Enhancing humanitarian logistics under uncertainty: A data-driven distributionally robust optimization approach with worst-case mean-CVaR
针对灾害中设施选址、物资分配和疏散规划问题，提出一种数据驱动的两阶段分布鲁棒优化模型，采用最坏情况均值-CVaR准则应对极端情景，在加拿大麦克默里堡野火案例中验证了模型有效性，发现避难所间协作策略可使关键物资未满足需求降低约40%。
With the rise in global disasters, improving humanitarian supply chains and evacuation planning is essential for saving lives and delivering help quickly and fairly. This study proposes a model that integrates facility location, relief item distribution, and evacuation operations while accounting for critical social parameters such as demographic vulnerability and regional accessibility in affected areas. The inter-shelter collaboration logistics strategy is incorporated into the framework to address challenges in optimizing resource allocation and minimizing disruptions caused by blocked roads and uncertain demands. This research also develops a data-driven two-stage distributionally robust optimization (DRO) model, employing the worst-case mean-conditional value-at-risk criterion to ensure robustness against extreme scenarios. The model’s performance is assessed through out-of-sample analysis, demonstrating the DRO model’s enhanced robustness and effectiveness compared to the traditional two-stage stochastic programming model. The model is applied to the real case of the Fort McMurray wildfire in Alberta, Canada, to validate its practical applicability in disaster management. The results emphasize that prioritizing relief items, addressing social factors, and employing the inter-shelter collaboration strategy together improve evacuation efficiency and enhance resilience in disaster management, with the inter-shelter collaboration strategy contributing, for example, to approximately a 40% reduction in the unmet demand for a critical item.