Cooperative evolution of support vector machine empowered knowledge-based radio resource management for 5G C-RAN

Abstract

Cloud-radio access network (C-RAN) is an innovative approach in 5th Generation communication networks, where baseband processing units (BBUs) are dissociated from the remote radio heads (RRHs), and a remote cloud-based centralized pool of BBUs is formed. This paper provides an efficient multi-class classification radio resources management (RRM) scheme based on the cooperative evolution of support vector machine (SVM) to assign the spectrum resources of macrocellular users (MUE) i.e. sub-channel of variable bandwidth, to remote-head users (RUEs) and D2D pairs, such that sub-channels can be reused without compromising the quality of service (QoS). First, the C-RAN resource allocation problem is modeled as a NILP problem, with CUs and D2D nodes sharing the same channel. Following that, the proposed strategy is used to allocate resources. A realistic dataset is built for assessing machine learning-based techniques using 5G experimental prototype based on Open Air Interface (OAI). Finally, experimental results demonstrate the effectiveness of the proposed cooperative evolution of SVM-based, multi-class classification RRM over alternative schemes in terms of network throughput, prediction performance, and system utilization.