Enhanced Routing Scheduling in SDN Using Dueling Double DQN with Prioritized Experience Replay

To address routing and scheduling challenges in data center networks, this paper proposes a dynamic path planning algorithm leveraging deep reinforcement learning and software-defined networking (SDN). The algorithm integrates Dueling Double Deep Q-Networks (DQN) with Prioritized Experience Replay. Link information is obtained through Software-Defined Networking (SDN), processed, and then provided to the Dueling Double DQN agent for simulation training to determine the optimal path for the current state. Experimental results indicate superior performance over DQN and Double DQN in terms of packet loss rate, average latency, and average bit rate.