In this paper, a railway control algorithm is proposed for a coordinated stop of a convoy at stations in a virtual coupling system. The control algorithm uses a decentralized robust model predictive control in a novel application to railways of a predecessor-leader following communication topology. Thus, each train in the convoy takes its control actions based on different in-front-train references, obtaining one optimal solution that takes the train just ahead as a reference and a second optimal solution that takes the leader of the convoy as a second reference. Between the two optimal solutions, the final control strategy selects the minimum of the two control actions. The simulations of the control algorithm corroborate that this control algorithm enhances a coordinated stop of the convoy at stations, while it does not affect the behavior of the convoy in normal operation. The simulations pointed out that the algorithm might also affect the behavior of the convoy under perturbed conditions, but its real impact needs to be investigated. Future research lines include further communication topologies and centralized controllers exclusively involving the trains of the convoy linked through the communication topology.