Reproduce “A Probabilistic Digital Twin of UK En Route Airspace for Training and Evaluating AI Agents for Air Traffic Control” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Nick Pepper, Adam Keane, Amy Hodgkin, Dewi Gould, Edward Henderson, Lynge Lauritsen, Christos Vlahos, George De Ath and 6 more
The paper presents a probabilistic digital twin of London Area Control Centre en-route airspace for testing AI decision-support agents. It combines data, a physics-informed aircraft model, probabilistic trajectories, metrics, and controller review. For descending B738s in a week, mean-mode MAE was 16.77 versus 22.95 kt for CAS and 566.05 versus 1,009.22 ft/min for ROCD against BADA, validating the twin, not agent safety.
Key insights
- Combining historical operational data with a probabilistic, physics-informed machine learning model allows for accurate reproduction of real-world traffic scenarios and their inherent uncertainties.
- A structured assurance case, following the Trustworthy and Ethical Assurance (TEA) framework, provides quantitative evidence for the Digital Twin's accuracy and fidelity.
- The Digital Twin forms a unified environment supporting AI agent development, benchmarking, and competency-based human-in-the-loop assessment by qualified Air Traffic Control Officers (ATCOs).
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.