In this work, a mixed integer nonlinear programming model combining direct service links, demand uncertainty and congestion effects is proposed. This model is efficiently solved by Generalized Benders Decomposition, for instances of moderate sizes and reasonable number of scenarios. The deployed algorithms are further used for re-designing the Brazilian air transportation network, enabling the analysis of future demand scenarios and providing decision support about the optimal investment policy for Brazil.
hub-and-spoke networks; Benders Decomposition; large scale optimization; mixed integernon-linear programming; demand uncertainty