Bi-Level Optimization Framework for Urban Low-Altitude UAV Delivery Ensuring Target Level of Safety

In recent years, urban low-altitude UAV logistics has emerged as a pivotal solution for last-mile delivery. To enhance operational efficiency, it is essential to address the Coupled Task Allocation and Trajectory Optimization Problem (CTA-TOP) for heterogeneous UAVs. However, existing research, which often approaches this problem using multi-objective weight assignment methods, suffers from two primary shortcomings: (1) It overlooks the inherently multi-objective nature of the problem, where improper weight allocation leads to solutions that deviate from practical needs (e.g., low risk weights may ignore excessive local risks, while high weights unduly sacrifice efficiency). (2) Although ground risk is recognized as a fundamental constraint for Urban Air Mobility (UAM), weight-based optimization cannot guarantee that segment path risks consistently remain below the Target Level of Safety (TLS). To this end, this paper proposes a bi-level optimization framework based on improved algorithms: the TC-NSGA-III algorithm is employed for task allocation to minimize total ground risk, delivery time cost, and UAV workload balance, while the RG-FMT ${}^{\ast }$ algorithm is used for trajectory planning to ensure path risk remains below the TLS. Simulation results demonstrate that the proposed RG-FMT ${}^{\ast }$ algorithm achieves a 100% TLS compliance rate in trajectory planning with the shortest computation time. The overall framework significantly outperforms comparative methods across key metrics such as total risk and time cost. Further large-scale urban scenario simulations validate the algorithm’s generalization capability and scalability. This study provides an effective solution for safe and efficient coordinated scheduling in urban low-altitude logistics.