NASA Logo, National Aeronautics and Space Administration

+NASA Home

+Ames Home

Guillaume Brat to Present in the Next SAE G34 — AI in Aviation — on May 20, 2021
Intelligent Systems Division Banner

Guillaume Brat to Present in the Next SAE G34 — AI in Aviation — on May 20, 2021

A few weeks ago, the European union Aviation Safety Agency (EASA), which handles certifications, released their first usable guidance for Level 1 Machine Learning (ML) applications. This is the first document contributing to the certification of a system with some ML-based components. Currently, it is restricted to supervised machine-learning systems that have been trained offline, not those adapted in deployment. There are some differences with the current trend coming out of the Federal Aviation Administration (FAA). For example, there is no recommendation to use runtime monitoring even though the FAA sees runtime monitoring as the first step towards the deployment of ML-based systems. Despite these differences, it is an opportunity to see how NASA research on software assurance for autonomy in aviation fits in the context of these first “certification” standards and guidance. Dr. Brat will do so in the context of talks at SAE G34 — Artificial Intelligence (AI) in Aviation — which focuses on drafting standards for certification of ML-based systems in aviation.

BACKGROUND: NASA has been working for the past 12 years on software tools for the assurance of software in aviation-critical systems. For two years now, NASA has focused more on the use of AI-based techniques in aviation than the traditional software systems used in the past. The primary focus has been on ML, and more specifically on supervised ML systems that have been trained offline. NASA’s research has been driven by case studies such as a vision-based centerline tracking system (implemented using deep neural networks) and the new generation of collision-avoidance systems developed under FAA guidance, i.e., the family of Aircraft Collision Avoidance Systems 10 (ACAS-X) products. Since EASA has recently released its first usable guidance for Level 1 machine learning applications, this is an opportunity to see how NASA research is mapping to this initial ML guidance. In his talk, Dr. Brat will use the EASA guidance document as a guide to present the past, present, and future of tools and techniques being developed at NASA. The intent is to not only provide an overview of the NASA research effort, but also to see how this effort is addressing the concerns listed in EASA’s first usable ML guidance.

POINT OF CONTACT: Guillaume Brat, guillaume.p.brat@nasa.gov

First Gov logo
NASA Logo - nasa.gov