Lecture Notes: AI in Aviation - Key Topics and Discussions (Day 1) - Various Speakers

Introduction

Focuses on how AI can support aviation safely.
Importance of collaboration across industry, regulators, academia, and researchers.
EASA (European Union Aviation Safety Agency) is leading in AI for aviation, working with the FAA (Federal Aviation Administration) and other stakeholders.

IPC Projects (Innovation Partnerships Contracts) between EASA and companies for innovation:
- 2020-2021: Computer vision-based runway & traffic detection with a Swiss startup.
- 2022: Formal verification methods for neural networks, working on maintenance applications with Collins Aerospace.
- 2023: Moving towards Level 2 AI, auto-taxiing projects with Boeing.

Research under Horizon Europe and the MLeap machine learning project led by Bundeswehr University Munich.
Emphasis on inter-disciplinary cooperation between different sectors.
AI applications range from anti-collision systems to natural language processing and camera-based automatic inspections.

AI in aviation certification projects initiated for simpler, general aviation aircraft to eventually cover larger aircraft.
2022: Published the first AI special condition SC AI-01 and the first EASA AI paper.
Common denominator: Trustworthiness framework is essential.
AI Concept Papers will consolidate in the rule-making task 0742.

Divided into two days; Day 1 focuses on the EASA AI roadmap and framework consolidation. Day 2 focuses on machine learning research projects and findings.
Importance of timelines and industry push for practical applications of AI.
Closing with ethical aspects of AI; ethics is crucial and led by EASA's PhD research.

Guiding Principles: Focus on safety, seeing AI as an engineering component, and using existing regulations where possible.
Learned AI vs Learning AI: Priority on learned AI due to its deterministic behavior. Learning AI is for future considerations.
Incremental approach: Starting with low-criticality applications to understand AI deeply before expanding.
Conformance Assessment: Conformance assessments for safety certifications remain a key consideration.

Scientific committee supports EASA on AI and Automation issues, per its annual reports.
Work Programs: Human-machine collaboration, ethics in AI, and levels of automation in various fields (e.g., aviation, automotive, rail, etc.).
Importance of human factors and ethics in AI safety and trustworthiness.

Research and industry consultations feed into AI Assurance objectives and concept papers.
Challenges: Reinforcement learning, explainability, continuous safety and security risk assessment including ethics-based assessments.
Future explorations include expanding the concept paper to cover hybrid AI, symbolic AI, and robust safety and security measures.

Alignment between EASA's concept paper and industry standards (e.g., EURO-CAE/SAE working groups on ML standards).
Emphasis on industry collaboration to create actionable, scalable AI standards in aviation.

Tales: Conflict detection and resolution, digital co-pilot assistance, and contrail mitigation using AI.
uAvionix: Human-AI teaming and decision-making assistance for pilots.
Deep Blue: Human factors in AI, intelligent assistant prototypes for aviation, and ethical training.
CAE: AI-driven performance assessment in pilot training with physiological monitoring.
Boeing: Automated taxi systems and real-time AI system demonstration.

Survey Insights: Comfort, trust, and acceptance of AI by aviation professionals; high emphasis on ethical deployment of AI.
Risk of De-skilling: Major concern among professionals, implications for training and competence maintenance.
Regulation and Responsibility: Importance of shared responsibility and accountability in AI-driven systems.

AI is a transformative force in aviation, requiring a balanced approach blending regulation, ethical considerations, and industry collaboration.
Future steps include continuous stakeholder engagements, further research, and iterative development of standards to align with evolving AI capabilities and challenges.
Ongoing efforts on harmonizing standards globally, particularly between EASA and FAA, aim for universal safety standards.