5G Network Resource Allocation Using Reinforcement Learning

Abstract: The rapid evolution of 5G and LTE networks, coupled with the increasing diversity of user requirements, has necessitated intelligent and dynamic resource allocation techniques. This paper explores the application of reinforcement learning (RL) algorithms—specifically Deep Q-Network (DQN), Proximal Policy Optimization (PPO), and Mean Field Qlearning (MFQ)—to optimize network resource allocation based on critical Quality of Service (QoS) parameters such as packet loss rate and packet delay. Utilizing a real-world-inspired dataset that includes multiple dimensions like time of day, user equipment categories, and usage scenarios across various sectors (healthcare, public safety, smart cities, etc.), an OpenAI Gymcompatible environment is developed to simulate real-time allocation challenges. Each RL agent learns to allocate network resources efficiently under changing conditions, aiming to minimize packet loss and delay while maximizing system throughput. The paper further includes the deployment of these trained models using Django on a local server, with a user-friendly Reactbased frontend that accepts input features and returns the allocation results from all three models. Comparative analysis of cumulative rewards, convergence rates, and action selections helps determine the effectiveness of each algorithm. The system demonstrates a scalable, intelligent approach to managing the growing demands of next-generation wireless networks, emphasizing practical feasibility and deployment readiness