Data-driven national strategic traffic assignment models for road network congestion management

The national management of road congestion is a complex and multi-disciplinary challenge. In previous decades the solution was to build more capacity; however, in densely populated countries this is no longer an easy option due to the cost and environmental issues entailed. Pro-active traffic management is one key to improving the performance of the road infrastructure going into the future, when not only the economic productivity but also the environmental impacts of transport will be under increasing scrutiny. Models to analyse congestion need to be developed that can be effectively applied to large national networks within the constraints of accuracy, efficiency and data-privacy. This work seeks to investigate how to use the existing cross-sectional traffic data that highway authorities readily have access to for the creation of data-driven traffic assignment models. These models can assess the performance of key national road infrastructure and strategic interventions to reduce congestion. Such data is currently used to reactively manage traffic with action taken after congestion has started. This work first looks at extracting the building blocks of a data-driven model for the English motorway network. This includes a degenerate topographic representation via map generalisation. Techniques for the estimation of the key components of traffic assignment models are then developed to work with the data restrictions. The use of density-based road-specific congestion functions is proposed and compared to the state of the art to enable the efficient and accurate calculation of traffic patterns. A new technique utilising network modularity community detection is developed that divides the network and estimates the demand profile of its drivers from the measured road flows, reducing the network size restrictions of current approaches. Finally, the developed techniques are applied to a national strategic road network to evaluate network inefficiency from selfish driving and potential targeted intervention strategies.