During operations a lot of things can go wrong. If the planning is robust,
small disturbances will have a minor effect in the operations. However, large
disturbances can never be taken into account in the planning. For instance, if
an accident happens at a certain line and no train traffic is possible anymore
for a certain period of time, decisions have to be made which trains need to
be canceled between which locations (e.g. the whole line or only partly). Of
course, this can have consequences both for the rolling stock circulation and
for the crew schedules. Notice that these problems have similarities with the
corresponding problems in the short-term scheduling phase, but the major
difference is that decisions need to be made in a short period of just a few
minutes: heuristic approaches are required. Moreover, there may be a lot
of uncertainty; both on the position of rolling stock and crews, and on the
duration of the disturbance.
In the railway industry there has not been much research on this topic yet.
However, in other modes of transport (especially airlines) there have been
some first attempts to solve real-time scheduling problems. An interesting
reference to this topic is Stojkovi´c & Soumis (2001).