面向QoS需求的分簇自组织网络路由算法  

QoS Routing Algorithm in Clustered Self-Organizing Networks

作　　者：杨灿 罗涛 刘颖 李泽旭 徐永庆 YANG Can;LUO Tao;LIU Ying;LI Zexu;XU Yongqing(China Electronic Technology Group Corporation Seventh Research Institute,Guangzhou 510000,China;School of Information and Communication Engineering,Beijing University of Posts and Telecommunications,Beijing 100876,China)

机构地区：[1]中国电子科技集团公司第七研究所,广州510000 [2]北京邮电大学信息与通信工程学院,北京100876

出　　处：《北京邮电大学学报》2022年第1期1-6,共6页Journal of Beijing University of Posts and Telecommunications

基　　金：科技部国家重点研发计划项目(2017YFB 0503000)。

摘　　要：基于分布式分簇的网络管理架构,网络节点可以被划分成多个管理域,并由相应区域的簇首进行协同管理。为实现分布式网络场景中,业务差异化的服务质量(QoS)需求与多维度网络资源之间的高效按需匹配,提出了一种基于强化学习的路由调度算法,以降低端到端的时延和防止网络拥塞为目标,优化调度路径。所提算法可以通过簇首集中式和节点分布式2种方式实现,可以解决分布式环境下全局资源信息不完备的问题,有效保证跳变环境下网络的健壮性。将100个节点划分为4个管理域进行仿真验证。仿真结果表明,所提算法可以有效地降低业务的平均时延,并且在业务拒绝率、网络资源利用率方面均优于传统方法。Based on the distributed and clustering network architecture,network nodes can be divided into multiple clusters,which can be managed by their corresponding cluster heads in a collaborative manner.In order to achieve on-demand,efficient matches between the differentiated quality of service(QoS)requirements and the multi-dimensional network resources,a reinforcement learning-based routing algorithm is proposed.The proposed algorithm aims to reduce end-to-end delay and prevent congestion by optimizing the routing path,which can be implemented in both centralized cluster heads and distributed nodes,so as to guarantee the robustness in a dynamic environment.The performance is evaluated by numerical simulations in a network with 100 nodes divided into four regions.The simulation results illustrate that the proposed algorithm can reduce average latency significantly.Besides,the algorithm proposed is superior to the minimum-hop method in terms of rejection rate and resource utilization.

关 键 词：多跳网络 服务质量路由 强化学习 

分 类 号：TN929.52[电子电信—通信与信息系统]