Safety Assurance for Artificial Intelligence/Machine Learning in Safety Critical Airborne Systems

Details

• Creators:
  Sharma, Prashin ; McFarland, Chad ; Jahn, Matthew
• Corporate Creators:
  Boeing Company
• Corporate Contributors:
  United States. Department of Transportation. Federal Aviation Administration. William J. Hughes Technical Center
• Subject/TRT Terms:
  Artificial Intelligence Aviation Safety Contingency Management Critical Systems Machine Learning Reinforcement Learning Risk Analysis Run-Time Assurance Running Time Standardization Uncertainty Uncertainty Quantification Validation Values In Measurement

  More +
• Publication/ Report Number:
  DOT/FAA/TC26/2
• DOI:
  https://doi.org/10.21949/v3xp-7q86
• Resource Type:
  Tech Report
• Geographical Coverage:
  United States
• Edition:
  Final Report 9/2021 – 9/2024
• Corporate Publisher:
  United States. Department of Transportation. Federal Aviation Administration. William J. Hughes Technical Center
• Abstract:
  This report addresses the challenges of integrating artificial intelligence (AI) and machine learning (ML) technologies into aviation while ensuring safety and reliability. AI/ML applications in areas such as autonomous flight control and collision avoidance show great potential, but there is uncertainty around how to employ them and certify them confidently. This report evaluates current verification and validation (V&V) methods and explores new approaches to safety assurance for AI/ML-enabled systems. The project assessed AI/ML assurance techniques, implementation frameworks, and risk mitigation strategies, including run-time assurance (RTA) and contingency management (CM). It highlights the need for improved methods to manage uncertainties in AI/ML models. Recommendations include the development of new AI-specific standards and certification processes, along with continued research into AI/ML safety, particularly in human-AI interaction and real-time monitoring. The findings emphasize that while AI/ML technologies have transformative potential for aviation, significant advancements in safety assurance frameworks and certification processes are required to support their safe and effective integration into safety-critical systems.
• Format:
  PDF
• Funding:
  692M15-21-T-00035
• Collection(s):
  FAA William J. Hughes Technical Center for Advanced Aerospace
• Main Document Checksum:
  urn:sha-512:97a139acad6a4a19b2466f454d72fbb48c33b484717606df0c7bccb9dd7856059ccf9c7eace697c62220f1e392f6810c770515fa1e521dbd347a9b4712c7c365
• Download URL:
  https://rosap.ntl.bts.gov/view/dot/92273/dot_92273_DS1.pdf
• File Type:
  [PDF - 4.55 MB ]