Leveraging AI to assess inequities in pavement maintenance and rehabilitation strategies
Abstract
Pavement maintenance and repair (M&R) is an important social need that should be accessible to all regardless of social or economic circumstances. In this project, pavement condition is analyzed through the use of machine learning algorithms and then compared to the socioeconomic factors of the surrounding communities. This analysis is crucial for assessing inequities that disadvantaged communities often face, especially pertinent toward pavement M&R policies. In addition to the goal of determining equity, a secondary goal of this project was to develop a methodology that can be repeated across any city. Utilizing an image-set provided by Google Street-View's API and census data of Kansas City, Missouri, a comparative analysis was conducted to determine whether equity between advantaged and disadvantaged groups was present. To aid in pavement distress identification, YOLOv5 (You only look once), a popular deep learning algorithm, was used to identify seven unique pavement distresses across Kansas City Road segments. The resulting methods were able to demonstrate a comparison between pavement conditions and socioeconomic metrics, demonstrating a trend indicating road segments in disadvantaged communities show slightly worse conditions. The strongest correlative factor borne out of analysis shows that median household income demonstrates the greatest gap between advantaged and disadvantaged census blocks.
Degree
M.S.