Given an airline schedule and demand forecasts, the Fleet Assignment Problem consists in determining how to assign aircraft types to flight legs in the best possible way. This assignment has a major impact on the profit of an airline, since it determines the quantities of seats available over the itineraries of the flight network, along with the associated operating cost. Decades of research on this problem have improved the formulations to be more and more realistic. This thesis extends the ongoing work, considering the problem of doing Fleet Assignment taking demand volatility into account. We first propose a study involving the two models of the literature that are the most widely used by the industry, FAM and IFAM. We show that FAM can be seen as a Lagrangian Relaxation of IFAM, with particular Lagrangian multipliers. We implement this relaxation, and we apply known results to extend it in a column generation based on a Dantzig-Wolfe decomposition of IFAM. We then study the effects of forecasts inaccuracy over the performance of IFAM, and we present a novel approach for modeling the Fleet Assignment Problem. Our model, Market Driven Fleet Assignment Model (MDFAM), makes the itinerary demands part of the decision variables. We propose to constraint these variables rather than consider them as a fixed input of the problem, and we call the resulting constraints Market Constraints. We illustrate the flexibility of this approach through various examples, and we provide a series of experiments in order to determine which Market Constraints give the best results. We compare the different models, and we show that MDFAM can reach a performance which is similar to IFAM's, while being easier to use and to implement.
