Stefan Flügel (TØI) outlines a new research project on machine learning and traffic predictions (Machine learning for computational efficient predictions of long-term congestion patterns in large-scale transport systems)
Delays, lost leisure time, increased air pollution: Traffic jams are a source of significant socio-economic costs for companies and individuals in Norway, especially in the Oslo area. Urban planners need tools to evaluate where investments in road capacity are most effective in reducing traffic jams. A challenge with current tools (so-called strategic transport models) is that only highly specialized professionals can use them and that the calculation time can be up to several days.
In this project, we aim to build a new type of prediction tool. Our tool will be based on machine learning. In order to be able to apply machine learning for long-term predictions, we have to train the model on data where population and road capacity vary. Our idea is to establish such training data based on agent-based traffic simulation models.
The tool will be available open-access and it is expected to be much faster and much more user-friendly than current transport models.
We will start up in August 2021 and plan to conclude in August 2025. The project is a collaboration between the Institute of Transport Economics, the University of Bergen (UiB), the Swedish National Road and Transport Institute, the AI-company Epigram AS, as well as the Norwegian Public Road Administration. Our project includes a PhD-project at the department of informatics at UiB.