A Local Search Algorithm for Train Unit Shunting with Service Scheduling

In this paper we consider the train unit shunting problem extended with service task scheduling. This originates from Dutch Railways, which is main railway operator in Netherlands. Its urgency stems upcoming expansion of rolling stock fleet needed to handle ever-increasing number passengers. The consists matching units arriving on a yard departing trains, scheduling tasks such as cleaning and maintenance available resources, parking trains tracks that can operate conflict-free. These different aspects lead computationally extremely difficult problem, combines several well-known NP-hard problems. paper, present first solution method covering all problem. We describe partial order schedule representation captures full local search algorithm utilizes ordering. proposed compared an existing mixed integer linear program computational study realistic instances provided by Railways. show our solve real-world complete It even outperforms current algorithms when considered isolation, is, without tasks. Although was developed for case it applicable any or location, irrespective its layout, uses self-propelling does not have passing trains.