8 May 2020 in Journal of Lightwave Technology
A data-driven fiber channel modeling method based on deep learning (DL) is introduced in an optical communication system. In this study, bidirectional long short-term memory (BiLSTM) is selected from a diverse range of DL algorithms to perform fiber channel modeling for on—off keying and pulse amplitude modulation 4 signals. Compared with the conventional model-driven split-step Fourier (SSF)-based method, the proposed method yields similar results based on the comprehensive comparison of multiple characteristics associated with the generated optical signals, including the optical amplitude and phase waveforms in the time domain, optical spectrum components in the frequency domain, and eye diagrams after detection in the electrical domain. Additionally, the effects of multiple factors on the modeled fiber channel have also be investigated, including fiber length, fiber nonlinearity, dispersion, data pattern, pulse shaping, and sample rate. The satisfactory fitting results and acceptable mean square errors indicate that the approximate transfer function of the fiber channel is learned by the BiLSTM. Moreover, compared with repetitive iteration SSF, the computing time is significantly reduced by the BiLSTM owing to its independence on fiber length and insensitivity to data size and launch power. Our aim is to demonstrate the BiLSTM is comparable with the conventional model-driven SSF-based method for direct-detection optical fiber system. We think the proposed method could be a supplementary technique that can be used for the existing simulation system and could also be a potential option for future simulation methods.
12 July 2017 in IEEE Photonics Technology Letters
An $m$ -ary adaptive demodulator based on machine learning for light beams carrying orbital angular momentums (OAMs) over free-space turbulence channels is proposed and demonstrated. Benefiting from natural advantages in the image recognition, convolutional neural network (CNN) is selected to construct the adaptive demodulator. Without extra space light modulators and digital signal processing at the reception, the adaptive demodulator transforms the sequence of intensity patterns of received Laguerre-Gaussian beams carrying different OAM modes into initial signals efficiently. As comparison, K-nearest neighbor (KNN), naive Bayes classifier (NBC), and back-propagation artificial neural network (BP-ANN) are also studied. Furthermore, the demodulating accuracy of 4-, 8-, and 16-ary OAM is investigated with the comprehensive consideration of the atmospheric turbulence, OAM mode spacing, and transmission distance. The simulation results show that the demodulating error rate (DER) of CNN outperforms KNN, NBC, and BP-ANN, especially under stronger turbulence and longer distance. The DER of CNN is ~0.86% for the 1000-m 8-OAM system under strong turbulence, ~30 % less than those of KNN, NBC, and BP-ANN.
16 October 2019 in IEEE Photonics Journal
21 August 2019 in IEEE Photonics Journal
In the coherent long-reach passive optical networks (LRPON), it is crucial to propose cost-effective digital signal processing (DSP) technologies to reduce the overall complexity and power consumption. This paper has proposed a low-complexity chromatic dispersion (CD) estimation scheme based on deep neural networks (DNN) and the error vector magnitude (EVM). To add comparisons, the performances of CD estimation schemes using other two well-known machine learning algorithms including the k-nearest neighbor (KNN) and the decision tree (DT) have also been investigated. The simulation results show that the proposed CD estimation scheme is effective in the coherent LRPON with the quadrature phase shift keying (QPSK) and 16-ary quadrature amplitude modulation (QAM) systems at 14Gbaud rate, 28Gbaud rate and 56Gbaud rate. The comprehensive performances of the DNN outperform those of the KNN and the DT. The mean estimation error of the DNN is less than 20ps/nm within the 100km access distance in the 28Gbaud QPSK/16QAM systems. What's more, compared with classical methods using the CD scanning and frequent domain equalizers (FDE), the computation complexity of proposed CD estimation schemes based on the DNN-EVM has been respectively reduced by 72.3 times, 86.7 times and 2.8 times about the amount of multipliers, adders and comparators.
31 January 2023 in Journal of Lightwave Technology
Digital twin (DT) as one promising technology has been introduced in optical networks, whose role is to achieve interactive functions and intelligent control by building a bridge between physical space and digital space. In this paper, a multifactor-associated network topology portrait (NTP) scheme is proposed to draw a dynamic, real-time, and comprehensive topology representation for DT optical network (DTON). Multiple parameters from nodes and links are jointly evaluated, including link distance, link loss, optical signal-to-noise ratio, bandwidth utilization rate, load capacity, node failure rate, and link failure rate, involving both network layer and physical layer. Based on the NTP, the performance of dynamic routing computation combined with six different routing algorithms is studied, Here, the time for routing computation is sacrificed a little bit in exchange for significant improvements in outage probability, outage latency, and the number of service requests.
5 March 2025 in Journal of Lightwave Technology
As the key component that facilitates long-haul transmission in optical fiber communications by increasing capacity and reducing costs, accurate characterization and gain settings of erbium-doped fiber amplifiers (EDFAs) are essential for quality of transmission estimation and system configuration optimization. However, it is difficult to construct accurate and reliable EDFA models due to complex physical mechanisms and dynamic loading conditions. Although some mathematical and data-driven models have been proposed, their practical applications will face limitations of intricate parameter measurements and high data requirements, respectively. To overcome limitations of both methods, a physics-informed machine learning (PIML) method for parameter identification and gain estimation of EDFA is proposed, which greatly reduces the data requirements by embedding physical prior knowledge in the neural network. In this approach, the gain of EDFA can be accurately estimated by a physics-informed neural network (PINN)-based forward model when parameters including absorption, gain, saturation, and background loss are known. For practical scenarios where parameters are unknown, PINN-based inverse models are established first to identify actual values of parameters from only several sets of input-output data pairs, and PINN-based forward models are accordingly established for gain estimation with identified values. Moreover, an experimental system is constructed to verify the feasibility and performance of proposed method in practical scenarios. Results show that PIML-based method can effectively identify physical parameters from measured data, and better gain estimation results are achieved with mean absolute error of 0.127 dB and standard deviation of 0.065 dB using identified values than typical values of parameters.