An Innovative Fuzzy Logic And Reinforcement Learning Hybrid Model For Adaptive Traffic Management In Software-Defined Networks

Abstract

Software-Defined Networking (SDN) has been recognized as a disruptive technology in contemporary networking for the provision of dynamic traffic control and centralised control. However, cost-effective management of networks traffic with high demand and dynamic environments is still challenging. To address such limitations, this paper proposes a hybrid Fuzzy-Reinforcement learning traffic management model to strike a balance between scalability, convergence speed and energy efficiency. The proposed scheme makes use of fuzzy logic for an initial quick decision making and reinforcement learning for on-line optimization so it can adjust instantly to network dynamics. The approach is based on performance metrics, such as latency, throughput, packet loss rate, bandwidth utilization and energy consumption, and is applied on the model, under different traffic conditions. Comparison of the new model with existing models, such as DFRDRL and AVRO, further demonstrates the effectiveness and efficiency of the new approach. The performance results show that our approach achieves up to 33% end-to-end latency reduction, 30% energy savings, and better throughput than the conventional methods. As the network scale and traffic load grew, the model kept good performance, which proved the scalability and robustness. And, its rapid convergence rate (adaptively learning and monitoring) enables dynamic stability of the network. Furthermore, the results confirm the effectiveness and the feasibility of the proposed model as a practical and novel solution for SDN traffic management in response to technical and sustainability issues in recent network systems.