In April 2016, Amazon won a patent for a flying warehouse that deploys unmanned aerial vehicles (UAVs) to deliver parcels to customers. In this system, the airborne fulfillment center (AFC) remains or moves at a high altitude, and UAVs are deployed from the AFC to deliver orders to customers. Smaller airships, also known as shuttles, may be used to replenish the AFC with inventory, UAVs, fuel, supplies, and staff. Amazon’s AFC delivery system is a novel and challenging system that can convert same-day deliveries into 10 min deliveries and provide a wide range of UAV delivery services for Amazon Prime members. However, to operate the AFC delivery system efficiently, a quantitative operation methodology needs to be developed, which considers the related regulations and consumable replenishment. In this study, mixed-integer linear programming is developed to derive optimal operation schedules for AFC, shuttles, and UAVs simultaneously. The proposed model can be used in real time to address system changes and derive new operation schedules. The effects of system components are analyzed via a numerical experiment to provide managerial guidelines for the implementation of the AFC delivery system. Furthermore, cooperation between the new AFC delivery system and the existing stationary UAV delivery service was investigated. The results demonstrate the possibility of complementary cooperation between the two systems.

In response to increasing congestion and inefficiencies in urban freight transportation systems, driven by the rapid rise of e-commerce and urbanization, underground logistics are emerging as a promising alternative. This study introduces the Vehicle Routing Problem with Underground Transportation (VRP-UT), a novel model that integrates traditional vehicle routing with underground logistics facilities to optimize urban deliveries. To address the model’s complexity, we propose a hybrid solution approach that combines Q-learning with pruning techniques, enhancing route optimization and minimizing unnecessary operations. Comprehensive numerical and sensitivity analyses further validate the model’s effectiveness, positioning VRP-UT as a scalable and efficient solution to meet growing urban logistics demands while supporting sustainable development. A case study, including sensitivity analysis based on Seoul Subway Line 3, demonstrates the practical applicability of the VRP-UT model, highlighting its potential to alleviate surface-level traffic congestion, reduce operational costs, and shorten delivery times in densely populated urban areas.