Real-Time Traffic Density Estimation Using Connected Vehicle Data
Real-Time Traffic Density Estimation Using Connected Vehicle Data
Jonathan Ron-Edoute, M.Sc. student at the department of Industrial
Advisor: Prof. Irad Ben-Gal
Abstract:
Accurate real-time traffic density estimation is essential for optimizing traffic management systems, reducing congestion, and enhancing road safety. This study addresses a critical gap in traffic density estimation by integrating machine learning (ML) models with connected vehicle (CV) data to create a scalable and robust solution. Unlike traditional methods dependent on costly stationary sensors or models constrained by physical assumptions, our approach combines CV data, geographic segment information, and stationary sensor data to achieve high accuracy with full spatiotemporal coverage. Using a Random Forest Regressor (RFR) as the primary model, we evaluated its performance in two experimental settings: known-segment validation and in-state generalization. The results demonstrate that the RFR model performs exceptionally well, achieving an MAE of 4.0 vehicles and an RMSE of 6.0 vehicles for segments shorter than 1500 meters, even under real-world conditions where CV penetration rates are as low as 13.25%. The model consistently outperforms traditional methods in low-penetration-rate scenarios while maintaining high spatial and temporal coverage, providing a significant advantage over existing approaches. This research highlights the practical utility of ML-driven traffic density estimation, offering a solution that is accurate, adaptable, and scalable. Our findings pave the way for advancements in real-time traffic management systems and underscore the potential for further research, including testing model robustness across diverse traffic environments, enhancing feature engineering, and extending applications to urban areas with dynamic traffic conditions.
Bio:
Jonathan Ron-Edoute, M.Sc. student at the department of Industrial Engineering department, Bachelor of Science in Industrial Engineering with great honors from Tel Aviv University as well. His professional background includes rich experience in data science and analytics, having worked as a data scientist and data analyst at Fiverr. Currently, he is conducting research in the field of Machine Learning, focusing on Real-time traffic density estimation using connected vehicle data. His expertise lies at the intersection of data-driven problem-solving and cutting-edge research in transportation systems.
