ARL-RA: Efficient Resource Allocation in 5G Edge Networks: A Novel Intelligent Solution Using Approximate Reinforcement Learning Algorithm

Abstract- The rapid proliferation of fifth-generation (5G) technology has resulted in a wide range of applications, posing challenges in managing network resources effectively and efficiently. To address these challenges, network slicing (NS) and Fog-Radio Access Networks (F-RAN) have emerged as key technologies, enabling the creation of isolated virtual networks on shared physical infrastructure to support high-bandwidth and low-latency communication. However, allocating network resources to latency-sensitive applications like self-driving cars and remote surgery, while ensuring a quality experience, is complex due to stringent latency requirements and limited availability. In this paper, we propose a novel approach, leveraging the Q-learning algorithm, specifically the Approximate Reinforcement Learning for dynamic resource allocation (RA-ARL), in the context of 5G environments. Our modified algorithm takes into account crucial network attributes, introduces innovations such as service type classification based on latency sensitivity, and considers time-varying resource conditions and service demands. We propose an RL model to optimize network utility, focusing on the F-RAN model. Our experimental results demonstrate the effectiveness of ARL-RA in terms of convergence, resource utilization, and the ability to handle user request rejections. This work contributes to the advancement of efficient and effective resource allocation in dynamic 5G networks, particularly for latency-sensitive applications with stringent quality requirements.