EUVIMEDDescoperă EUVIMED
Crossref

BMG-Q: Localized Bipartite Match Graph Attention Q-Learning for Ride-Pooling Order Dispatch

Yulong Hu, Siyuan Feng, Sen Li

IEEE Transactions on Intelligent Transportation Systems · 2025-10 · 26 (10) · 15407-15421

DOI: 10.1109/tits.2025.3595653 ↗

At a glance

Journal article in IEEE Transactions on Intelligent Transportation Systems, 2025-10. Written by Yulong Hu and 2 co-authors. Cited 8 times so far (Crossref); 53 references. Funded by: Hong Kong Research Grants Council, National Natural Science Foundation of China, General Research Fund from the Research Grants Council of Hong Kong Special Administrative Region.

RezumatDate bibliograficeCiteRelated studies

Abstract (original language)

This paper introduces Localized Bipartite Match Graph Attention Q-Learning (BMG-Q), a novel Multi-Agent Reinforcement Learning (MARL) algorithm framework tailored for ride-pooling order dispatch. BMG-Q advances ride-pooling decision-making process with the localized bipartite match graph underlying the Markov Decision Process, enabling the development of novel Graph Attention Double Deep Q Network (GATDDQN) as the MARL backbone to capture the dynamic interactions among ride-pooling vehicles in fleet. Our approach enriches the state information for each agent with GATDDQN by leveraging a localized bipartite interdependence graph and enables a centralized global coordinator to optimize order matching and agent behavior using Integer Linear Programming (ILP). Enhanced by gradient clipping and localized graph sampling, our GATDDQN improves scalability and robustness. Furthermore, the inclusion of a posterior score function in the ILP captures the online exploration-exploitation trade-off and reduces the potential overestimation bias of agents, thereby elevating the quality of the derived solutions. Through extensive experiments and validation, BMG-Q has demonstrated superior performance in both training and operations for thousands of vehicle agents, outperforming benchmark reinforcement learning frameworks by around 10% in accumulative rewards and showing a significant reduction in overestimation bias by over 50%. Additionally, it maintains robustness amidst task variations and fleet size changes, establishing BMG-Q as an effective, scalable, and robust framework for advancing ride-pooling order dispatch operations.

Funding

  • Hong Kong Research Grants Council
  • National Natural Science Foundation of China
  • General Research Fund from the Research Grants Council of Hong Kong Special Administrative Region

Cite

APA

Hu, Y., Feng, S., & Li, S. (2025). BMG-Q: Localized Bipartite Match Graph Attention Q-Learning for Ride-Pooling Order Dispatch. IEEE Transactions on Intelligent Transportation Systems, 26(10), 15407-15421. https://doi.org/10.1109/tits.2025.3595653

Vancouver

Hu Y, Feng S, Li S. BMG-Q: Localized Bipartite Match Graph Attention Q-Learning for Ride-Pooling Order Dispatch. IEEE Transactions on Intelligent Transportation Systems. 2025;26(10):15407-15421. doi:10.1109/tits.2025.3595653

BibTeX
@article{Hu2025,
  author = {Hu, Yulong and Feng, Siyuan and Li, Sen},
  title = {BMG-Q: Localized Bipartite Match Graph Attention Q-Learning for Ride-Pooling Order Dispatch},
  journal = {IEEE Transactions on Intelligent Transportation Systems},
  year = {2025},
  volume = {26},
  number = {10},
  pages = {15407--15421},
  doi = {10.1109/tits.2025.3595653}
}
Start for free →