As the relevance of rail transport continues to grow, efficient operational planning becomes increasingly critical. This is also evident at stations, where unused trains must be moved to the yard and returned to the platform when needed. To achieve this, available drivers must be assigned to trains in a way that respects the time constraints of their shifts. This thesis explores the approach of combining the shunt routing problem (SRP), where times for routes of shunt trains are calculated, with the drivers' problem, which deals with creating feasible driver schedules. The resulting Constraint Programming (CP) model is based on an existing CP model for the SRP. This combined model is tested on four stations in the Netherlands and solves the problem in under 200 seconds for the Vlissingen and Enkhuizen stations, while higher runtimes for Enschede and Amersfoort are observed.
Experiments with an alternative model structure highlight how the linking of the train and driver variables influences runtime significantly. Several variable and value selection strategies are examined on multiple variables, improving performance for certain instances. Furthermore, the analysis reveals that runtime is highly dependent on the station and instance characteristics.
By combining the SRP with the drivers’ problem, this thesis delivers a functional CP model applicable across all stations and contributes to a deeper understanding of the integration of the two problems.

The Shunt Routing Problem (SRP) is an important logistic problem at nearly every railway station. It is a subproblem of an even bigger train scheduling problem, the Train Unit Shunting Problem (TUSP). The objective of SRP is to schedule conflict-free routes between the platforms and the yards, for currently non-operational trains that have completed their current journeys and have not yet commenced the next one.
We developed a Constraint Programming (CP) model for the Shunt Routing Problem at NS. The model is intended to replace the current algorithm for constructing the initial solution for the Local Search, addressing the TUSP at NS. The model incorporates all essential feasibility requirements and produces conflict-free solutions, in contrast to its forerunner. Moreover, it can be applied to any instance and station layout.

Initially, the model exhibited an average performance of 7-8 minutes on a real-life instance at Enkhuizen station. We managed to enhance the model and reduce the computation time to less than a second. The key improvements were achieved by changing the solver search strategy, imposing additional search guidelines, or further restricting the domains of influential variables. During experimentation, an optimized second version of the model was proposed.
The evaluation undoubtedly confirmed that it outperforms the first version.

Both versions of the model, without any additional enhancements, perform exceptionally well on a second real-life instance at Amersfoort station, solving it in milliseconds. Even with a considerable number of additional routes, the performance of the second version remained reasonable, with a computation time of 18 seconds.

A final experiment revealed the remarkable performance of Gecode, achieving computation time in milliseconds with the second version of the model implemented for Enkhuizen in MiniZinc. This performance is an order of magnitude faster compared to the default performance of Google OR-Tools on Enkhuizen with the second version.

The planning of railway operations is a very complex process, in which timetables and other logistical plans need to be both reliable and robust to effectively prevent and cope with disturbances. However, research in the robustness of initial stabling plans, designed in an earlier planning phase, during the following planning phases, has been lacking, with the models created in research in the Train Unit Shunting Problem (TUSP) being generally deterministic in nature, even though stabling plans could quickly become infeasible when the stabling demand in the form of arriving and departing trains changes.
This thesis therefore proposes a definition of the robustness of an initial stabling plan to changes in the stabling demand, such as changes in train lengths, as well as provide an assessment method of said robustness. The created Robustness Assessment Model (RAM) first generates an initial stabling demand and stabling plan, then performs a Monte Carlo simulation to generate a set of stabling plans created for variations of the initial stabling demand, based on changes in stabling demand such as train length. Finally, the RAM estimates the robustness of the initial stabling plan by analysing the differences between the initial and these generated stabling plan variations and how efficiently the initial plan is able to change to these plans to optimally facilitate the variations of the initial stabling demand.
Running this model across three locations, each with three capacity utilisation scenarios, has shown that the model is able to estimate the robustness of an initial stabling with acceptable confidence, and furthermore is able to give insight into how capable the stabling plan creation method is in generating a robust solution. Furthermore, the RAM can also be used to investigate patterns in stabling plans which could predict the robustness of said stabling plans. Current shortcomings to the RAM are its relatively long runtime, the simple stabling plan creation process used in the RAM, and that only one-sided stabling yards without servicing scheduling have been incorporated. Further research is therefore recommended for extending the RAM with other stabling yard layouts and the scheduling of services, as well as improve the stabling plan generation process to take more constraints into account.