基于Q学习的无线通信网多目标智能路由策略  

Q-Learning Based Multi-Objective Intelligent Routing Strategy for Wireless Communication Networks

作　　者：于佳禾 胡春燕 周园 YU Jia-he;HU Chun-yan;ZHOU Yuan(Beijing Aerocim Technology Co.,Ltd,Beijing 102308,China)

机构地区：[1]北京航天晨信科技有限责任公司,北京102308

出　　处：《计算机仿真》2024年第3期431-435,447,共6页Computer Simulation

摘　　要：针对无线通信网络易受环境影响、拓扑结构变化迅速带来的服务质量(Quality of Service,QoS)保障需求和快速响应需求,提出一种基于Q学习的智能路由策略,利用强化学习技术能与环境实时交互的优势,以转控分离、网络可编程的软件定义网络(Software Defined Networking,SDN)架构为路由策略的部署平台,考虑路由跳数和网络流量对时延、带宽、丢包率等QoS的需求设计状态、动作、奖励等强化学习的基本要素,构建智能体,并搭建基于SDN的仿真环境,利用SDN控制器收集网络节点信息、链路状态信息开展智能体训练与测试,打造智能化的路由方案。仿真结果表明,所提方法能够根据网络中的实时信息和用户需求提供定制化的路由策略,进行动态智能选路,并进行快速响应。Aiming at the quality of service(QoS)guarantee requirements and fast response requirements brought about by rapid changes in the topology of wireless communication networks and being easily affected by the environment,this paper proposes an intelligent routing strategy based on Q-learning.Utilizing the advantage of reinforcement learning technology for real-time interaction with the environment,a software defined networking(SDN)architecture with transfer control separation and network programmability is used as a deployment platform for routing strategies,considering the number of routing hops and network traffic on delay,bandwidth,and packet loss rate to meet the requirements of QoS,design the basic elements of reinforcement learning such as state,action,and reward,build an agent,and build an SDN-based simulation environment.The SDN controller is used to collect network node information and link status information for intelligent agent training and testing,and create intelligent routing schemes.The simulation results show that the method in this paper can provide customized routing strategies according to the real-time information in the network and user requirements,perform dynamic intelligent routing,and respond quickly.

关 键 词：无线通信网 软件定义网络 路由算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]