Cluster-enabled cooperative scheduling based on reinforcement learning for high-mobility vehicular networks

Youhua Xia, Libing Wu*, Zhibo Wang, Xi Zheng, Jiong Jin

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

19 Citations (Scopus)


It is important to transmit data reliably and efficiently in vehicular networks. Existing works usually study routing strategies and cooperative scheduling to improve the efficiency of transmission. However, the data transmission remains inefficient because of the lack of full use of communication resources. The transmission is unreliable because information cannot be completely transmitted to the destination vehicles. Moreover, the increasing number of connected vehicles and the limitation of available communication resources make task scheduling challenging in vehicular networks. In this work, we propose Cluster-enabled Cooperative Scheduling based on Reinforcement Learning(CCSRL) to improve the communication efficiency and reliability of vehicular networks, with the goal of maximizing the information capacity. In particular, we leverage the stability to select a cluster head vehicle to enhance data transmission efficiency, and a reinforcement learning-based auxiliary transmission is further designed to guarantee the reliable communication among vehicles. The experimental results demonstrate that the performance of the proposed scheduling algorithm, especially the performance of the packet delivery ratio and node packet loss ratio, is better than that of the state-of-the-art algorithm.
Original languageEnglish
Pages (from-to)12664-12678
Number of pages15
JournalIEEE Transactions on Vehicular Technology
Issue number11
Early online date8 Oct 2020
Publication statusPublished - Nov 2020


  • Clustering
  • information capacity
  • reinforcement learning
  • vehicular networks


Dive into the research topics of 'Cluster-enabled cooperative scheduling based on reinforcement learning for high-mobility vehicular networks'. Together they form a unique fingerprint.

Cite this