers and much more to the central controller, be it via Ethernet, controller area network (CAN) bus, mobile industry processor Rory Jackson discovers how flight controllers are evolving to become highly intelligent, safety-rated systems rather than simple ‘black boxes’ Boxing clever April/May 2024 | Uncrewed Systems Technology AI, modularity, open architectures and redundancies are must-have elements for uncrewed vehicle autopilots in 2024 (Image courtesy of Auterion)
RkJQdWJsaXNoZXIy MjI2Mzk4