GOALI: Demand Driven Fleet Management Analysis, Models, and Algorithms for the Airline Industry

Virginia Polytechnic Institute And State University, Blacksburg VA

Investigators

Abstract

This research project proposes to focus on an essential aspect of airline supply management, namely, the fleet assignment decision, and its effective coordination with demand under conditions of uncertainty. The intent is to develop and implement general models and algorithms in concert with our industrial partner: United Airlines, in order to ensure the transfer of technology to the airline industry. The current airline practice is to assign aircraft capacity to scheduled flights well in advance of departures. This is due to the dependency between the fleet assignment decision and other airline processes. At such an early stage, however, the high uncertainty of demand poses a major impediment for airlines to best match aircraft capacities with the final demand. On the other hand, the accuracy of demand forecasts improves markedly over time. Thus, it becomes advantageous to postpone irrevocable fleeting decisions to as great a future extent as possible, and to inject sufficient flexibility in the prescribed decisions so as to facilitate the ability to make future revisions within the operational constraints. To accomplish this, we propose a three-stage dynamic supply management framework that coordinates the initial fleet assignment model with a pair of downstream demand driven re-fleeting and swapping models in a manner designed to exploit the flexibility in the system, while incorporating progressively more accurate demand forecasts obtained as departures approach. The intent is to provide sufficient flexibility in the initial fleet assignment decisions to permit effective subsequent re-fleeting decisions so as to promote a better matching of demand with supply. If successful, the impact of our proposed research contributions will be: (1) to promote improved supply management decisions in the airline industry, while considering the interactions between the initial fleeting, later re-fleeting, and swapping models within this framework; and (2) to prescribe effective solution approaches for the various models at each stage of this process, which can be extended to other similar manufacturing and service industrial environments. In conducting this study, we propose to investigate several model enhancements to improve representability in practice by incorporating various features such as path-level and fare-class based demands, recapture of spilled passengers, day-of-week demand variations, and maintenance considerations. We hope thereby to help eliminate several critical weaknesses associated with the traditional fleet assignment process, lend insights into improving operations, and achieve a demonstrable advance in the state-of-the-art in practice.