Hybrid Fleet Management and Vehicle Repositioning in Mobility-as-a-Service Operations

• 成果类型: Article; Early Access
• 署名作者: Zhang, Guowei; Zhu, Ning; Jia, Ning; Zhao, Long; He, Qiao-Chu
• 署名单位: Dalian University of Technology; Chinese Academy of Sciences; University of Science & Technology of China, CAS; Tianjin University; National University of Singapore; Southern University of Science & Technology
• 刊物名称: PRODUCTION AND OPERATIONS MANAGEMENT
• ISSN/ISSBN: 1059-1478
• DOI: 10.1177/10591478251386538
• 发表日期: 2025
• 关键词: Robust Optimization; systems

摘要

While mobility-as-a-service platforms have revolutionized urban transportation and fundamentally transformed travelers' experience and engagement, they encounter a significant challenge in maintaining a temporal and spatial balance between supply and demand, particularly with the inclusion of crowd-sourced freelance drivers. In this study, we propose a hybrid supply-side management problem that accounts for heterogeneous revenue dependent on surge pricing, where surge pricing not only determines per-trip earnings but also influences demand elasticity and overall revenue generation. This problem combines the physical repositioning of in-house vehicles with an incentive-driven approach for freelance vehicles. A multinomial logit model is employed to model the repositioning behavior of crowd-sourced drivers, and reformulated as a low-level equilibrium problem embedded in a bi-level program. To characterize the surge price-dependent uncertain demand, a tailored residual-based Wasserstein ambiguity set is constructed. Notably, the proposed residual-based Wasserstein DRO model is demonstrated to satisfy both finite sample guarantee and asymptotic optimality. A linear decision rule approximation facilitates a tractable reformulation, and it is shown to incur no loss of optimality for the single-period case. We validate the practical applicability of our model using a dataset from RideAustin. We find that the hybrid fleet approach gets the best of both worlds by increasing average revenue using crowd-sourced vehicles and enhancing system robustness using in-house vehicles. Interestingly, we show that a sweet-spot may exist as an optimal ratio between the number of in-house and crowd-sourced vehicles that maximizes the overall revenue.