|Ph.D Student||Mansour Omar|
|Subject||Traffic Modelling with Application to Dynamic Tolling|
|Department||Department of Civil and Environmental Engineering||Supervisor||Professor Tomer Toledo|
|Full Thesis text|
Dynamic network loading (DNL) models have been widely used in the fields of economics, transportation and environmental engineering as tool to evaluate long-term and short-term plans in urban areas. With the technology advancements, DNLs are used to estimate and predict traffic conditions in real-time to make appropriate control actions, where both computational timing and accuracy should be considered.
The most notable parameter for indicating congestion and level of service in transportation networks is travel time. This study explores the travel time dynamics for a network scale and shows the correlation with the LWR theory. Based on the new perspective a DNL based on travel time is established. The model incorporates the travel time dynamics model to advance the traffic through a link and a node model to distribute the traffic in junctions. The simulation model is tested on various case studies and its performance is reviewed.
An additional DNL approach is developed in this study. This approach simulates traffic dynamics by tracking the changes in the states of the nodes in the network. Each change in the node’s traffic state generates information packages (IP) that travel through connected links to neighboring nodes. These IPs propagate in different velocities depending on the type of information they carry and the traffic state within the link. The developed model is tested on various case studies and compared with the travel time-based DNL. Both are theoretically identical and show similar behavior in simulating the traffic state.
Finally, a real-time simulation-based control framework is developed to determine dynamic toll rates in order to optimize an operator’s objective, subject to various operational and contractual constraints, such as smooth toll rate changes and maintaining prescribed levels of service on the toll lane. Using the DNL models as an engine, the process generates a dataset of optimal tolling trajectories for various traffic scenarios, each action in the trajectories is linked to a measurable previous traffic conditions. A flexible function (policy function) that links the optimal action to the available data is fitted (trained) using the dataset. This process incorporates models to predict the drivers’ choice whether or not to use the toll lanes as a function of the toll rate and travel times presented to drivers within the information system and a DNL model to predict the resulted traffic conditions for each tolling strategy. A case study demonstrates the use of this framework and its potential to generate useful toll settings system.