Review of AI-Driven Adaptive QoS Framework for Enhancing Performance in 5G and Beyond Mobile Networks

Author:   Mohamed jassim, Suhad Faisal Behadil

Download PDF

In this paper, we present a comprehensive review of AI and ML techniques for improving adaptive QoS for 5G and beyond networks. Dynamic resource allocation, scalability, and security are bottleneck issues for 5G networks with ultra-reliable low-latency communication (URLLC), massive connectivity, and diversified vertical applications. Traditional static techniques are not adequate to deal with such a complex and heterogeneous environment, and AI-type solutions are required. The paper abstracts and categorizes the application of artificial intelligence together with software-defined networking and network function virtualization (SDN and NFV) concerning the AI/ML techniques of supervised, unsupervised, reinforcement, deep reinforcement, and federated learning in relation to the resource management and optimization, network slicing, resource allocation, traffic engineering, anomaly detection, and predictive maintenance of the systems. Literature contributions are mainly aimed at enhancing network latency, minimizing loss, optimizing bandwidth usage, and securing mechanisms, but there are still domain cross-silo, operational cross-vendor, computational, data, and energy dependency burdens. These aspects, together with the challenges of deploying AI models in practice, model interpretability, and unsupervised yet complex, intelligent, and self-sustainable cross-silo active inter-functioning systems, are kept in focus in this paper to inform the upcoming research of next-generation wireless networks in a secure, scalable, and robust manner.