Strategies for delay management in scheduled railway traffic

[leenasuhl]

Unavoidable disruptions induce delays in timetable-driven rail traffic and thus imply the necessity of operations control and dispatching tasks. There are many reasons for disruptions, such as a missing signal, weather, and problems with organization of traffic, for example maintenance planning. In such a case often the original schedule cannot be executed, and dispatchers have to make decisions about changes to the schedule or resource assignments often in a matter of minutes.

In case of a train delay because of a disruption, a connection conflict occurs if the feeder train arrives too late to reach the connecting train on time. In case of such a conflict the responsible dispatcher has to make a decision if the connector train waits for the delayed feeder train. If a connector does not wait, all transit passengers of this specific connection will be delayed, and if it waits, all passengers in the connector get delayed and face the risk of missing their connections at following stations.

With dispatching strategy we mean every algorithm capable of determining dispatching decisions that consist of waiting instructions for single connecting trains waiting for late feeder trains in case of a connection conflict. For example, two basic strategies are “Always Wait” and “Never Wait”. Always Wait means that passengers always make their connections because each train waits for all delayed feeder trains. Never Wait denotes the opposite case where all trains always leave on time, regardless whether there are delayed transit passengers or not. From a practical point of view, these two are extreme strategies, and the strategies that are relevant for practice are somewhere between them.

In an earlier study we have compared rather simple thumb rule strategy based on waiting time rules and the so called ReOpt strategy that reoptimizes the schedule after each disruption [Kliewer, Suhl 2009]. We have shown that the thumb rule strategy usually is better than ReOpt, but that some more advanced strategies taking the passenger numbers into account are better than the thumb rules.
The thesis continues this research in cooperation with Deutsche Bahn AG. Data is given or gathered about delays, passenger numbers, maintenance, wagon types and train structures, as well as the distributions of these numbers. Furthermore, information is needed about costs involved, degrees of freedom, business processes, time restrictions, sizes of affected areas, and so on. The decision problem is to be formalized as an optimization model. Involving stochastic components and scenarios into the model is to be evaluated.
The goal of the thesis is then to develop and evaluate decision support methods in order to minimize the effect of disturbances according to an adequate measure. Algorithmic methods from Operations Research, such as Mathematical Optimization, Heuristics and Metaheuristics, Constraint Programming and Simulation will be evaluated according to various criteria. The best methods are to be implemented for a decision support system and tested for real-life data in passenger and freight traffic.

Reference:
Kliewer N., Suhl L.: A Note on the Online Nature of the Railway Delay Management Problem. Manuscript, accepted for publication in Networks, 2009.