Abstract
Device-to-device (D2D) enabled cloud radio access networks (C-RANs) are potential solutions for further improving spectral efficiency (SE) and decreasing latency by allowing direct communication between two users. However, due to the acquisition of global channel state information (CSI) and the execution of centralized algorithms, heavy burdens are placed on the fronthaul and the baseband unit (BBU) pool. To alleviate these burdens, a distributed approach to mode selection and resource allocation for potential D2D pairs under pre-determined resource allocation of C-RAN users is proposed, in which pairs of users are endowed with decision-making capabilities. The proposed procedure is divided into three stages: communication mode and subchannel selection, utility value determination, and reinforcement learning based strategy update. The core idea is that D2D pairs selfoptimize the mode selection and resource allocation without global CSI under several practical constraints. Simulation results show that enabling D2D can significantly improve SE for CRANs. Furthermore, the impacts of the fronthaul capacity, the centralized signal processing capability of the BBU pool, and the distance between the D2D transmitter and the remote radio head are demonstrated and analyzed.
Funding Information
  • Chinese State Major Science and Technology Special Project (2016ZX03001020-006, 2017ZX03001025-006)
  • Chinese National Program for Special Support of Eminent Professionals
  • National Science Foundation (CNS-1702808, ECCS-1647198)

This publication has 42 references indexed in Scilit: