Bi-Level Optimization Framework for Two-Echelon UAV Scheduling in Urban Low-Altitude Logistics

Abstract

Efficient urban UAV logistics require deep synergy between task allocation and 3D path planning, which traditional methods often decouple. Traditional ground-based wheeled logistics is constrained by road traffic efficiency, leading to frequent delivery delays and continuously rising costs. Additionally, existing studies usually decouple task allocation and path planning, and ignore the dynamic energy consumption differences of UAVs under actual full-load and empty-load states. To address this, this study proposes a novel heterogeneous bi-level optimization framework, Bi-AGWO, to solve the two-echelon multi-UAV cooperative delivery problem (2E-MDVRP) in 3D urban environments. The upper-layer model utilizes ant colony optimization (ACO) to handle discrete combinatorial task dispatching and cross-hub sharing sequences. Meanwhile, the lower-layer model integrates the grey wolf optimizer (GWO) with cubic smoothing splines for continuous 3D obstacle avoidance and precise flight distance evaluation. A novel iterative closed-loop feedback mechanism directly returns high-fidelity lower-layer physical distances to drive the upper-layer heuristic search. Simulation results factoring in UAV empty weight and power consumption demonstrate superior global convergence stability. In dense obstacle scenarios, Bi-AGWO achieves an optimal global energy consumption of 23.42 kWh, significantly outperforming state-of-the-art baselines like Bi-ALNS and Bi-PSO. It autonomously evolves an efficient near-neighbor clustering topology. Furthermore, parameter sensitivity analysis reveals a S-shaped energy decay response to maximum payload capacity and a fierce monotonic escalation with empty weight increases. Ultimately, the framework serves as a robust decision-making tool for urban instant logistics enterprises.