Types and Challenges of Resource Allocation for Reinforcement Learning in Wireless Networks: A Review

Abstract

– The allocation of resources is one of the primary issues in wireless networks, as it has a direct impact on spectrum efficiency, energy efficiency, latency, and Quality-of-Service (QoS) provisioning. The growing complexity and heterogeneity of the current state of wireless environment, such as dense deployments and time varying channels, have revealed weaknesses in the traditional optimization-based approaches to assigning resources, which usually assume a simplified model and require perfect knowledge of the entire system. This has led to reinforcement learning (RL), which has become a promising alternative because it is able to learn adaptive decision-making policies by interacting with the environment. This is a structured and comprehensive review article about the resource allocation techniques by RL in wireless networks. This paper presents the principles of RL and explains why it must be adopted in this sphere. Subsequently, the review divides the current literature into the main categories of resources, such as spectrum allocation, power allocation, and time-slot scheduling. Moreover, the following critical issues are analyzed: dynamic channel conditions, interference management, scalability, computational complexity, and strict QoS and latency requirements. This review will be a valuable source of information to researchers and practitioners who want to create intelligent resource allocation solutions for the next generation wireless networks by pointing out existing research trends, limitations, and open issues.