Green Heterogeneous Computing Powers Allocation Using Reinforcement Learning in SDN-IoV

Yingjie Liu, Dan Wang, Bin Song, Xiaojiang Du

Research output: Contribution to journalArticlepeer-review

6 Scopus citations

Abstract

In the era of 6G-driven Internet of Vehicles (IoV), the form of services centered on computing power has promoted the development of applications such as In-Vehicle Infotainment (IVI) and smart driving assistance. The collaboration of computing power deployed between edge and end provides computing services for IoV applications. However, due to the diversity of computing tasks and the heterogeneity of computing power resources, it is a challenge to allocate computing power resources reasonably to meet user needs. To solve this problem, we propose a green heterogeneous computing power allocation method to reduce the energy consumption. First, we design a Heterogeneous Computing Power-IoV (HCP-IoV) architecture based on Software Defined Network (SDN) to realize the deep integration of computing power and network through the hybrid network architecture of vehicle-road collaboration and road-road collaboration. Then, to ensure the reasonable allocation of different computing tasks, we design an intelligent heterogeneous computing power allocation scheme to reduce energy consumption on HCP-IoV. Furthermore, we formulate the joint communication and computing power allocation as a Markov Decision Process (MDP) and use the improved Proximal Policy Optimization (PPO) method to achieve an adaptive allocation of heterogeneous computing powers. Simulation results show the effectiveness of the proposed method.

Original languageEnglish
Pages (from-to)983-995
Number of pages13
JournalIEEE Transactions on Green Communications and Networking
Volume7
Issue number2
DOIs
StatePublished - 1 Jun 2023

Keywords

  • IoV
  • heterogeneous computing power
  • multivariate Gaussian distribution
  • reinforcement learning

Fingerprint

Dive into the research topics of 'Green Heterogeneous Computing Powers Allocation Using Reinforcement Learning in SDN-IoV'. Together they form a unique fingerprint.

Cite this