PhD Defence B.T.C. van Rossum

Abstract

In many real-life settings, the size and complexity of work allocation problems necessitate the use of operations research (OR) techniques. Traditionally, OR models have prioritised efficiency objectives, such as cost minimisation. However, there is a growing interest in methods that ensure fair work allocations. This thesis applies OR techniques to design models and methods that achieve fairness in work assignments.

The first part of this thesis focuses on railway crew planning, addressing practical problems that arise in the proposed crew planning process of Netherlands Railways. The first study considers tactical crew scheduling, introducing a Benders decomposition approach for robust template selection. The second study examines fair operational crew scheduling under the assumption that template-based rosters have already been constructed. It proposes a tailored column generation heuristic to construct individual crew schedules that are fair over time. The third study presents an efficient exact pricing algorithm to accelerate column generation algorithms for basic railway crew scheduling problems.

The second part of this thesis takes a more theoretical perspective, developing general optimisation methods for fairness-oriented work allocation. The first study centers on fairness over time in settings where work must be assigned online to homogeneous workers. It provides theoretical and experimental justifications for using an intuitive work allocation policy. The second study investigates branch-and-price methods for minimising the range and other order based objective functions, introducing a generic branching rule that enables the use of classical, efficient branch-and-price methods for this type of problem.