Big data processing and storage framework for its: A case study on dynamic tolling

Urban and national road networks in many countries are severely congested, resulting in increased travel times, unexpected delays, greater travel costs, worsening air pollution and noise levels, and a greater number of traffic accidents. Expanding traffic network capacities by building more roads is both extremely costly and harmful to the environment. By far the best way to accommodate growing travel demand is to make more efficient use of existing networks. Portugal has a good but underused toll highway network that runs near to an urban/national road network that is free to use but congested. In choosing not to pay a toll, many Portuguese drivers are apparently accepting greater risk to their safety and longer travel times. As a result, the urban/national road network is used far more intensively than projections anticipated, which raises maintenance costs while increasing levels of risk and inconvenience. The main idea behind the work presented here, is to motivate a shift of traffic from the overused network to the underused network. To this end, a model for calculating variable toll fees needs to be developed. In order to support the model, there is the need to accurately predict the status of road networks for real-time, short and medium term horizons, by using machine learning algorithms. Such algorithms will be used to feed the dynamic toll pricing model, reflecting the present and future traffic situations on the network. Since traffic data quantity and quality are crucial to the prediction accuracy of road networks' statuses, the real-time and predictive analytics methods will use a panoply of data sources. The approach presented here, is being developed under the scope of the H2020 OPTIMUM, a European R&D project on ITS.