A University of Houston Professor of Civil and Environmental Engineering is using artificial intelligence to improve road safety by connecting datasets that are typically analysed in isolation.
In a study funded by the Texas Department of Transportation, Professor Lu Gao used large language models to analyse large-scale roadway condition data, including pavement structure, surface condition, roadway geometry and crash records drawn from police reports.
By bringing these data sources together, the study helps identify roadway segments where pavement or roadway conditions may be linked to higher crash risk.
Turning crash narratives into structured data
At the core of the research is the use of large language model-based crash narrative analysis to identify and quantify pavement-related crash risk. The LLM component converts unstructured police crash narratives into structured, mechanism-specific labels – such as hydroplaning or curve-related loss of control.
This approach enables the extraction of outcomes directly linked to pavement and roadway-surface conditions that are often missing from conventional structured crash data fields. These details tend to be embedded in free-text narratives, making large-scale extraction difficult through manual review or simple keyword rules.
“A case study using over 24,000 police crash narratives linked to a pavement management dataset of approximately 180,000 data records demonstrates strong associations between friction and texture measures and wet-pavement crash mechanisms,” Gao explained.
The research results have been published in the journal Accident Analysis and Prevention.
Pavement conditions and crash risk
The study focused on pavement conditions such as roughness and skid severity, which play roles in influencing both crash occurrence and severity. Prior research has found that rough pavement contributes to increased crash frequency and that skid resistance has a negative correlation with crash occurrence, particularly under wet conditions.
“Understanding how pavement conditions relate to crash outcomes can help inform segment screening and treatment selection, especially by identifying high-risk segments where elevated crashes are more strongly associated with pavement-related conditions and may therefore be responsive to pavement-focused treatments,” Gao added.
The goal of the research is to help transportation agencies determine which road segments are most in need of maintenance or safety improvements, allowing limited resources to be directed to locations where they may have the greatest impact on reducing crash risk.
Building on LLM applications in traffic safety
Gao noted that recent work in traffic safety has explored large language models for crash narrative understanding and structured information extraction, providing a foundation for the narrative-based analyses used in this study.
The research ultimately aims to help transportation agencies select candidate pavement-safety projects by identifying conditions associated with elevated crash risk and prioritising targeted, cost-effective countermeasures.
The University of Houston is a public research university that enrols nearly 49,000 students and serves the Houston and Gulf Coast region through faculty research, experiential learning and industry partnerships.
Last Updated on July 19, 2026 by Nick Ross
