During operations, railway systems tend to suffer from performance drops linked to unex-
pected events. These events often generate conflicts. Specifically, a conflict occurs when two
or more trains would claim a track section concurrently if they ran at the planned speed. In
case of conflict, trains need to slow down or stop to maintain a safe train separation, and this
leads to delay propagation. Dispatchers are in charge of traffic management. Their objective
is to prevent delay propagation by making sensible re-routing and re-scheduling decisions as
traffic evolves. This is formalized in the operations research literature as the real-time Railway
Traffic Management Problem (rtRTMP) [2]. In practice, dispatchers tackle real-time traffic
management in a centralized fashion, mostly relying on their experience alone. Also most lite-
rature addresses the problem in a centralized manner, proposing different optimization models
and algorithms. Conversely, the European project SORTEDMOBILITY [3] proposes a para-
digm shift : the traditional centralized envisioning of real-time railway traffic management is
replaced by multiple agents (the trains) self-organising and partaking in decision making. This
will bring increased flexibility and responsiveness to the decision-making process in perturbed
traffic scenarios.
In this work, we showcase the original design for the self-organization process proposed in
SORTEDMOBILITY. Furthermore, we provide a proof of concept on the proposed method's
applicability by considering a realistic case study, based on a French railway control area.