There is a growing demand of transport by train – be it freight or passengers. Expanding the rail infrastructure is costly, takes a long time to be realised and is often just not possible due to a lack of space. Based on a joint future vision of NS and ProRail the researchers of the Rail lab are researching ways to increase the efficiency and use the maximum capacity of the rail whilst at the same time make the system more robust and less fragile for disturbances. They are focusing at how all rail traffic at so called hubs in the network best can be managed with the use of combined expertise of human experts and AI. Besides the scheduled passenger & freight trains this includes shunting trains from and to a station, the choice for a specific shunting yard at a hub and shunting trains at yards, between and during maintenance and cleaning. 
The researchers are going to use AI to optimise how more trains can be moved to and from the yards and serviced and parked on a yard. One of the biggest challenges for the project is how the algorithms will deal with disturbances in the service. You don’t know when disturbances will occur, and if they do occur this is often on short notice. So how can they design an algorithm that will produce more robust schedules with room for errors or a plan B, whilst still be as optimal and efficient as possible?

At the end of the project the researchers aim to have a first prototype of a new generation of a planning support tool that the human planner can interact with. Additionally the researchers aim to see if the algorithms developed in this project can be applied for fitting-in trains that are outside the daily train schedule – for instance at the port where cargo trains are loaded for further transport of for example containers. The next step is a further integration of the scheduling processes of NS and executing processes of ProRail enabeling both organisations to quickly make decisions during an execution phase.
Other partners: This lab is part of the NWO Long-term project ROBUST AI, consisting of in total 17 of these labs.