KOI - Better Train Traffic Prediction

SICS researchers have developed a new method to predict delays in train traffic. With the use of traffic history based on GPS data and probabilistic calculations it is possible to estimate, with good precision, when the trains will arrive and change of trains can be made.

It has not escaped anyone that train traffic is depending on weather and other factors that are hard to predict. Still, so far, Trafikverket (the Swedish Transport Administration) has had no tool to estimate arrival time with any precision. SICS has addressed this challenge and developed a new method which will make it much easier to know when a train can be expected to arrive and passengers and cargo can be delivered.

When the train is on its route, every new position given by the GPS on the train, narrows the span of time when the train can be expected to arrive at the stations along the route. Making such information more reliable, precise and available is of high value for traffic control, train operators, cargo customers and passengers. Cargo customers and passengers can for instance get the information through an app and the Trafikverket can use it for a number of purposes. A specific goal of the project is to support train dispatchers with information such as the trains estimated future progression, the uncertainty in the trains future positions and the risks of spread of delays.

The approach to use historical data for building statistical models of train behaviours is new for making train traffic prognosis. The project has developed a method that compares the current traffic situation with historical data and estimates a continuous future traffic development together with an uncertainty at some given risk level. Besides prognosis this could also mean to support time tabling, detect anomalous train movements, comparing different dispathing measures and warn when an unwanted event is at risk. The predictions do not only apply to normal sitiations. Also in severely disrupted situations, such as a downed overhead line, the approach may prove successful.