Airline reserve crew scheduling under uncertainty

This thesis addresses the problem of airline reserve crew scheduling under crew absence and journey time uncertainty. This work is primarily concerned with the allocation of reserve crew to standby duty periods. The times at which reserve crew are on duty, determine which possible crew absence or delay disruptions they can be used to absorb. When scheduling reserve crew, the goal is to minimise the expected levels of delay and cancellation disruptions that occur on the day of operation. This work introduces detailed probabilistic models of the occurrence of crew absence and delay disruptions and how reserve crew are used to absorb such disruptions. Firstly, separate probabilistic models are developed for crew absence and delay disruptions. Then, an integrated probabilistic model of absence and delay disruptions is introduced, which accounts for: delays from all causes; delay propagation; cancellations resulting from excessive delays and crew absence; the use of reserve crew to cover such disruptions given a reserve policy; and the possibility of swap recovery actions as an alternative delay recovery action. The model yields delay and cancellation predictions that match those derived from simulation to a high level of accuracy and does so in a fraction of the time required by simulation. The various probabilistic models are used in various search methodologies to find disruption minimising reserve crew schedules. The results show that high quality reserve crew schedules can be derived using a probabilistic model. A scenario-based mixed integer programming approach to modelling operational uncertainty and reserve crew use is also developed in this thesis and applied to the problem of reserve crew scheduling. A scenario selection heuristic is introduced which improves reserve crew schedule quality using fewer input scenarios. The secondary objective of this thesis is to investigate the effect of the reserve policy used on the day of operation, that is, determining when and which reserve crew should be utilised. The questions of how reserve policies can be improved and how they should be taken into account when scheduling reserve crew are addressed. It was found that the approaches developed for reserve crew scheduling lend themselves well to an online application, that is, using them to evaluate alternative reserve decisions to ensure reserve crew are used as effectively as possible. In general it is shown that `day of operation' disruptions can be significantly reduced through both improved reserve crew schedules and/or reserve policies. This thesis also points the way towards future research based on the proposed approaches