An Automated Framework for Lane Closure Detection on Highway Using Connected Vehicle Data and Machine Learning Models

Abstract

Lane closures on highways present significant challenges, including traffic disruptions, increased crash risks, and economic losses. The traditional methods, which use primarily manual reporting or sensor-based methods, can be error-prone, inefficient, and costly. This study introduces an innovative real-time lane closure detection approach using connected vehicle (CV) data and machine learning techniques. Our methodology analyzes CV data metrics such as speed variations and lateral waypoint positioning relative to road reference lines, comparing these across road segments with and without closures. We employ two machine learning models—support vector machines (SVMs) and k-nearest neighbors (K-NN)—trained on features extracted from these metrics to detect lane closures and provide insights into location and time of start and end. This research was extended to encompass the entire state of Iowa, utilizing annual data to comprehensively assess lane closure detection capabilities across diverse geographical and traffic conditions, demonstrating its potential for scalability and broader implementation. Challenges encountered during state-wide implementation were addressed, proposing practical solutions to mitigate them. A visual dashboard was also developed to validate the models’ accuracy in detecting lane closures, aiding informed decision-making by DOT officials and other stakeholders. Our research highlights potential applications, including scalable solutions for accurate lane closure detection, driver alerts in connected cars, crash risk analysis, and support for naturalistic driving studies in lane closures. This data-driven method offers a cost-effective, real-time alternative to conventional detection methods. By advancing lane closure detection methods, this paper contributes to enhancing road safety and optimizing traffic management and catalyzes the evolution of autonomous vehicle technologies within modern transportation systems.