Data-driven robust two-stage ferry vehicle management at airports

In the face of substantial uncertainties in flight schedules, driven by factors such as heavy traffic flow, extreme weather conditions, and climate change, efficient management of ground support vehicles at airports becomes a critical challenge. This paper delves into the ferry management problem (FMP), where a fleet of ferries, comprising both regular and backup vehicles, is tasked with servicing flights within specified time windows before their arrival or departure. The central aim of the FMP is to optimize ferry vehicle allocation, minimizing total operational cost while ensuring punctual and effective service for each flight. A novel two-stage scenario-based robust model is introduced to effectively capture the potential uncertainties. We present four solution strategies to solve the FMP. The initial two methods, the sample average approximation (SAA) and its robust version (RSAA), focus on reducing computational demands through a selective sampling of scenarios. Our third approach, built on the column-and-constraint generation (C&CG) procedure, guarantees the solution quality by progressively incorporating critical scenarios into the master problem, benefiting from the strategic limitation of scenarios and the transformation of subproblems into minimum-cost maximum-flow problems for efficient solution approximation. Lastly, we introduce a data-driven, on-the-fly heuristic that dynamically adjusts scheduling plans, boosting adaptability to real-time operational fluctuations. Our comprehensive experiments, utilizing real-world datasets, validate the robustness, efficiency, and effectiveness of the proposed algorithms, showcasing their practical applicability in managing airport ground support under uncertain conditions.