基于遗传算法的边缘光纤传感网动态资源调度策略  

Dynamic Resource Scheduling Strategy of Edge Optical Fiber Sensor Network Based on Genetic Algorithm

作　　者：赵俊 谢岩[1] 苏轩 朱彦霖 罗曦 严敏 甘战 侯云川 常嘉瑞 ZHAO Jun;XIE Yan;SU Xuan;ZHU Yanlin;LUO Xi;YAN Min;GAN Zhan;HOU Yunchuan(Faculty of Electronic Information Engineering,Huaiyin Institute of Technology,Huaian Jiangsu 223003,China;National&Local Joint Engineering Research Center for Optical Sensing/Communications Integrated Networking,Southeast University,Nanjing Jiangsu 210096,China;Qujing Bureau of CSG EHV Power Transmission Company,Qujing Yunnan 655000,China)

机构地区：[1]淮阴工学院电子信息工程学院,江苏淮安223003 [2]东南大学光传感/通信综合网络国家地方联合工程研究中心,江苏南京210096 [3]中国南方电网有限责任公司超高压输电公司曲靖局,云南曲靖655000 [4]不详

出　　处：《传感技术学报》2025年第1期82-87,共6页Chinese Journal of Sensors and Actuators

基　　金：航空科学基金项目(2019ZH069003);南京市科技计划项目(202202025);中国南方电网公司科技项目(010800KK52220001)。

摘　　要：提出一种基于异步光包交换及波分复用技术的分布式边缘光纤传感系统(AOPS-WDM-DEFSS)方案,可支持边缘计算场景下海量、异构、突发感知业务的交换和传输需求,面向AOPS-WDM-DEFSS边缘侧业务特性,提出基于遗传算法的边缘计算集群动态资源与任务调度策略,建立了动态资源调度集群模型与任务模型,定义了完成时间T、通信负载L和使用费用C三个优化目标函数,仿真分析了任务数n、任务相关度REV、传输处理速度比RAT、及权重(α、β和γ)对目标函数的影响。结果显示,当n增加至50时,T、L、C分别增加至15.11 s、641 MB和¥443.2;当REV增加至5时,T和L分别增加至20.1 s和1932 MB;当RAT增加至25时,T和L分别增加至11.4 s和1724 MB;α、β和γ分别达到0.8和0.9后,T、L、C的归一化值趋近于1。A distributed edge fiber sensing system(AOPS-WDM-DEFSS)scheme based on asynchronous optical packet switching and wavelength division multiplexing technology is proposed,which can support the switching and transmission requirements of massive,het-erogeneous,and burst sensing services in edge computing scenarios.Based on the service characteristics of the edge side of AOPS-WDM-DEFSS,a dynamic resource and task scheduling strategy for edge computing clusters using genetic algorithms is proposed.The dynamic resource scheduling cluster model and task model are established,and the three optimization objective functions of completion time(T),communication load(L)and usage cost(C)are defined,and the effects of task number(n),task relevance(REV),transmission processing speed ratio(RAT),and weights(α,βandγ)on the objective function are simulated.The results show that when n increases to 50,T,L and C increase to 15.11 s,641 MB and¥443.2,respectively.When REV increases to 5,T and L increase to 20.1 s and 1932 MB,respectively.When RAT increases to 25,T and L increase to 11.4 s and 1724 MB,respectively.Afterα,βandγreach 0.8 and 0.9,respectively,the normalized values of T,L and C tend to 1.

关 键 词：光纤传感网 边缘计算 分布式 遗传算法 动态资源调度 

分 类 号：TN915[电子电信—通信与信息系统]