• Examines the safety of AI-enabled drone runway debris detection at airports. • Uses multi-criteria ranking to sequence safety actions for implementation. • Links recommendation ranking criteria directly to traced system safety findings. • Emphasises operator training, situational awareness, and human oversight. Unmanned Aerial Vehicles (UAVs) supported by Artificial Intelligence (AI) offer a promising capability for detecting Foreign Object Debris (FOD) on runways in non-segregated airport operations. However, they are also introducing new operational and safety challenges. This study conducts a systems-based hazard analysis for such operations using the System-Theoretic Process Analysis (STPA) method. The safety recommendations generated from STPA findings are then ranked to support sequencing and tiering using the PROMETHEE method as decision support, providing a transparent basis for implementation under practical constraints without compromising comprehensive hazard coverage. Recommendations were ranked using 15 criteria, 11 of which were operationalised from STPA artefacts (hazards, unsafe control actions and loss scenarios) to quantify risk contribution and traceability. This study identified 8 hazards, 149 unsafe control actions, 188 loss scenarios and 161 safety recommendations. The combined STPA and PROMETHEE analysis emphasises the importance of human-automation coordination, operator training, situational awareness and system-level integration in ensuring the safe deployment of FOD detection UAVs. Overall, the study demonstrates an STPA-grounded multi-criteria decision-making prioritisation workflow in which recommendation scores are derived from, and traceable to, STPA artefacts, enabling transparent near-term implementation planning under practical resource and operational constraints for AI-enabled UAV runway inspection.
Bounou et al. (Sun,) studied this question.