“In 15 minutes, there is a 90% chance that the Turku motorway to the west will be congested at Veikkola.” Such congestion forecasts may soon be common in Finland,
which aims to be a leader in traffic automation. In order to achieve this goal, the Ministry of Transport and Communications has prepared a roadmap, and one of its key goals is to increase the use of transport knowledge capital and data. One of the projects related to the goal is a traffic congestion and traffic emergency forecasting project, for which the Finnish Transport Agency selected Loihde Advance as its analytics partner. The project was carried out in cooperation with Digia.
“In cooperation with Loihde Advance, we have developed a forecast model that shows fairly reliably how the average speed of vehicles develops on the roads. The test phase is now underway, but the preliminary results are promising,” says Pekka Kinnunen, Analytics Specialist at the Finnish Transport Agency. With promising results, Kinnunen means that the program was able to predict vehicle speeds, for example, in the tested sections with a probability of more than 99%.
In the test phase, models were created for two TMS points on Ring III in both directions of traffic. Open data was used for congestion modeling and forecasting. TMS data collected by the Finnish Transport Agency, traffic weather station data, and the calendar serve as the source data. The TMS point registers vehicles that cross a certain point, providing the time of passing, the direction of travel, lane, speed, length of the vehicle, the time difference between consecutive vehicles, and the vehicle category of each vehicle.
Using linear regression analysis, the data is used to predict the change in the average speed of traffic 15 minutes into the future. The change determines whether the road section in question is becoming congested. The long-term average speed for each day of the week and time of the day can also be calculated. Congestion forecasts are made almost in real time. The results are stored in a database and visualized through the MS PowerBI user interface implemented by Digia.
Regular congestion can be modeled fairly reliably, but the aim is that irregular traffic circumstances, such as collisions, can also be detected quickly through the application. When congestion can be predicted, efforts can also be made to prevent them. Traffic congestion can be communicated with changing signal boards and traffic can be directed to alternative routes. Efforts can also be made to prevent congestion through light control or changing speed limits. More accurate congestion forecasts mean smoother and safer traffic.
The pilot project opens up interesting opportunities for further development. “We are now modeling TMS-point-specific models, but in the future, the process should be developed so that we have a more general model that could be utilized at every point. The ideal situation for us would be to have a generic self-learning model that would become more precise as location-specific data is collected,” says Kinnunen.
In addition to the traffic centers, citizens may also be able to take advantage of better congestion forecasts provided by the project in the future, for example, through the Finnish Transport Agency’s Traffic Situation service.