Learning-based Robust Resource Allocation for Ultra-Reliable V2X Communications

Weihua Wu, Runzi Liu, Qinghai Yang, Hangguan Shan, Tony Q.S. Quek

Research output: Contribution to journalArticlepeer-review


Vehicle-to-everything (V2X) communications face a great challenge in delivering not only the low-latency and ultra-reliable safety-related services but also the minimum throughput required entertainment services, due to the channel uncertainties caused by high mobility. This paper focuses on the robust resource management of V2X communications with the consideration of channel uncertainties. First, we formulate a transmit power minimization problem, whilst guaranteeing the different quality-of-service (QoS) requirements. To achieve the robustness of QoS provisions against channel uncertainties, a statistical leaning approach is developed to learn the uncertainties from the data samples of the random channel coefficients as a convex ellipsoid set, which is also called high-probability-region (HPR). Then, the highly intractable power minimization problem is converted into a second-order cone program by the robust optimization approach. Afterwards, we propose a joint set partitioning and reconstruction mechanism to further reduce the total transmit power by pruning the rough HPR into a more precise uncertainty set, which leads to a trackable second-order cone program and a linear program. Finally, we prove that the network performance can be effectively enhanced by the improvement mechanism. Simulation results verify the effectiveness of the robust resource allocation approaches over the non-robust one.

Original languageEnglish
JournalIEEE Transactions on Wireless Communications
Publication statusAccepted/In press - 2021

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Electrical and Electronic Engineering
  • Applied Mathematics


Dive into the research topics of 'Learning-based Robust Resource Allocation for Ultra-Reliable V2X Communications'. Together they form a unique fingerprint.

Cite this