Data-Driven Optimization of Truck–Drone Collaborative Delivery with Shared Fleet Allocation

Truck–drone collaborative delivery (TDCD) refers to a coordinated logistics paradigm in which drones are deployed from delivery trucks to serve nearby customers, enabling parallelized last-mile operations. Much of the existing TDCD literature relies on synthetic datasets and manufacturer-declared drone specifications, which may overestimate performance in real-world operations. This study develops an empirically informed, route-based Mixed-Integer Linear Programming (MILP) framework that integrates empirically derived drone performance models with constrained fleet allocation decisions. Using delivery routes from the Amazon Last Mile Routing Dataset (2021), we consider three electric trucks departing from a common depot, each equipped with drones drawn from a shared fleet of 10 units. Drone flight time and energy consumption are modeled using regression functions calibrated with real flight test data from a DJI Matrice 100 platform, capturing observed variations due to payload and operational conditions. The optimization jointly determines truck stop selection, customer assignments, and drone allocation while minimizing a weighted combination of route makespan, total energy consumption, and fleet size under operational and energy constraints. The results indicate that coordinated truck–drone delivery can achieve substantial reductions in both delivery completion time and energy consumption relative to conventional truck-only delivery. These findings demonstrate the effectiveness of coordinated truck–drone operations under realistic constraints and highlight the importance of data-driven modeling and fleet-level resource allocation in improving last-mile delivery performance.