Abstract—Simulation using a Network Simulator-3 (NS-3)-based setup enables realistic testing under varying network densities and channel conditions. The proposed Deep Q-Network (DQN)-based approach is benchmarked against recent deep reinforcement learning methods, achieving performance improvements of up to 25.7% in throughput, 31.5% latency reduction, and 27.4% energy savings. Q-Learning offers simpler scalability and faster convergence in low-complexity scenarios. These findings confirm the promise of reinforcement learning as a core enabler for autonomous and adaptive 5G resource management and as an efficient and scalable method for intelligent 5G management. The success of DQN in this domain can be attributed to its integration of deep neural architectures that can learn intricate interactions among channel quality, user demand, and interference patterns, in conjunction with experience replay and target networks. Notably, this method attained significant fairness and energy efficiency without the need for explicit fairness constraints in the reward formulation, underscoring the robustness of the generalizability of the learned policies.
 
Keywords—fifth-Generation (5G), reinforcement learning, Q-Learning, Deep Q-Network (DQN), resource allocation, network slicing


Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).