Truck-Drone Collaborative Path Planning in Large-Scale Urban Scenarios

Biao Xu

Keywords: truck-drone collaborative delivery; multi-objective optimization; deep reinforcement learning; path planning

Abstract: With the vigorous development of e-commerce, the timeliness requirements for logistics distribution have been continuously elevated. Customers' specific demands for delivery time windows have rendered it crucial to complete delivery services within the specified timeframes. However, in large-scale urban scenarios characterized by dense customer distribution and complex delivery environments, traditional distribution models are struggling to meet the demands for efficient and on-time logistics. Additionally, delivery time windows, as hard constraints, narrow the scope of feasible solutions and increase the difficulty of path planning. To address these challenges, this study explores the Truck-Drone Collaborative Delivery Problem with Multiple Time Windows (TDCDP-MTW) and its solution algorithms. A hybrid distribution model is proposed, which categorizes customers into two types: dense and sparse. For dense customer clusters, fixed trucks are employed as UAV (Unmanned Aerial Vehicle) deployment centers; for sparse customer areas, trucks carry UAVs for delivery. This approach enhances UAV utilization and delivery efficiency. Meanwhile, by integrating deep reinforcement learning and attention mechanism, an encoder-decoder architecture is constructed to capture spatiotemporal dependencies. This architecture is applied to optimize path planning schemes in large-scale customer scenarios, improving computational efficiency and generalization ability. Experimental results demonstrate that this study provides a collaborative path planning solution that balances efficiency and flexibility for high-density urban logistics, and resolves the coupling dilemma between multi-objective optimization and dynamic decision-making in large-scale scenarios.

Submission Number: 42