Two-stage stochastic planning of rural truck–UAV delivery networks under weather and demand uncertainty
Abstract
Rural express delivery is challenged by dispersed settlements, long road detours, limited demand density, and weather-sensitive operations. This study develops a two-stage stochastic planning framework for a county-level truck–unmanned aerial vehicle (UAV) delivery network under joint weather and demand uncertainty. Residual unmet demand is incorporated into the objective function through a penalty term, while service rate is reported as an operational performance indicator. The first stage determines township forward station locations, heterogeneous UAV fleets, and backup truck capacity, while the second stage determines scenario-dependent village assignment, UAV delivery, truck recourse, and residual unmet demand. A scenario-aware adaptive large neighborhood search (SA-ALNS) is developed by combining adaptive operator weighting with service deficits, resource utilization, and weather-related UAV infeasibility in the destroy-and-repair process. The results show that station performance depends on spatial complementarity and its interaction with heterogeneous fleet capacity. Demand surges mainly increase fleet-time pressure, whereas adverse weather reduces UAV feasibility and transfers additional workload to backup trucks. The baseline analysis further indicates that truck–UAV delivery does not automatically outperform truck-only delivery when station investment, vehicle costs, supporting resources, and service outcomes are considered. These findings support a conditional and performance-oriented interpretation of rural truck–UAV deployment and provide a planning framework for phased air–ground logistics development.