The focus of this study was on the development of functional and system requirements for an integrated, realtime temporary condition Traffic Management System for work zones on Canadian roadways (TMS-Can). TMSCan uses a combination of electronic sensors, software, wireless and cable communications networks, and electronic signs and variable message signs to manipulate traffic flow. The travel information on traffic routing displayed by these variable message signs is computed by a computer agent driven by Reinforcement Learning. Reinforcement Learning is an approach whereby the control agent directly learns to map sensed system states to optimal actions. A simple but powerful reinforcement learning method known as Q-learning is used. A micro simulation tool – Paramics – was utilized to train the agent in an offline mode within a simulation environment in order to make it ready for field implementation. The approach developed in this research was rigorously evaluated under simulated conditions. Results from the simulation are very encouraging and have demonstrated the effectiveness and superiority of the technique in reducing congestion in work zones. Deployment of the developed system at the field trial stage was not completed under this study. (Author/publisher)
Abstract