Thesis Summary
Global road networks are under growing strain, yet conventional pavement assessment practices remain limited: they often emphasize surface distresses, under-represent subsurface health, and face practical barriers to deploying advanced sensing and automated analytics at scale.
This PhD research develops an integrated, multi-sensing, non-destructive framework that combines UAV and LLIS imagery for surface characterization with GPR for subsurface evaluation. Within this framework, automated, data-driven models are developed for four critical distresses: surface cracks, rutting, layer thickness deficiency, and subsurface cracks.
Key methodological advances include data-centric crack assessment strategies (transfer learning and generative AI) to reduce annotation effort, harmonization of rutting measurements across sensing systems via empirically validated conversion equations, and automated GPR interpretation using statistical and deep learning models. Subsurface conditions are quantified through two new indices: the Thickness Condition Index (TCI) and the Subsurface Crack Index (SCI).
The research further introduces two Integrated Pavement Performance Models (IPPM) — the Integrated Deductive-Based Performance Index (IDPI) and the Integrated Fuzzy Performance Index (IFPI) — which synthesize surface and subsurface indicators across five performance criteria (rideability, safety, user costs, deterioration rate, and environmental impact) using FAHP. Large Language Models are leveraged both to automate qualitative PASER-style assessment (Auto-PASER) and to help structure expert judgement for model weighting.
Research Context
