Since the dawn of the nineteenth century, development of railway systems has taken a huge importance in many countries. Over the years, the number of trains, the number of tracks, the complexity of networks increase and are still increasing. Directing trains on efficient routes, stopping and cancelling them are some actions that railway operators must take in their everyday life in order to regulate the traffic. However, with its continual growth, the consequences of such actions become rapidly hard to predict. Bad decisions can lead to disastrous situations such as accidents or, in the best cases, to unnecessary delays leading to financial losses. Decisions and actions that could be taken manually in the past are now hard combinatorial problems that require computer based methods for their solving. In this context, the need of a reliable and efficient railway traffic management is crucial. Like any transportation system, three aspects must be considered: safety, availability and fluidity. Safety and availability belong to verification engineering while fluidity is related to optimisation. A plethora of research on this field already exist. However, most of it suffers of a lack of scalability. They can only be used for small or medium stations. This thesis presents innovative approaches for tackling this problem. For each aspect, we propose a method, that is feasible in practice for stations of any size. Concretely, verification of safety is performed with a dedicated algorithm while availability is verified with Statistical Model Checking. Fluidity optimisation is carried out with Constraint Programming. The performance of these methods are analysed through three stations of the Belgian railway network. (FSA - Sciences de l'ingénieur) -- UCL, 2017