Algorithmic Decision Support for Shunt Planning

In order to provide train services with a high quality, much coordination is required and a complex planning process is carried out. One of the last elements of this planning process is operational shunt planning. Shunt planning focuses on the logistics within a station and its surroundings. Since demand for transportation fluctuates over a day, a railway operator typically has a surplus of rolling stock outside the rush hours, and especially during the night. In general, the idle rolling stock is parked at a shunt yard, thereby keeping the main railway infrastructure available for other train services. Besides parking of rolling stock, matching of arriving to departing rolling stock, routing over local railway infrastructure, cleaning of rolling stock, and crew planning are part of shunt planning. Every change in a previous step of the planning process is likely to require changes in shunt plans at one or more stations. Therefore, many planners at NS Reizigers are currently involved in shunt planning. In addition, high-quality shunt plans enable a smooth start-up of the railway operations in the morning. A smooth start-up decreases the chances of disturbances in the morning. It is well known that such disturbances spread out easily in time and space. Therefore, the quality of shunt plans influences the quality of the services offered to passengers. The relevance of research on shunt planning from a societal, managerial and scientific point of view is therefore clear. ``Algorithmic Decision Support for Shunt Planning" introduces relevant aspects of shunting and provides a first step for quantitative models and algorithms to support shunt planning. The algorithms for solving the models contain algorithms that resemble the current practice of shunt planners as well as algorithms that are somewhat farther away from current practice. Computational tests on real-life data show that high-quality solutions are typically found within minutes of computation time. In addition, these algorithms are designed to interact with shunt planners. They provide a firm basis for an advanced planning system to support shunt planners.

Doctoral thesis download 

Contact information:

Olga Novikova

Email