Download files
Abstract
Schedule disruptions commonly affect airline operations and cause a great disparity between the expected and actual operational costs. Many disruption management methods have been proposed to address this disparity, establishing the classes of proactive and reactive approaches. While the efficacy of reactive approaches is greatly affected by the level of recoverability resulting from proactive decisions, there is little research regarding the integration of the disruption management classes. Recoverable robustness is one such method that bridges the gap between the proactive and reactive approaches. This thesis aims to demonstrate the potential recoverability improvements from applying recoverable robustness to airline planning problems.
The recoverable robust tail assignment and aircraft maintenance routing problems are introduced to demonstrate the potential of this framework. These problems are formulated as stochastic programs, which are efficiently solved by integrating column generation and Benders' decomposition. The development of enhancement techniques is required to solve the large-scale optimisation problems resulting from large flight schedules and sets of disruption scenarios. In addition, the aircraft maintenance routing problem introduces a novel modelling approach designed to minimise the effect of disruptions that occur on preceding days.
A general framework for column-and-row generation is developed in this thesis to improve the solution runtime and quality compared to a standard column generation approach. This framework is presented as an alternative solution approach to Benders' decomposition.
An explicit evaluation of column-and-row generation against column generation is performed using the integrated airline recovery problem as an example. This evaluation demonstrates an improvement in solution runtimes and assesses the suitability of employing the integrated airline recovery problem in the recoverable robustness evaluation stage. A novel modelling approach for passenger recovery is also proposed, attempting to improve the evaluation stage feedback. This modelling approach reallocates passengers to alternative flights following flight cancellations, effectively reducing operational costs and increasing passenger flow.
This thesis demonstrates the ability of recoverable robustness to improve the recoverability of various airline planning problems. We show the necessity of the many enhancement techniques developed in this thesis to achieve the best results from applying the recoverable robustness framework.