Dataset of an airline-driven flight rescheduling problem

This data was derived from Luo and Yu (1997), Bard and Mohan (2008), and Brunner (2014). The data is from American Airlines’ daily flight schedule at the Dallas/Fort Worth Airport. There are 71 flights in GDP and the GDP spanned three hours in duration. The planning period of the airline-driven flight rescheduling problem is a half day for one specific season. The data was considered as the base scenario on which a GDP had been issued, and there is no change of the GDP. Including the base scenario, some hypothetical scenarios on subsequent GDPs were assumed and then used to attain the solution. The hypothetical scenarios correspond to new situations in which there are changes in GDP due to weather conditions or external restrictions. The minimum turnaround times of plane and crew were fixed with 30 minutes and 20 minutes, respectively. The maximum allowed time for delays was set to 500 minutes. The misconnection cost of crew members and the cancellation cost of a flight were 50 and 500, respectively. For the incremental delay cost function, each interval, in which different unit costs occurred, was fixed by 60 minutes. There were three intervals. The cost per unit delay time is set to 2, 3, and 4, and associated with each interval, respectively. Correspondingly, the constant value of the incremental delay cost function was set to 0, -60, and -180, and associated with each interval, respectively. The initial flight schedule and the schedule revised by a GDP are presented in "base scenario.csv". The current GDP has been issued at 0. The initial flight schedule defines the arrival and departure times of flights. The connections of crews between flights are also described in the initial flight schedule. In the revised schedule, the arrival times represent the scheduled arrival times of slots. For possible scenarios, ten scenarios were randomly generated. The scheduled arrival times revised by the subsequent GDP in each scenario are presented in scenario##.csv". All the constrained flow rates are 1.81, 1.88, 1.58, 1.64, 1.59, 1.46, 1.49, 1.46, 1.4, and 1.45, respectively. All the next time points are 103, 103, 123, 112, 151, 168, 145, 139, 76, and 103, respectively. In each scenario, the starting times of slots was calculated by adding the inverse of the constrained flow rate to the initial arrival times after the time point, cumulatively. The probability of the base scenario, θ, was set to 0.2, and the probability of the other scenarios was set to 0.08.本数据集源自Luo与Yu（1997）、Bard与Mohan（2008）以及Brunner（2014）的研究成果。数据取自美国航空位于达拉斯/沃思堡机场的每日航班时刻表。该数据集包含地面延误程序（Ground Delay Program，GDP）下的71架次航班，该程序持续时长为3小时。本次航空公司航班调度重排问题的规划周期为某一特定季节的半天时长。该数据被作为已发布地面延误程序的基准场景，且该地面延误程序未发生变更。除基准场景外，研究人员还假设了后续地面延误程序的若干假想场景，并将其用于求解该问题。假想场景对应因天气条件或外部限制导致地面延误程序发生变更的全新工况。 飞机与机组人员的最小周转时间分别固定为30分钟与20分钟。最大允许延误时长设定为500分钟。机组人员的中转延误成本与航班取消成本分别为50与500。针对增量延误成本函数，不同单位成本对应的时长区间均以60分钟为间隔划分，共设3个区间。各区间对应的单位延误成本分别为2、3与4。与之对应，各区间增量延误成本函数的常数项分别设定为0、-60与-180。 初始航班时刻表及经地面延误程序调整后的时刻表收录于"base_scenario.csv"。当前地面延误程序于时刻0发布。初始航班时刻表定义了航班的到达与出发时刻，同时也载明了航班间的机组人员衔接关系。在调整后的时刻表中，到达时刻代表机位的计划到达时刻。针对各类假想场景，研究人员随机生成了10组场景。各场景中经后续地面延误程序调整后的机位计划到达时刻收录于"scenario##.csv"文件中。所有约束流量率分别为1.81、1.88、1.58、1.64、1.59、1.46、1.49、1.46、1.4与1.45。所有后续时间点分别为103、103、123、112、151、168、145、139、76与103。在每组场景中，机位的起始时刻通过以下方式计算：将约束流量率的倒数依次累加到该时间点之后的初始到达时刻中。基准场景的概率θ设定为0.2，其余场景的概率均设定为0.08。