针对传统网络路由算法在网络负载增加、流量变化快速时常常出现效率不高或性能下降的问题,将Q-Learning算法与网络拓扑(Network Topology)相结合,提出一种基于强化学习的路由算法(NTQ-Learning)并应用于软件定义网络(Software Defined Networking,SDN)中.将强化学习算法引入SDN的路由决策过程,通过定义状态空间、动作空间和奖励机制,使其在SDN中动态学习并找到最优路由策略,以在不同的网络负载下最小化网络延迟,从而提高网络的整体性能和效率.仿真实验结果表明,与传统算法Dijkstra相比,NTQ-Learning算法能够快速适应流量变化,在网络负载增大时有效降低时延.
Routing Algorithm Based on Reinforcement Learning in SDN
Traditional routing algorithms often have low efficiency or performance degradation when faced with challenges such as increasing network load and rapid traffic changes.It was combined Q-Learning with Network Topology to propose a reinforcement learning-based routing algorithm called NTQ-Learn-ing,which was applied in Software Defined Networking(SDN).Reinforcement learning algorithm was in-troduced into the routing decision-making process of SDN.By defining state space,action space and re-ward mechanism,reinforcement learning algorithm can dynamically learn and find the optimal routing strategy in SDN to minimize network delay under different network loads,thus improving the overall per-formance and efficiency of the network.Simulation results show that compared with Dijkstra algorithm,NTQ-Learning algorithm can quickly adapt to traffic changes and effectively reduce the delay when the net-work load increases.
software defined networkingreinforcement learningrouting decision
万方数据
