AI-Based 5G Traffic Management: Simulation, Comparison, and Real-World Performance Analysis

International audience The evolution of 5G networks has introduced complex and high-density traffic conditions requiring intelligent management solutions. Traditional simulation and control systems lack the dynamic adaptability to manage real-time traffic fluctuations efficiently. This paper introduces a simulated framework that uses an AI-enabled model for controlling 5G traffic. The simulation framework was developed in Python and deployed using TensorFlow and Keras libraries. It mimicked a realistic 5G network environment with multiple traffic slices and varying Quality of Service (QoS) requirements. The simulation incorporates real-world network parameters, enabling accurate and adaptive traffic management. Traditional network simulation tools such as NS-3 and OMNeT++ have long served as essential platforms for modelling wireless traffic behaviour. While they provide foundational insights, these tools rely heavily on deterministic and stochastic models, which often fail to address the dynamic and heterogeneous nature of ultra-dense 5G environments. The framework incorporates real-world parameters and deep learning models to dynamically adapt to traffic variations and optimise resource utilisation. A comparative analysis demonstrates that AI-based simulation outperforms traditional traffic control methods in terms of scalability, efficiency, and responsiveness. Metrics, including latency, packet loss, and throughput, were evaluated. Experimental evaluation using real-world network traffic data demonstrates substantial improvements across three critical metrics: 38% reduction in latency, 27% increase in throughput, and 41% decrease in packet loss rates compared to conventional approaches. The results validate the effectiveness of AI-integrated simulations in enhancing Quality of Service (QoS) and managing dynamic traffic in ultra-dense 5G environments. In conclusion, this research underlines that AI-driven frameworks are not just a theoretical advantage but a practical necessity for achieving the ambitious performance targets of 5G and beyond.