Project Details

Abstract

This project proposes an agentic artificial intelligence (AI) framework for pavement assessment that brings together multimodal deep learning, depth-aware evidence, and standards-aligned reasoning within a single workflow. Although recent automated methods have improved distress detection, segmentation, and severity labeling, they still function largely as isolated tasks and therefore do not support the broader engineering reasoning required for maintenance decision-making. As a result, an important gap remains between model prediction and practical pavement assessment. Building on prior advances in pavement condition index estimation, dense captioning, and annotation-efficient segmentation, the proposed framework will detect and classify pavement distress, estimate severity and depth-related condition indicators, identify the appropriate ASTM-aligned assessment pathway, and generate clear engineering-oriented reports. In this way, the project connects perception, interpretation, pathway selection, and reporting in a coordinated system. Ultimately, it aims to move pavement AI beyond passive prediction toward a more interpretable, scalable, and decision-ready tool for transportation practice.

Project Word Files