【作者】 李金林； 徐丽萍；

【Author】 LI Jinlin;XU Liping;School of Management and Economics,Beijing Institute of Technology;

【机构】 北京理工大学管理与经济学院；

【摘要】 针对民航客运收益管理中单航段航班多票价级别的舱位控制问题,根据顾客预订和退订的动态性,以及顾客不登机的可能性和超售惩罚成本,用离散时间马尔科夫决策过程描述允许超订下的舱位控制过程,并包含群订情况。顾客到达率和退订概率的预测误差可导致转移概率分布的不确定性,为了系统性地减小最优策略对预测误差的敏感性,建立稳健控制的R-MDP模型。文中引入最大稳健收益的概念,当转移概率分布的不确定性用Kullback-Leibler散度型置信集合表示时,证明了最大稳健收益函数关于剩余座位数的单增性和凹性;提出两种舱位控制的稳健策略,讨论最大可预订数的性质,嵌套的稳健保护水平的存在条件,并给出求解算法。更多还原

【Abstract】 According the dynamics of passenger reservation and cancellation process, the possibility of no-shows and the overbooking penalty, this paper models the problem of the airline capacity control with overbooking a discrete time Markov decision process, for a single-leg flight with multiple fare classes, including group reservations. The forecast errors of the probability of a request arrival and a cancellation will induce the uncertainty of the underlying transition probability distributions. In order to systematically mitigate the sensitivity of the optimal policy to the forecast error, the R-MDP model to achieve robust control is formulated. A concept of maximum robust revenue is introduced, and it is demonstrated increasing and concave in remaining seats under the uncertainty sets of transition probability distributions describing in Kullback-Leibler divergence. Two kinds of robust control polices is proposed. The property of maximum efficient reservation number and the condition of the nested protection levels are discussed. The algorithm to achieve the maximum robust revenues and robust polices is given.更多还原