Quality of service‐aware adaptive radio resource management based on deep federated Q‐learning for multi‐access edge computing in beyond 5G cloud‐radio access network

Abstract

Due to the exponential increase of mobile data services, both industry and academia have turned their focus for improving the quality of service (QoS) provisioning for the beyond 5G (B5G) networks. In this paper, an adaptive QoS-aware resource allocation scheme is developed for multi-tenant and multi-service cloud-radio access networks (C-RAN) mobile edge computing scenario with the aim of maximizing network performance over time. First, a multi-user C-RAN resource allocation architecture for task computation and result collection is developed, where edge nodes can take on user equipments' (UEs) computation-intensive activities. Then deep reinforced learning (DRL) based multi-user resource allocation scheme, that is, deep federated Q-learning (DFQL) is developed for determining the best resource allocation policy. To maximize the validation accuracy of the global model in the FL while satisfying resource restrictions, this technique learns resource allocation immediately in contrast to the present allocation mechanisms. A realistic dataset is built for assessing the proposed method using 5G experimental prototype based on open air interface (OAI). At last, the effectiveness of the proposed scheme is demonstrated, where the performance of the scheme is compared with the existing resource allocation techniques.