As airborne systems become more and more complex – partly in order to “off load” aircrews, the certification of such systems becomes even more challenging, as the emphasis shifts from the complexity of human behavior to the even greater complexities of systems and software. More complex and sophisticated techniques, such as Artificial-Intelligence (AI) / Machine-Learning (ML) are now making their first steps into aviation safety-critical systems, both airborne and on the ground, and this trend is matched by new certification requirements presented by worldwide regulators such as the FAA and EASA.
In response to these emerging industry and regulatory needs, an entirely new certification eco-system is now emerging, starting with direct papers from EASA and the FAA, proceeding with the SAE G-34 & EUROCAE WG-114 committees, currently developing new “means of compliance” for aviation AI/ML applications. These committees have already produced a statement of concerns (AIR6988/ER-022) and a taxonomy document (AIR6987/ER-027), as a prelude to the means of compliance document currently being developed (ARP6983/ED-324), and additional supporting projects are developed in parallel.
