Flexible Real-time Railway Crew Rescheduling using Depth-first Search

Abstract

Crew rescheduling is one of the most important factors one must consider during the recovery process following the disruption of a railway service. If it is not properly considered when the timetable is adjusted, it can jeopardise the prompt return to stable service. A new method, namely depth-first search crew recovery (DFSCR), is developed and proposed as a satisfactory way of dealing with the real-time crew rescheduling problem. DFSCR takes into account practical considerations of the railway environment and implements flexible, parameterised rescheduling constraints in the model to generate multiple solutions. The method has been tested on real-world scenarios and its capabilities are tested against two pre-existing methods. Sensitivity tests on five parameters (maximum permitted working hours, for example) are conducted and results indicate that small adjustments to the allowed ranges of parameter values can often generate acceptable solutions where there would have been none otherwise. This parameter relaxation is implemented via a feedback mechanism in DFSCR which takes results of a run of the algorithm to inform which parameters should be relaxed and by how much in order to maximise one's chances of finding a solution in the subsequent run of the algorithm.