Network-wide Traffic Speed-Flow Estimator

Traffic congestion is an everyday occurrence of city. To monitor and improve the road traffic conditions efficiently, a research team aims to develop a network-wide traffic speed-flow estimator for provision of real-time traffic speeds and flows simultaneously for the entire study network by time of day. A model-based data-driven approach is proposed in the project, a new traffic assignment model that captures queuing effects, network uncertainties and temporal variations of traffic flows and speeds, with integration of traffic fundamental diagrams and traffic model results into the deep learning methods, to estimate the network-wide traffic speeds and flows, especially for the unobserved links without any traffic detector data. 

The real-time traffic information generated by this estimator is necessary for the development of Intelligent Transportation Systems (ITS). Route guidance service providers can use the estimation of traffic speeds to recommend drivers the optimal paths based on different criteria, such as maximizing on-time arrival probability in congested road networks. The authority can examine the effects of different traffic management measures through monitoring the traffic flows. Both public and private sectors can make use of this estimator to carry out traffic impact assessment for justification of their proposed development projects. 

Furthermore, a multi-type traffic detector location model will be developed for determining the optimal number, location and type of traffic detectors with consideration of multi-period traffic conditions, in order to suggest where and what type of the detectors should be installed and how many detectors are required for estimating efficiently the network-wide traffic speed and flow by time of the day. In the long run, the installation costs of the traffic detectors for ITS and traffic management could be reduced significantly.