Predicting traffic

Nearly all of the vehicle traffic in the Netherlands is registered continuously. In the computer lab of Prof. ir. Hans van Lint, researchers are using this source of big data for mathematical models that describe and predict traffic.

“In the DiTT lab (Delft integrated Traffic & Travel Laboratory), we are collecting as many data as possible on traffic and transport to serve as a foundation for simulation models of various scales’, explains the Professor of Traffic and Transport Simulation. The data that Van Lint is collecting come from various sources, including the induction loops in the road which measure the presence and speed of vehicles. ‘Another source is data from the vehicles themselves. More and more cars are equipped with loggers, which transmit data periodically. This functionality is also included in many apps, including the popular Flitsmeister’.

The DiTT lab is also the research laboratory for the the National Data Warehouse for Traffic Information (NDW). All data from the road network in the Netherlands are brought together in this context. In addition to the data from the loops, it includes data on road construction and the locations of accidents, as well as data on weather conditions and major events. ‘This should yield smart tips that would be useful to road authorities, municipalities, provinces and services providers (such as the ANWB)’.

Van Lint gives the example of a province that would like to improve traffic safety. ‘They first want to identify accident hotspots and the factors that are associated with them. These could include events or local weather conditions, possibly even in relation to each other. In the ideal situation you would have a map containing all data in a form that can be run backwards and forwards in order to determine how such situations emerge and how conditions change during the situation. This is what we will be building’.

In the meantime, the researchers have learned how traffic flows and how tailbacks are formed. However they have less insight into motorists. ‘Social media can provide contextual information, as can observation systems in vehicles. These systems record acceleration, braking and steering behaviour, and they film what the driver is looking at. These data can be used to create predictive microscopic traffic models. The combination of these models with studies of behaviour in a driving simulator can contribute to the improvement of theories and models of driving and travelling behaviour’.

The DiTT lab is receiving support from the ICT company CGI, which provides and configures the necessary hardware and software. ‘We have several racks of servers here’, explains Van Lint. ‘The server that we will be using next year is in a refrigerator at CGI. It is an IBM Netezza, a monster of a machine, with 256 processors coupled to dedicated hard drives. That thing is fast enough to run experiments on all data from the road network in the Netherlands for an entire year in no time at all’.

Van Lint has applied for funding from the European Research Council (ERC). ‘It would be great if I could receive that grant. Then we’d be able to couple two driving simulators to the system, and I’d have enough money to add more doctoral candidates’.

Photo © Sam Rentmeester

Photo © Sam Rentmeester

Stay informed about the research

Receive the Delft Outlook newsletter 4 times a year