Scholarlink

<p>Software-Defined Networking (SDN) has emerged as a key enabler for next-generation communication systems, offering flexible network management and dynamic traffic control. With the increasing demand from latency-sensitive and high-bandwidth applications such as virtual reality and online gaming, achieving Quality of Service (QoS) guarantees in SDN routing has become a critical challenge. Traditional routing mechanisms often fall short in adapting to highly dynamic traffic conditions, leading to degraded performance. To address these challenges, this paper presents a Deep Q-Network with Reward Shaping (DQN-RS) approach for QoS-aware routing in SDN. The proposed method intelligently selects optimal paths for incoming traffic by incorporating a reward-shaping mechanism that guides the learning process to simultaneously satisfy multiple QoS constraints, including bandwidth, delay, and packet loss. The DQN-RS framework is extensively evaluated on both synthetic and real-world network topologies, such as MIRA. Experimental results demonstrate that DQN-RS achieves superior performance across multiple metrics, improving the request acceptance rate by approximately 2–3% compared to the best existing algorithms, while ensuring better network efficiency and reliability.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

DQN-RS: QoS-aware SDN routing optimization using deep Q-networks joining with reward shaping

  • 期刊论文

  • 发表日期：

    2026年6月29日

查看全文
  • Shixin Ge,
  • Mahdi Mir

摘要

Software-Defined Networking (SDN) has emerged as a key enabler for next-generation communication systems, offering flexible network management and dynamic traffic control. With the increasing demand from latency-sensitive and high-bandwidth applications such as virtual reality and online gaming, achieving Quality of Service (QoS) guarantees in SDN routing has become a critical challenge. Traditional routing mechanisms often fall short in adapting to highly dynamic traffic conditions, leading to degraded performance. To address these challenges, this paper presents a Deep Q-Network with Reward Shaping (DQN-RS) approach for QoS-aware routing in SDN. The proposed method intelligently selects optimal paths for incoming traffic by incorporating a reward-shaping mechanism that guides the learning process to simultaneously satisfy multiple QoS constraints, including bandwidth, delay, and packet loss. The DQN-RS framework is extensively evaluated on both synthetic and real-world network topologies, such as MIRA. Experimental results demonstrate that DQN-RS achieves superior performance across multiple metrics, improving the request acceptance rate by approximately 2–3% compared to the best existing algorithms, while ensuring better network efficiency and reliability.