Dr. Nourhan Bachir Successfully Defends Her PhD on AI-Driven Criticality in Urban Traffic Networks

Dr. Nourhan Bachir’s PhD Defense
Date of Defense: September 18, 2025
Thesis Title: “Criticality as a Learnable Property: Scalable AI Techniques for Critical Link Analysis in Urban Traffic Networks”
Supervisors: Roland Billen (ULiège, Belgium) and Chamseddine Zaki (AUM, Kuwait)

Nourhan Bachir’s thesis fundamentally reframes road-link criticality as a learnable property derived from data, rather than a fixed index. Her work introduces a robust framework that integrates spatial features, graph analysis, stress-tested simulation signals, and novel trajectory-based movement patterns.

The work advances a powerful, modular pipeline through three core contributions: PEMAP, a post-event traffic redirection framework (simulation, GIS, AI, VANETs); VeTraSPM, a trajectory-mining approach introducing SFI, CIS, and SIS to capture behavioral criticality; and SMaL-CLIP, a scalable ML pipeline that trains on a small subset of the data to accurately predict the rest, demonstrating strong cross-city generalization.

Collectively, these methods improve reliability, scalability, and generalization in critical link analysis and support more resilient, data-driven urban mobility planning. 

The thesis demonstrates how moving from static, formula-based indices to learning criticality from data enables efficient and transferable solutions to real-world network disruptions.

Access the full thesis: Criticality as a Learnable Property