The development of Internet brings a great challenge to the survivability of the supporting distributed intelligent optical networks. The emergence of Active-Fault-Alarm (AFA) technology makes it possible for the syst...The development of Internet brings a great challenge to the survivability of the supporting distributed intelligent optical networks. The emergence of Active-Fault-Alarm (AFA) technology makes it possible for the system to be aware of the incoming or possible fault in advance and to provide possibility to develop a more active restoration mechanism. On this base, an Active Segment Pre-Restoration (ASPR) mechanism for distributed optical network is proposed. ASPR allows establishing a Segment Pre-Restoration Path (SPR-Path) as a work path, which is initiated by the local node, in advance of potential fault occuring and keeps the SPR-Path only during the low-quality or fault period. Simulation results show that the ASPR mechanism has better restoration performance compared with that of Active Restoration (AR) scheme.展开更多
光分配网络(optical distribution network,ODN)网络是FTTH网络中的重要组成部分,其质量好坏直接关系到客户使用宽带网络的体验。为了对哑资源进行高效精准管理,改善网络质量,提升宽带业务开通效率,降低维护成本,设计优化了哑资源智能...光分配网络(optical distribution network,ODN)网络是FTTH网络中的重要组成部分,其质量好坏直接关系到客户使用宽带网络的体验。为了对哑资源进行高效精准管理,改善网络质量,提升宽带业务开通效率,降低维护成本,设计优化了哑资源智能检测及清查系统。该系统基于深度学习中的YOLOX算法,设计改进提出随机梯度下降动量及Nestrov动量(stochastic gradient descent with momentum and nestrov,SGDMN)优化器,该优化器能有效抑制振荡并且在加速训练的同时对梯度进行校正,以此对参数进行更新;在不脱离真实场景的前提下选择数据增强方式,以此实现对分光器、尾纤、标签、二维码等关键信息进行目标检测,进行分类标识,维护人员可基于这些关键信息配合资管系统信息完成哑资源清查。通过其他经典目标识别算法进行对比实验,结果表明改进后的YOLOX算法精度更高,满足哑资源智能检测及清查实际工程需求。展开更多
We propose and experimentally validate an optical true time delay beamforming scheme with straightforward integration into hybrid optical/millimeter（mm）-wave access networks. In the proposed approach, the most compl...We propose and experimentally validate an optical true time delay beamforming scheme with straightforward integration into hybrid optical/millimeter（mm）-wave access networks. In the proposed approach, the most complex functions, including the beamforming network, are implemented in a central office, reducing the complexity and cost of remote antenna units. Different cores in a multi-core fiber are used to distribute the modulated signals to high-speed photodetectors acting as heterodyne mixers. The mm-wave carrier frequency is fixed to 50 GHz（VBand）, thereby imposing a progressive delay between antenna elements of a few picoseconds. That true time delay is achieved with an accuracy lower than 1 ps and low phase noise.展开更多
基金supported in part by National High Technical Research and Development Program of China (863 Program)under Grant No.2009AA01z255, 2009AA01A345National Key Basic Research Program of China (973 Program) under Grant No.2007CB310705+1 种基金National Natural Science Foundation of China under Grant No. 60932004Specialized Research Fund for the Doctoral Program of Higher Education of China under Grant No.200800130001
文摘The development of Internet brings a great challenge to the survivability of the supporting distributed intelligent optical networks. The emergence of Active-Fault-Alarm (AFA) technology makes it possible for the system to be aware of the incoming or possible fault in advance and to provide possibility to develop a more active restoration mechanism. On this base, an Active Segment Pre-Restoration (ASPR) mechanism for distributed optical network is proposed. ASPR allows establishing a Segment Pre-Restoration Path (SPR-Path) as a work path, which is initiated by the local node, in advance of potential fault occuring and keeps the SPR-Path only during the low-quality or fault period. Simulation results show that the ASPR mechanism has better restoration performance compared with that of Active Restoration (AR) scheme.
文摘光分配网络(optical distribution network,ODN)网络是FTTH网络中的重要组成部分,其质量好坏直接关系到客户使用宽带网络的体验。为了对哑资源进行高效精准管理,改善网络质量,提升宽带业务开通效率,降低维护成本,设计优化了哑资源智能检测及清查系统。该系统基于深度学习中的YOLOX算法,设计改进提出随机梯度下降动量及Nestrov动量(stochastic gradient descent with momentum and nestrov,SGDMN)优化器,该优化器能有效抑制振荡并且在加速训练的同时对梯度进行校正,以此对参数进行更新;在不脱离真实场景的前提下选择数据增强方式,以此实现对分光器、尾纤、标签、二维码等关键信息进行目标检测,进行分类标识,维护人员可基于这些关键信息配合资管系统信息完成哑资源清查。通过其他经典目标识别算法进行对比实验,结果表明改进后的YOLOX算法精度更高,满足哑资源智能检测及清查实际工程需求。
基金founded by H2020 ITN CELTA under Grant No.675683 of Call:H2020-MSCA-ITN-2015
文摘We propose and experimentally validate an optical true time delay beamforming scheme with straightforward integration into hybrid optical/millimeter（mm）-wave access networks. In the proposed approach, the most complex functions, including the beamforming network, are implemented in a central office, reducing the complexity and cost of remote antenna units. Different cores in a multi-core fiber are used to distribute the modulated signals to high-speed photodetectors acting as heterodyne mixers. The mm-wave carrier frequency is fixed to 50 GHz（VBand）, thereby imposing a progressive delay between antenna elements of a few picoseconds. That true time delay is achieved with an accuracy lower than 1 ps and low phase noise.
