Scheduling is the process of allocating resources to activities over time. In a scheduling problem, resources are scarce and constrained in various ways (e.g. in the order of activities and the capacity of resources), and one is looking for a schedule of the activities that both satisfies the constraints and is optimal according to some criteria. This paper deals with an extension of a constraint-based scheduling model of real time management of train traffic through junctions and large stations. This extension allows the improvement of the model of conflicts between trains running in opposite directions with the use of state resources. To tackle large instances, a temporal decomposition has been used. We tested the model and algorithms by applying to problem instances of a real terminal station. Computational results show that with the temporal decomposition method we are able to obtain good solution in short computation time