Intelligent Routing Optimization Model For QoS-aware Software-Defined Networks Using Deep Reinforcement Learning

Abstract

This thesis presents an intelligent routing model for Software-Defined Networks (SDNs), taking into account Quality of Service (QoS) using Deep Reinforcement Learning (DRL). The agent, based on a DQN network, selects the best paths according to real-time network conditions. Results show significant improvement over the Dijkstra algorithm: up to 60% reduction in delay, 50% less jitter, and a 207% increase in throughput. The model is flexible, scalable, and suitable for data centers, IoT, and 5G networks.