Log-in here if you’re already a subscriber
Two leading developers of artificial intelligence for aviation are partnering to advance common approaches to certifying airborne machine learning (ML) applications.
California-based Xwing and Switzerland’s Daedalean announced their “pre-competitive collaboration” on November 28 at an AI/ML workshop sponsored by the standards organization RTCA in Washington, D.C. Xwing is using an ML-based system in its uncrewed Cessna Caravan to detect runways from the air and ensure they’re clear of obstacles, while Daedalean is developing a more comprehensive suite of computer vision products for traffic detection, landing guidance and positioning.
Related: Xwing is wading into the shallow end of deep learning AI
Civil aviation regulators have yet to certify a safety-critical application based on machine learning, and Xwing and Daedalean have independently published proposals for how that might be done. Their collaboration will build on the common philosophy underlying their approaches, based on their consensus that some alternative approaches — including formal verification and explainability — are fundamentally untenable.
“Overall we have I think the same philosophy and roughly similar methods to get to the end goal,” said Maxime Gariel, Xwing’s president and chief technology officer, in a conversation with The Air Current and Daedalean co-founder and CEO Luuk van Dijk.
Subscribe to continue reading...