Acta Optica Sinica, Volume. 45, Issue 14, 1420016(2025)

Research Progress of Channel Waveform Modeling in Optical Transmission System Based on Deep Learning (Invited)

Minghui Shi, Zekun Niu**, Hang Yang, Kaiyan Jin, Xinyi Zhou, Zhongyuan Sun, Zhaoyan Zhang, Lilin Yi, and Lilin Yi*
Author Affiliations
  • State Key Laboratory of Photonics and Communications, School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
  • show less
    Figures & Tables(12)
    The structure of a coherent optical transmission simulation system
    The processing of dataset construction
    Classification standards of DL-based optical fiber transmission system channel waveform modeling schemes. (a) Long-haul transmission mode; (b) deep learning model structure; (c) incorporating physical prior knowledge
    Framework diagram of BiLSTM-based fiber channel model[41]
    Structure of CGAN[42]
    Fourier neural operator framework[75]
    Process of modeling nonlinear dynamics in fiber optics using PINN[57]. (a) Temporal-spatial 2D region and exact evolution data generated by SSFM as a reference; (b) discrete-points sampling for PINN; (c) the PINN structure constructed for solving the NLSE in fiber
    Coherent WDM system and FDD modeling scheme details[67]. (a) Schematic of WDM simulation system; (b) transmission processes of FDD-based and SSFM-based optical fiber channel models; (c) generation process for the training dataset; (d) FDD-BiLSTM structure for nonlinearity modeling
    Optical fiber channel modeling frameworks based on the Seq2Seq-FDD scheme[81]. (a) The structure and parameters of optical fiber transmission WDM simulation systems and feature decoupled distributed schemes; (b) the structures of Bi-LSTM employed in Vanilla-FDD and Seq2Seq-FDD; (c) Seq2seq framework implemented by transfer learning processing
    Optical fiber communication simulation system based on S-SSFM-FNO[82]
    Comparison of Q-factor of signals modeled using SSFM and DL-based optical fiber channel models under different launch powers, after processing with CDC, 10 StPS DBP, and 50 StSP DBP algorithms
    • Table 1. Classifications and performance metrics of DL-based channel waveform modeling schemes in optical fiber transmission systems

      View table

      Table 1. Classifications and performance metrics of DL-based channel waveform modeling schemes in optical fiber transmission systems

      Scheme

      Long-haul

      transmission mode

      DL model

      structure

      Incorporation of physical information

      Channel

      number

      Symbol rate /

      GBaud

      Distance /kmTime ratioRef.
      BiLSTM1108020% of SSFM41
      CGAN13010002% of SSFM4270
      FCNN2302003.5% of SSFM71-72
      Multi-head attention1401000

      2.37% of C-SSFM

      dz=0.005 km)

      73-74
      FNO128120025% of SSFM75
      FNO53080076
      DeepONet52880077
      PINO1143200.1% of SSFM78
      FDD-BiLSTM41301040

      3% of NP-SSFM

      dϕ=0.005)

      6779
      FDD-Encoder-Only-Transformer21301200

      1.3% of NP-SSFM

      dϕ= 0.05)

      68
      FDD-Co-LSTM1321600

      10% of C-FFSM

      dz=0.1 km)

      80
      Seq2Seq-FDD514080

      0.5% of NP-SSFM

      dϕ=0.005)

      81
      S-SSFM-FNO21642400

      1.5% of NP-SSFM

      dϕ=0.05)

      82
    Tools

    Get Citation

    Copy Citation Text

    Minghui Shi, Zekun Niu, Hang Yang, Kaiyan Jin, Xinyi Zhou, Zhongyuan Sun, Zhaoyan Zhang, Lilin Yi, Lilin Yi. Research Progress of Channel Waveform Modeling in Optical Transmission System Based on Deep Learning (Invited)[J]. Acta Optica Sinica, 2025, 45(14): 1420016

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Optics in Computing

    Received: Apr. 15, 2025

    Accepted: May. 27, 2025

    Published Online: Jul. 16, 2025

    The Author Email: Zekun Niu (zekunniu@sjtu.edu.cn), Lilin Yi (lilinyi@sjtu.edu.cn)

    DOI:10.3788/AOS250914

    Topics