Doctor of Philosophy (PhD)


Construction Management

Document Type



Retroreflectivity plays a crucial role in pavement markings as it enhances nighttime visibility for drivers. Yet, due to budget constraints, many U.S state agencies rarely monitor retroreflectivity of their markings, and instead, restripe markings based on visual inspection or fixed schedule. Addressing this, Federal Highway Administration (FHWA) introduced a new final rule, effective September 6, 2022, that established a new minimum standard for pavement marking retroreflectivity. This rule requires state agencies to implement a method within four years for maintaining pavement marking retroreflectivity at or above minimum levels. The ultimate objective of this study was to assist state agencies comply with this new rule by developing a new decision-making tool for predicting retroreflectivity of pavement markings for up to three years using only measured initial retroreflectivity and other readily available project data.

To fulfill this objective, transverse skip retroreflecyivty (RS) data and other key variables from National Transportation Product Evaluation Program (NTPEP) were retrieved. Six robust machine learning algorithms (Decsion Tree, Random Forest, Categorical Boosting, Extreme Gradient Boosting, Artificial Neural Network, and Deep Neural Network) were employed to develop a total of 66 models in Python. Findings indicated that Random Forest achieved the highest accuracy among six algorithms. As expected, prediction accuracy decreased over time. Within the first year, Random Forest models predicted retroreflectivity with a coefficient of determination (R2) of 0.97 and Root Mean Square Error (RMSE) of ±31.12 mcd/m2/lux. Within the second and third years, R2 and RMSE were 0.92/±43.7 mcd/m2/lux, and 0.91/±36.0 mcd/m2/lux, respectively, which are more accurate as compared to previous studies. These models were calibrated utilizing the long-line retroreflectvity (RL) data, collected from test decks in the vicinity of Louisiana State University (LSU) campus.

An implementation-ready decision-making tool was developed incorporating the Random Forest models. In this tool, Python was integrated with Visual Basic for Applications (VBA) Macros within Microsoft Excel. With this tool, the users can practically employ the proposed machine learning models effortlessly without any expertise in coding software. Overall, the proposed tool is expected to serve state agencies as a preliminary guide to schedule future restriping activities for their marking products.



Committee Chair

Marwa Hassan

Available for download on Thursday, October 29, 2026