Work Zone Traffic Management on Highways

The main focus of this project was on the application of Intelligent Transportation Systems to work zone traffic management on highways, for the purposes of promoting safety, relieving congestion, improving mobility, and minimizing negative impacts due to transportation. This research introduced the use of Reinforcement Learning, an Artificial Intelligence method from Machine Learning, to provide optimal control for a freeway-arterial or Express/collector corridor with work zones. It presented the methodology, development, and simulated testing and results of the Artificial Intelligence based computer agent, in preparation for real time implementation in the field. The approach focused on providing real time route recommendations that could be relayed to motorists through Variable Message Signing (VMS) in order to direct traffic effectively to minimize system wide delay and congestion due to construction. The developed system, named TMS-Can, uses a control strategy driven by the Reinforcement Learning Agent to provide traffic management capabilities and traveler information on traffic routing, and detours to drivers. With real-time information, drivers can be aided in route selection allowing them to optimize the available network capacity restrictions imposed due to work zone construction. Simulated experiments with the Learning Agent conducted for the study area in the morning peak period with existing traffic conditions have shown considerable improvement over the base case of current traffic status. The results of simulation tests have proved that a real time dynamic like the Q-Learning approach can provide much better control than that of the fixed diversion.