Approaches for AI/ML Based Airborne Systems Compliance
2022-26-1224
05/26/2022
- Event
- Content
- Abstract—Technologies such as Artificial Intelligence (AI) encompassing Machine Learning (ML), Deep Learning (DL) techniques have proved to be a game changer in multiple domains including aerospace. The avionics systems are becoming more complex, as software content continues to grow, pushing the cost of certification higher. The non-transparent nature of AI/ML systems further adds to this complexity which, throws new challenges in certification of safety critical systems. It is difficult to verify AI/ML design and to explain or interpret the behavior during the operation. This paper attempts to discuss simpler approach of showing compliance to AI/ML systems with the existing ARP4761, ARP4754A, DO178B/C, DO-254, and EASA/FAA guidelines. Further, this paper provides possible references and aids to evaluate the AI/ML based systems in aerospace. Keywords—Artificial Intelligence (AI), Machine learning (ML), Deep learning (DL), EASA, FAA, Compliance, Safety Critical System.
- Citation
- Paramasivam, P., Kumar, S., Shrivastava, R., and Sudalaimani, N., "Approaches for AI/ML Based Airborne Systems Compliance," SAE Technical Paper 2022-26-1224, 2022, .