How Can Drones Revolutionize Package Transport Between Warehouses?

In recent years drones have become logistics stars, driven by the surge in fast, sustainable deliveries. Yet their role goes beyond the last mile: a new study focuses on flying drones between sorting hubs to move parcels more efficiently. The concept is simple: if a warehouse handles more packages, drones should visit it more often. To achieve this, researchers imagined two drone types: "local" ones that circle all hubs, and "express" ones that fly only to the busiest centers. They turned the problem into a mixed integer linear programming (MILP) model that aims to minimize total transport time. Unlike old models with fixed routes, this allows flexible paths, including direct flights or loop circuits. To cope with large networks, they added an approximation method: first they pre‑select high‑demand hubs based on package volume and location, cutting down the number of decision variables. Numerical tests show the approach works well even as the number of bases grows. In practice, drones could become invisible "drivers" shuttling parcels between warehouses, making the supply chain faster and greener.

The proposed model relies on a Mixed‑Integer Linear Programming (MILP) formulation that combines continuous variables—representing flight times and distances—with binary variables indicating whether a drone visits a specific hub within a given interval. The objective is to minimize the total travel time, subject to constraints on drone capacity, hub availability, and adherence to loading/unloading time windows. For local drones, the model enforces that each hub is served at least once within the planning cycle, while for express drones a pre‑selected candidate set is introduced, drastically reducing the number of binary variables. Route flexibility is managed through flow variables that allow non‑linear paths, including direct jumps between non‑adjacent hubs, and closed loops that optimize the return to the starting point. To ensure scalability, a Benders’ decomposition algorithm was implemented, separating the master problem (selection of high‑demand hubs) from the sub‑problem (flight scheduling). Additionally, a heuristic pre‑filtering approach was introduced: spatial centrality metrics (e.g., proximity index) and package volume are computed to create a shortlist of candidate hubs. This reduces the problem to a manageable size even for networks with hundreds of hubs, while maintaining solution quality within 5 % of the global optimum. Experimental results on synthetic and real datasets show a total transport time reduction of up to 30 % compared to static strategies, with computation times under 10 minutes for networks of 200 hubs. The approach paves the way for dynamic aerial logistics systems capable of adapting in real time to demand fluctuations and operational conditions such as wind or airspace restrictions.