A5074609945- Optical Network Technologies
- Advanced Optical Network Technologies
- Advanced Photonic Communication Systems
- Semiconductor Lasers and Optical Devices
- Software-Defined Networks and 5G
- Photonic and Optical Devices
- Software System Performance and Reliability
- Advanced Fiber Laser Technologies
- Network Traffic and Congestion Control
- Neural Networks and Applications
- Time Series Analysis and Forecasting
- VLSI and FPGA Design Techniques
- Atomic and Subatomic Physics Research
- Scientific Computing and Data Management
Nokia (France)
2016-2023
Nokia (United States)
2015
Computer Algorithms for Medicine
2014
Laboratoire d'Optique Appliquée
2012-2014
Orange (France)
2011-2012
In this paper, we propose to lower the network design margins by improving estimation of signal-tonoise ratio (SNR) given a quality transmission (QoT) estimator, for new optical demands in brownfield phase, based on mathematical model physics propagation. During greenfield phase and operation, collect correlate information QoT input parameters, issued from established initial available almost free elements: amplifiers output power SNR at coherent receiver side. Since have some uncertainties...
An optical network, like any system, has to be observable before it can become subject optimization, and this is the main capability that ORCHESTRA project introduces. ORCHESTRA's high observability will rely on information provided by coherent transceivers extended, almost for free, operate as software defined performance monitors (soft-OPM). Novel digital signal processing (DSP) OPM algorithms developed combined with a novel hierarchical monitoring plane, cross-layer optimization...
We develop an algorithm extension for a coherent receiver, coupled with machine learning to monitor mechanical stress optical fiber, recognizing fiber breaks before they occur. demonstrate event classification 95% accuracy over real-time PDM-QPSK testbed.
Using monitored physical parameters in a learning process, we decrease design margins by reducing uncertainties on the input of Quality Transmission (QoT) tool, improving accuracy signal-to-noise ratio prediction.
We illustrate the cost reduction during 10-year life of a core WDM network enabled by elastic transponders when accounting directly for end-of-life OSNR detection margins, compared to margins progressively growing with ageing.
By associating machine learning and an analytical model (i.e., the Gaussian noise model), we reduce uncertainties on output power profile figure of each amplifier in optical network. We leverage signal-to-noise ratio (SNR) all light paths network, monitored coherent receivers. The process is based a gradient-descent algorithm where uncertain input parameters are iteratively modified from their estimated values to match with SNR European design margin then reduced 0.1 dB for new traffic demands.
With the advent of novel elastic optical transponders allowing for fine rate granularity, network designers can maximize throughput current installed and future wavelength-division multiplexing (WDM) infrastructures. In this sense, data flow be properly optimized each connection, while avoiding unnecessary margins. Such a connection mode operation should rely on simple mechanism. This paper illustrates how SNR-driven self-optimization connections is relevant WDM networks, especially when...
Network design margins are introduced by quality of transmission estimator inaccuracies. Some those inaccuracies due to uncertainty on the fiber type deployed in optical networks, and value chromatic dispersion fibers. We propose, this paper, a method identify all unknown types (and estimate dispersion) an network reduce said uncertainties. monitor, collect, centralize, correlate values accumulated over each established light path, already measured coherent receivers autonomously without...
Using machine learning and Signal-to-Noise Ratio (SNR) monitoring, we reduce uncertainties on output power profile noise figure (NF) of each EDFA in an optical network. The Quality Transmission (QoT) tool margin is reduced to 0.1dB for new traffic demands. process based a gradient-descent algorithm where the parameters analytical model are iteratively modified match emulated light paths QoT European © 2020 Nokia Bell Labs.
In the present day, evaluation of machine learning (ML) as a candidate for substituting analytical quality transmission (QoT) estimators is done in compartmentalized way. The assessment not produced from global optical network design perspective and with accurate metrics; on contrary, heavily focuses physical layer impairment precision capabilities while underemphasizing effects at layer. this paper, we recommend suitable methodology evaluating QoT substitution based foundational idea that...
We evaluate the potential benefits of finer granularity rate-adaptive transponders to better progressively fit ageing margin in WDM networks. The underlying technology and cost savings are presented for two network core topologies.
We present a linear-to-nonlinear power ratio monitor based on shallow artificial neural network and the optical spectrum. The is trained with experimental pairs of input single-channel spectra output optimal corrections, i.e., amendments that lead to level maximizing performance in terms signal-to-noise ratio. technique tested shows capability providing up 1 dB gain ±3 region around actual power. Furthermore, does not recommend variations resulting severe penalty (max -0.12 dB). Here, we...
We investigate how a machine learning-based QoT estimator performs depending on different features selections, homogeneity of the learned light paths and uncertainty their span lengths using artificial database for France43 network.
We assess the benefits of transfer learning based on artificial neural networks (ANN) using unbiased training data sets for Qo T-estimation unestablished light paths. This study considers from CONUS topology and an set to Germany50 topology.
We present a new strategy of regenerator placement along with traffic growth in the elastic WDM networks, by introducing additional regenerators on already allocated connections only from when needed to accommodate extra demand capacity.
We compare cost savings when planning a WDM French backbone network based on SSMF or LEAF, with 32 GBaud elastic optical transponders adapting their capacity from 100 to 300 Gb/s the actual ageing margins.
We compare the cost efficiency of optical networks based on elastic transponders when accounting for network ageing that impacts link margins. Network design Germany50, France43, and India71 WDM backbone is simulated providing results during a ten-year period, given different throughputs, traffic growth rates erosions transponders. also investigate distribution margins modulation schemes might be changed life network, before its end-of-life. evaluate potential benefits finer granularity...
We present rare-event classification of polarization transients based on field measurements with data augmentation combined robot-generated fiber-disturbance data. compare machine learning methods for accuracy and required number training sample traces.
We compare the resilience to network parameters uncertainty between a GN-model based QoT tool and ML-based ones. study mean square error of estimated G-OSNR novel metric on overestimation probability.
We illustrate the cost savings during 10-year life of a core WDM network enabled by elastic transponders when accounting for low design margins which are progressively growing with ageing, compared to end-of-life OSNR margins.
We present an experimental proof-of-concept on just-in-time resource allocation in elastic optical networks to provide seamless path restoration. Our method relies state of polarization monitoring via standard coherent receiver paired with machine learning for proactive fiber cut detection.