Mixed Reality & Explainable AI for Firefighting
TU Wien × Imperial College London • 2025 - 2026
Master's Research / Visiting Researcher
Project Overview
Human-centred AI decision support for safety-critical mixed-reality environments. As part of my Master's research and a visiting research stay at Imperial College London, I developed and evaluated an explainable-AI-based mixed-reality (MR-XAI) decision-support prototype for firefighting scenarios. The work was part of an international research collaboration involving TU Wien, Imperial College London, the Stanford Center at Incheon and Sungkyunkwan University. The core challenge was delivering useful, AI-supported hazard information inside a safety-critical mixed-reality environment without overloading or misleading the user. I focused the design on trust calibration, reliability awareness, spatial situatedness and temporal stability, and evaluated the prototype through user studies with UK firefighting stakeholders. The research resulted in a Master's thesis and an accepted ISMAR XR-SPro workshop paper.
Challenges
- •Delivering useful, AI-supported hazard information in a safety-critical mixed-reality environment without overwhelming the user
- •Designing explanations that support trust calibration and reliability awareness rather than blind reliance on the model
- •Maintaining spatial situatedness and temporal stability of AI overlays in a live MR headset context
- •Running meaningful user studies with real UK firefighting stakeholders under realistic constraints
- •Coordinating an international, multi-institution research collaboration across time zones and disciplines
Key Achievements
- •Built and demonstrated an explainable-AI mixed-reality decision-support prototype on Meta Quest 3
- •Designed explainability around trust calibration, reliability awareness, spatial situatedness and temporal stability
- •Conducted user studies with UK firefighting stakeholders to evaluate the prototype
- •Completed a Master's thesis on MR & XAI for firefighting at TU Wien
- •Authored an accepted paper at the XR-SPro Workshop at IEEE ISMAR 2026 (accepted, not yet presented)