06/ 2025: We are pleased to share that the paper “Path Pool-based Transformer Model in Reinforcement Framework for Dynamic Urban Drone Delivery Problem” by Chuankai Xiang has been accepted for publication in Transportation Research Part C: Emerging Technologies (TRC), a leading journal in the field of transportation systems and technologies.
This study addresses the critical challenge of real-time scheduling for large fleets of delivery drones in dynamic urban environments. To overcome the computational complexity associated with large-scale dynamic optimization, the paper proposes a novel Path Pool-based Transformer model integrated with Reinforcement Learning (PPTRL). Unlike traditional models that only use the last-visited node’s embedding, the proposed method introduces a Dependency Decay Pooling Strategy (DDPS) to incorporate the full path context, effectively capturing long-term dependencies in routing decisions.
Experimental results demonstrate that the PPTRL model achieves near-optimal solutions with significantly reduced computation times on small-scale problems, and outperforms state-of-the-art heuristics and learning-based approaches on larger-scale instances. The model also shows excellent scalability and robustness, highlighting its strong potential for real-world applications.
Congratulations to Chuankai on this important achievement!
Lishuai Li 