Building the Data Infrastructure for Monitoring Rural Roads

Automobiles are the dominant mode of transportation in the United States, especially in rural areas, supported by the nation’s widely distributed road network as a critical component of infrastructure. However, driving on deteriorated roads not only imposes financial burdens on motorists but also reduces rural communities’ mobility and, consequently, their access to jobs and economic development opportunities. Moreover, more than half of traffic fatalities in the U.S. are linked to deficient roadway conditions, such as fading pavement markings, missing guardrails or safety barriers, and damaged traffic signs, resulting in substantial crash-related costs.

This project aims to develop the data infrastructure necessary to expand the use of sensing technologies and artificial intelligence (AI) methods for inventorying road assets and monitoring rural road infrastructure. A cost-effective multimodal sensing device will be prototyped to capture driving videos, GPS data, and vehicle movement information, all with synchronized timestamps. Using this device, sample datasets will be collected in rural areas, annotated, and made publicly available through a sustainable data hosting and sharing platform. Building on these datasets, domain adaptation methods will be developed to implement and tailor deep learning models for rural contexts, enabling applications such as road asset inventorying and road condition monitoring.