Proactive Congestion Management

This project is applicable to corridors, both with and without managed lanes. We intend to combine data from conventional sources, such as loop detectors and traditional probe-based data, with newer sources, such as Bluetooth and connected vehicle (CV) data, to identify conditions that signal impending congestion. The objective is to forecast the likely occurrence of both recurrent and non-recurrent congestion. We will apply machine-learning techniques to produce data-driven models that rely on near- or real-time traffic measurements capable of generating predictions proactively based on complex and often subtle factors that trigger congestion. We believe that exploiting the advances in big-data science will result in success, whereas other efforts with similar objectives (but relying on more conventional analysis methods and data sources) have largely failed. We will design and test traffic control strategies that might mitigate the identified triggers of congestion or delay the onset of congestion, thereby reducing its duration and impact. Strategies considered will range from driver alerts to ramp metering, and CV-based variable speed limit and speed harmonization advisories.