Machine Learning-Based Medium Access Control Protocol for Heterogeneous Wireless Networks: A Review

Abstract

This research presents a comprehensive investigation of a Machine Learning (ML)-based Medium Access Control (MAC) protocol strategy for improving heterogeneous wireless network performance parameters and optimizing various (MAC) protocol issues like synchronization, bandwidth competency, error-prone broadcast channel, quality of service support, mobility of nodes, hidden and exposed terminal etc. All nodes in a wireless network use the same broadcast radio channel. The amount of bandwidth available for communication in such networks is restricted due to the radio spectrum's limitations. A unique set of protocols is necessary for managing access to the shared medium in order to improve reliability and quality of provision in such networks. Access to this shared media should be managed to ensure that all nodes receive a fair portion of the available bandwidth and that it is utilized effectively. The Medium Access Control (MAC) Protocol decides the accessible spectrum sharing among the users. For these purpose, various network management approaches have been developed to automate networking choices, notably on the MAC stage. To address these issues, the decentralized decision-making characteristic of Deep Reinforcement Learning (DRL) can be employed in current wireless communication and networking system to solve the key challenges that arise in such networks, such as the coexistence of several types of wireless connections serving different users. Reinforcement Learning (RL) and Deep Learning (DL) are combined in DRL where reinforcement learning has the ability to take right decisions and deep learning has the ability to perform same actions as the human brain do with the help of deep neural network (DNN), and these two techniques are the subsection of the machine learning (ML) and artificial intelligence (AI) technology.