Designing flexible service strategies for urban drone Delivery: A hybrid Simulation-Optimization framework

Urban drone delivery offers the potential to improve last-mile logistics efficiency, yet it introduces challenges in resource allocation and dynamic pricing under uncertain operational conditions. This study proposes a flexible service strategy design (FSSD) framework that integrates system dynamics (SD) modeling with reinforcement learning (RL) to enable adaptive deployment and pricing decisions. The SD model captures nonlinear feedback mechanisms and time-varying interactions within the delivery system, while a Deep Q-Network (DQN) is employed to learn pricing and resource allocation strategies that adapt to real-time fluctuations in demand and system constraints. A case study based on Meituan’s drone delivery pilot in Shenzhen, China, demonstrates a 14.88% increase in average daily cumulative profit along with a significant reduction in backlog volatility. The learned strategies are responsive to environmental variations and generalize effectively across different system states. Sensitivity analyses and extended experiments on economies of scale and dynamic pricing further identify key drivers of performance. The FSSD framework offers actionable managerial insights for designing adaptive, scalable, and efficient drone logistics systems.