Robust optimisation in network revenue management

Network revenue management is used extensively, particularly within the airline industry, to allocate dependent resources between different products. This work focuses on the situation where demand is uncertain and the aim is to determine booking limits that are robust to fluctuations in demand. Expanding on the work of Perakis and Roels (2010), we developed a genetic algorithm that finds booking limits that either minimize the maximum regret or maximize the minimum revenue for a number of different booking control policies: partitioned booking limits, nested booking limits and bid prices. We present results that demonstrate how these booking limits outperform those obtained via local descent methods and other traditional network models. Furthermore, we consider the uncertainty set for demand to be ellipsoidal further to the polyhedral as originally proposed. Finally, we introduce the formulation on network cruise revenue management application. We present the robust formulation for the cruise network setting and present numerical results that show that the robust control measures outperform standard approximation methods