Reinforcement Learning-based Computation Resource Allocation Scheme for 5G Fog-Radio Access Network

Fog computing has emerged as one of the key building blocks of fifth generation mobile networks (5G) because of its ability to effectively meet the demands of real-time or latency-sensitive applications. To introduce fog in 5G, particularly in the radio access network (RAN), intermediate network devices such as remote radio heads, small cells and macro cells are equipped with virtualised storage and processing resources to constitute the fog RAN (F-RAN). However, these resources are limited and inefficient management could cause a bottleneck for F-RAN nodes. To this end, this paper focuses on developing a dynamic and autonomous computing resource allocation scheme for F-RAN considering delay requirements of users at a node. The proposed algorithm uses reinforcement learning to optimise latency, energy consumption and cost in the F-RAN. The performance and computational complexity of the proposed algorithm will be evaluated as part of a simulation and the results compared with other algorithms from existing studies with a similar objective function.

I. Introduction

The ever-increasing need to achieve low latency and ultrahigh reliability has led to the emergence of the fog computing paradigm, whose fundamental principle is to bring cloud computing capabilities to the edge of the network closer to end devices and users [1].