Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards certifiable AI in aviation: landscape, challenges, and opportunities

Hymalai Bello, Daniel Geißler, Lala Ray, Stefan Müller-Divéky, Peter Müller, Shannon Kittrell, Mengxi Liu, Bo Zhou, Paul Lukowicz

TL;DR

This paper highlights the challenges of certifying AI development with an example to emphasize the need for qualification beyond performance metrics and presents a comprehensive mind map of formal AI certification in avionics.

Abstract

Artificial Intelligence (AI) methods are powerful tools for various domains, including critical fields such as avionics, where certification is required to achieve and maintain an acceptable level of safety. General solutions for safety-critical systems must address three main questions: Is it suitable? What drives the system's decisions? Is it robust to errors/attacks? This is more complex in AI than in traditional methods. In this context, this paper presents a comprehensive mind map of formal AI certification in avionics. It highlights the challenges of certifying AI development with an example to emphasize the need for qualification beyond performance metrics.

Towards certifiable AI in aviation: landscape, challenges, and opportunities

TL;DR

This paper highlights the challenges of certifying AI development with an example to emphasize the need for qualification beyond performance metrics and presents a comprehensive mind map of formal AI certification in avionics.

Abstract

Artificial Intelligence (AI) methods are powerful tools for various domains, including critical fields such as avionics, where certification is required to achieve and maintain an acceptable level of safety. General solutions for safety-critical systems must address three main questions: Is it suitable? What drives the system's decisions? Is it robust to errors/attacks? This is more complex in AI than in traditional methods. In this context, this paper presents a comprehensive mind map of formal AI certification in avionics. It highlights the challenges of certifying AI development with an example to emphasize the need for qualification beyond performance metrics.
Paper Structure (14 sections, 5 figures, 7 tables)

This paper contains 14 sections, 5 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Machine learning and artificial neural networks
  4. Avionics
  5. AI certification in Avionics
  6. Trustworthiness Analysis (TA)
  7. AI Assurance (AIA)
  8. Methods for data completeness and representatives
  9. Methods for AI generalization, explainability, and uncertainty assurance
  10. Human Factors for AI (HFAI)
  11. AI Safety Risk Mitigation (AIS)
  12. Certification Challenges of Classical AI Research
  13. Discussion and insights
  14. Conclusion

Figures (5)

  • Figure 1: General pipeline for developing Deep Neural Networks (DNN)
  • Figure 2: AI for avionics certification blocks according to the European Union Aviation Safety Agency (EASA) Artificial Intelligence Concept Paper: Guidance for Level 1 & 2 machine learning applicationsEASAConcept.
  • Figure 3: Objectives overview for a trustworthy analysis (TA) of artificial intelligence (AI) solutions in avionics
  • Figure 4: Overview of the learning assurance cycle using the European Union Aviation Safety Agency (EASA) W-shape model. The right side (blue color) shows the offline design of the machine learning (ML) and the left side (green color) represents the cycle towards online deployment.
  • Figure 5: Human factors for artificial intelligence (HFAI) overview