Artificial intelligence-driven autonomous optical networks: 3S architecture and key technologies  

作　　者：Yuefeng JI Rentao GU Zeyuan YANG Jin LI Hui LI Min ZHANG 

机构地区：[1]State Key Laboratory of Information Photonics and Optical Communications,School of Information and Communication Engineering,Beijing University of Posts and Telecommunications,Beijing 100876,China [2]State Key Laboratory of Information Photonics and Optical Communications,Institute of Information Photonics and Optical Communications,Beijing University of Posts and Telecommunications,Beijing 100876,China

出　　处：《Science China(Information Sciences)》2020年第6期3-26,共24页中国科学（信息科学）（英文版）

基　　金：supported by National Key R&D Program of China(Grant No.2018YFB1800802);National Natural Science Foundation of China(Grant No.61871051)。

摘　　要：In the optical networks,the dynamicity,the complexity and the heterogeneity have dramatically increased owing to the deployment of advanced coherent techniques,and the optical cross-connect technologies and diverse network infrastructures pose great challenges in the optical network management and maintenance for the network operators.In this review,we propose a"3 S"architecture for AI-driven autonomous optical network,which can aid the optical networks operated in"self-aware"of network status,"self-adaptive"of network control,and"self-managed"of network operations.To support these functions,a number of artificial intelligence(AI)-driven techniques have been investigated to improve the flexibility and the reliability from the device aspect to network aspect.Adaptative erbium-doped fiber amplifier(EDFA)controlling is an example for the device aspect,which provides a power self-adaptive capability according to the network condition.From the link aspect,adaptive fiber nonlinearity compensation,optical monitoring performance and quality of transmission estimation are developed to monitor and alleviate the link-dependent signal impairments in an automatic way.From the network aspect,traffic prediction and network state analysis methods provide the self-awareness,while automatic resource allocation and network fault management powered by AI enhance the self-adaptiveness and self-management capabilities.Benefit from the sufficient network management data,powerful data-mining capability and matured computation units,these AI techniques have great potentials to provide autonomous features for optical networks,including the network resource scheduling and the network customization.

关 键 词：artificial intelligence optical networks self-aware SELF-ADAPTIVE self-managed 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TN929.1[自动化与计算机技术—控制科学与工程]