Traffic congestion is a problem that many people encounter almost every day, causing unnecessary delay, additional pollution and increased fuel consumption. Smoover is an intelligent transportation system which provides personalized speed advice to car drivers using a smartphone application, such that trafic congestion can be reduced. The project was carried out in collaboration with Cygnify BV, Traxpert, Locatienet and Tessa Bouw Communicatie.
In this project I developed a novel reinforcement learning algorithm which automatically learns speed limit policies, taking into account the current and predicted traffic conditions. The speed advice that is shown in the application has been derived from these policies. More details about the reinforcement learning algorithms can be found in my MSc thesis and a journal article. For my thesis I received the Ngi-NGN Informatie Scriptieprijs 2014 from the Royal Holland Society of Sciences and Humanities.
Traffic flow optimization: A reinforcement learning approach
Erwin Walraven, Matthijs T. J. Spaan and Bram Bakker
Traffic Flow Optimization using Reinforcement Learning (abstract)
Proceedings of the 26th Benelux Conference on Artificial Intelligence, pp. 211–212, 2014.
Traffic Flow Optimization using Reinforcement Learning
Master's thesis, Delft University of Technology, 2014.
The Smoover app and our work on traffic flow optimization with reinforcement learning appeared in several magazines and newspapers. In the list below you can find references and links to articles and interviews.
|Interview Dutch magazine NewScientist||2015|
|Article in the quarterly of study association W.I.S.V. 'Christiaan Huygens'||2015|
|Article 'Proactief filevorming verminderen' in magazine Informatie||2015|
|Article on the website of IEEE Spectrum Magazine||2015|
|Radio interview Niek van der Bruggen, Q-music||2014|
|Interviews Algemeen Dagblad, Leidsch Dagblad and TU Delta||2014|
The video below shows the Smoover system in operation. Smoover also appeared in RTL Nieuws on November 19, 2014. The second video provides some additional information (in Dutch).