Due to high congestion in cities and growing demand for last-mile delivery services, several companies have been implementing two-echelon distribution strategies over the past few years. Notably, the installation of urban transshipment points has gained increasing attention, used by logistics operators to transfer goods from large freight trucks to smaller and more agile vehicles for last-mile delivery. Nevertheless, the main challenge is how to decide the number and location of these facilities under the presence of demand uncertainty. In this paper, we develop a two-stage stochastic program to design two-echelon last-mile delivery networks under demand uncertainty. This approach decomposes the problem into strategic decisions (facility location) and operational decisions (daily distribution of goods). To address large-scale instances, we solve the model through the sample average approximation (SAA) technique and estimate the optimal routing costs (of the SAA counterpart) using a continuous approximation method. Using a real-world case study with more than 1300 customers from New York City, our results provide several managerial insights regarding the mix of transportation modes, facility location, and the impact of allowing the outsourcing of customer demand. We provide extensive validation of the two-stage stochastic program results through a simulation-based approach and the calculation of the value of the stochastic solutions.

由于城市的高度拥堵和对最后一英里交付服务的需求不断增长，一些公司在过去几年中一直在实施两级配送战略。值得注意的是，城市转运点的安装越来越受到关注，物流运营商使用这些转运点将货物从大型货运卡车转移到更小、更灵活的车辆上，以进行最后一英里的交付。然而，主要挑战是如何在需求不确定的情况下确定这些设施的数量和位置。在本文中，我们开发了一个两阶段随机程序来设计需求不确定下的两级最后一英里交付网络。这种方法将问题分解为战略决策（设施位置）和运营决策（货物的日常分配）。为了解决大规模实例，我们通过样本平均近似 (SAA) 技术求解模型，并使用连续近似方法估计（SAA 对应物的）最优路由成本。通过对来自纽约市的 1300 多名客户的真实案例研究，我们的结果提供了一些关于交通方式组合、设施位置以及允许客户需求外包的影响的管理见解。我们通过基于模拟的方法和随机解决方案的值计算对两阶段随机程序结果进行了广泛验证。我们的结果提供了一些关于运输方式组合、设施位置以及允许客户需求外包的影响的管理见解。我们通过基于模拟的方法和随机解决方案的值计算对两阶段随机程序结果进行了广泛验证。我们的结果提供了一些关于运输方式组合、设施位置以及允许客户需求外包的影响的管理见解。我们通过基于模拟的方法和随机解决方案的值计算对两阶段随机程序结果进行了广泛验证。