Data analytics for robust crew pairing

数据分析可实现稳健的船员配对

• 批准号: 580589-2022
• 负责人: Gzara, FatmaF
• 金额: $ 2.38万
• 依托单位: University of Waterloo
• 依托单位国家: 加拿大
• 项目类别: Alliance Grants
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=758131
• 关键词: Data; analytics; robust; crew; pairing

Column generation and branch-and-price are well accepted approaches for several airline operations planning problems including crew pairing optimization. While most of the research developed over the last 60 years assumes known and constant flight characteristics, this is hardly true in practice. Airlines face daily changes in these characteristics like flight delays and crew absenteeism. In particular, crew schedules are highly impacted because of the strict regulations on crew safety, labor agreements, etc. Recent research addresses uncertainty in flight durations by modelling schedule recovery strategies and using robust or stochastic optimization, where assumptions are made on the uncertain parameters. With the abundance of historical flight data, there is a huge opportunity to leverage advances in machine learning tools to develop prediction models of flight characteristics, to augment existing crew pairing optimization solution methods using predicted information, and to develop new models and solutions that integrate learning and optimization.

列生成和分支与价格是解决包括机组配对优化在内的多种航空公司运营规划问题的广为接受的方法。虽然过去 60 年开展的大多数研究都假设飞行特性已知且恒定，但在实践中却很难成立。航空公司每天都面临着这些特征的变化，例如航班延误和机组人员缺勤。特别是，由于对机组人员安全、劳动协议等方面的严格规定，机组人员的时间表受到很大影响。最近的研究通过对时间表恢复策略进行建模并使用鲁棒或随机优化（其中对不确定参数进行假设）来解决飞行时间的不确定性。凭借丰富的历史飞行数据，我们有巨大的机会利用机器学习工具的进步来开发飞行特性的预测模型，使用预测信息增强现有的机组配对优化解决方案，以及开发集成学习和优化的新模型和解决方案。