pith. sign in

arxiv: 2606.04840 · v1 · pith:ZXXYEHE6new · submitted 2026-06-03 · ⚛️ physics.optics

Reinforcement Learning-Enabled Agent for Transmitter Optimization in Digital-Analog Radio-over-Fiber Fronthaul

Junhao Zhao , Huayuan Qin , Ouhan Huang , Zhongya Li , Chengxi Wang , Boyu Dong , Liangtao Chen , Xuyu Deng
show 9 more authors
An Yan Penghao Luo Renle Zheng Yongzhu Hu Aolong Sun Yinjun Liu Sizhe Xing Nan Chi Junwen Zhang
This is my paper

Pith reviewed 2026-06-28 04:44 UTC · model grok-4.3

classification ⚛️ physics.optics
keywords digital-analog radio-over-fiberreinforcement learningtransmitter optimizationfronthaulSNR improvementQAM modulationparameter tuningoptical communication
0
0 comments X

The pith

A reinforcement learning agent optimizes transmitter parameters in digital-analog radio-over-fiber systems from end-to-end SNR feedback alone.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper shows that an RL agent can tune four tightly coupled transmitter settings in DA-RoF fronthaul by treating SNR as its only reward signal. Traditional grid search over these parameters is too slow for practical use, so the agent learns through sequential adjustments without any differentiable model of the optical channel. Experiments confirm the agent raises SNR by roughly 2.7 dB over baseline methods for one- to four-order transmissions. The resulting SNRs support modulation formats up to 65536-QAM. The approach therefore supplies an online, hardware-efficient alternative to manual or exhaustive optimization.

Core claim

An RL-enabled agent architecture learns optimal values for rounding factor, scaling factor, geometric shaping factor, and pre-equalization tap coefficients directly from end-to-end SNR feedback, steadily improves SNR through sequential decisions, and outperforms baseline optimization by approximately 2.7 dB, reaching final SNRs of 35.8 dB, 42.9 dB, 53.8 dB, and 63.2 dB that support 1024-, 4096-, 16384-, and 65536-QAM, respectively, in 1- to 4-order DA-RoF experimental transmissions.

What carries the argument

The RL agent that selects transmitter parameter adjustments at each step based solely on observed SNR reward.

If this is right

  • Higher-order QAM formats become feasible in DA-RoF fronthaul without manual parameter tuning.
  • Optimization can run online during operation instead of requiring offline grid searches.
  • The same agent structure works across different transmission orders without redesign.
  • Hardware efficiency improves because no differentiable channel model is needed.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The method could be tested on other optical links where multiple analog and digital parameters interact.
  • If SNR feedback remains reliable under field conditions, the approach might reduce the need for periodic manual recalibration in deployed networks.
  • Extending the reward signal to include latency or power metrics would be a direct next measurement.

Load-bearing premise

End-to-end SNR feedback by itself is enough for the agent to discover good parameter settings and that the laboratory setup matches real-world fronthaul conditions.

What would settle it

Running the trained agent on a different fiber length or with added impairments not present in the original experiments and observing no SNR gain beyond the baseline or inability to support the claimed QAM orders.

Figures

Figures reproduced from arXiv: 2606.04840 by An Yan, Aolong Sun, Boyu Dong, Chengxi Wang, Huayuan Qin, Junhao Zhao, Junwen Zhang, Liangtao Chen, Nan Chi, Ouhan Huang, Penghao Luo, Renle Zheng, Sizhe Xing, Xuyu Deng, Yinjun Liu, Yongzhu Hu, Zhongya Li.

Figure 1
Figure 1. Figure 1: RL-enabled DA-RoF fronthaul architecture. Cloud Network CU/DU Fronthaul RUs RUs Agent-enabled CU/DU Updating Parameters Agent Optimized signal Fronthaul transmission Deployed RUs Reward feedback [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. Figure 3: System impairments in IMDD DA-RoF system: (i) amplitude nonlinear of digital and analog signals induced by amplifier; (ii) limited bandwidth induced by system components. ECL: external cavity laser; EA: electrical amplifier; MZM: Mach-Zehnder Modulator; SSMF: standard single-mode fiber; PD: photodiode. SSMF (i) PD (ii) ECL MZM EA Digital signal Power Amplification Amplitude nonlinear round(·) decision Deci… view at source ↗
Figure 4
Figure 4. Figure 4: Proposed RL-enabled DA-RoF fronthaul architecture. (a) DA-RoF fronthaul agent. (b) N-order DA-RoF fronthaul system. DA-RoF Mod. & GS & Pre-EQ. Update parameters Evaluate Q-network Target Q-network Wireless signal Cascade digital parts Residual analog part Fronthaul Recovered wireless signal DSP Return SNR as a reward TDM N-order DA-RoF Fronthaul System  Reply buffer Mini-batch  t 1 r + () , , , rf t t t … view at source ↗
read the original abstract

Digital-analog radio-over-fiber (DA-RoF) has emerged as a promising fronthaul solution that combines the high spectral efficiency of analog transmission with the robustness of digital transmission. However, the performance of DA-RoF critically depends on several tightly coupled parameters, including the rounding factor (RF), scaling factor (SF), geometric shaping (GS) factor, and pre-equalization taps coefficients, which jointly affect quantization noise, nonlinear distortion, and bandwidth-induced inter-symbol interference (ISI). Conventional grid search-based optimization is computationally prohibitive and impractical for optical communication. In this work, we propose a reinforcement-learning (RL)-enabled DA-RoF fronthaul agent architecture, capable of autonomously learning optimal transmitter parameters from end-to-end signal-to-noise ratio (SNR) feedback without a differentiable channel model. Experimental results demonstrate that the trained agent steadily improves SNR through sequential decision making and outperforms baseline, achieving ~2.7-dB SNR improvement for 1- to 4-order DA-RoF transmission, reaching final SNR of 35.8 dB, 42.9 dB, 53.8 dB, and 63.2 dB and supporting 1024-, 4096-, 16384-, 65536-quadrature amplitude modulation (QAM) format, respectively. These results validate that the proposed RL-enabled framework provides online, scalable, and hardware-efficient parameter optimization for DA-RoF fronthaul systems, paving the way toward high-order modulation format and intelligent next-generation radio access networks.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper proposes a reinforcement-learning (RL) agent for autonomous optimization of transmitter parameters (rounding factor RF, scaling factor SF, geometric shaping GS factor, and pre-equalization taps) in digital-analog radio-over-fiber (DA-RoF) fronthaul. The agent uses only scalar end-to-end SNR feedback without a differentiable channel model. Experiments claim the trained agent achieves a steady ~2.7 dB SNR improvement over baseline, reaching final SNRs of 35.8 dB, 42.9 dB, 53.8 dB, and 63.2 dB while supporting 1024-, 4096-, 16384-, and 65536-QAM, respectively.

Significance. If the experimental results hold under scrutiny, the work provides a practical demonstration of model-free RL for hardware-in-the-loop optimization of tightly coupled parameters affecting quantization noise, nonlinearity, and ISI in DA-RoF systems. This could enable scalable, online tuning for high-order modulation formats in next-generation fronthaul without exhaustive grid search or channel models, addressing a real deployment bottleneck.

major comments (2)
  1. [Abstract and results section] Abstract and results section: The central claim of reliable ~2.7 dB gain via sequential decision-making rests on the RL agent converging in the joint (RF, SF, GS, pre-eq) space using only scalar SNR. No details are supplied on state representation, action discretization/continuity, exploration schedule (e.g., ε-greedy or entropy), number of independent runs averaged, or learning curves, leaving open whether reported final SNRs reflect robust policy learning or favorable initialization.
  2. [Method and experimental setup sections] Method and experimental setup sections: The assumption that end-to-end SNR feedback alone suffices to escape local maxima caused by the interplay of quantization noise, nonlinearity, and bandwidth-induced ISI is load-bearing but unsupported by any ablation, sensitivity analysis, or comparison against alternative optimizers (e.g., Bayesian optimization) that would confirm global optimality in the reported operating regime.
minor comments (1)
  1. [Abstract] Abstract: The phrasing '1- to 4-order DA-RoF transmission' is ambiguous; clarify whether this refers to modulation order or transmission order.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address each major comment below and will revise the manuscript to incorporate additional details and analyses where appropriate.

read point-by-point responses

  1. Referee: [Abstract and results section] Abstract and results section: The central claim of reliable ~2.7 dB gain via sequential decision-making rests on the RL agent converging in the joint (RF, SF, GS, pre-eq) space using only scalar SNR. No details are supplied on state representation, action discretization/continuity, exploration schedule (e.g., ε-greedy or entropy), number of independent runs averaged, or learning curves, leaving open whether reported final SNRs reflect robust policy learning or favorable initialization.

    Authors: We agree that the current manuscript does not provide these implementation details. In the revised version, we will expand the Methods section to specify: state as a tuple of current parameter values plus recent SNR history; discrete action space with bounded adjustments per parameter; ε-greedy exploration with linear decay; results averaged over 5 independent training runs with reported standard deviation; and include learning curves in a new figure demonstrating consistent convergence behavior across runs. These additions will substantiate that the reported gains arise from learned policies rather than initialization. revision: yes

  2. Referee: [Method and experimental setup sections] Method and experimental setup sections: The assumption that end-to-end SNR feedback alone suffices to escape local maxima caused by the interplay of quantization noise, nonlinearity, and bandwidth-induced ISI is load-bearing but unsupported by any ablation, sensitivity analysis, or comparison against alternative optimizers (e.g., Bayesian optimization) that would confirm global optimality in the reported operating regime.

    Authors: The manuscript presents empirical evidence of steady SNR improvement through sequential decisions but does not include ablations or optimizer comparisons. While the multi-order modulation results indicate effective navigation of the coupled parameter space, we will add a dedicated subsection with (i) sensitivity analysis varying one parameter at a time while holding others fixed and (ii) a direct comparison of the RL agent against Bayesian optimization under identical hardware-in-the-loop conditions. This will provide quantitative support for the sufficiency of scalar SNR feedback. revision: yes

Circularity Check

0 steps flagged

No circularity; results are empirical RL training outcomes on hardware.

full rationale

The paper reports experimental SNR measurements from an RL agent trained on end-to-end feedback in a DA-RoF setup. No derivation chain, equations, or predictions are presented that reduce to inputs by construction. Claims rest on observed performance gains (e.g., ~2.7 dB improvement) rather than any self-definitional, fitted-input, or self-citation load-bearing steps. This is a standard experimental validation with no mathematical reduction to inspect.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The paper describes an experimental RL application with no new mathematical axioms, free parameters, or invented entities; it relies on standard RL techniques applied to the described system.

pith-pipeline@v0.9.1-grok · 5876 in / 1195 out tokens · 38851 ms · 2026-06-28T04:44:11.219820+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

42 extracted references · 27 canonical work pages · 2 internal anchors

  1. [1]

    6G Drivers and Vision,

    NGMN Alliance, “6G Drivers and Vision,” NGMN Alliance, White Paper, Apr. 2021. [Online]. Available: https://www.ngmn.org/wp-content/uploads/NGMN-6G- Drivers-and-Vision-V1.0_final.pdf

    2021

  2. [2]

    Exploring the key technologies and applications of 6G wireless communication network,

    P Li, J Fan, and J Wu, “Exploring the key technologies and applications of 6G wireless communication network,” iScience, vol. 28, no. 5, p. 112281, May 2025, doi: 10.1016/j.isci.2025.112281

    work page doi:10.1016/j.isci.2025.112281 2025

  3. [3]

    6G Wireless Networks: Vision, Requirements, Architecture, and Key Technologies,

    Z. Zhang et al., “6G Wireless Networks: Vision, Requirements, Architecture, and Key Technologies,” IEEE Veh. Technol. Mag., vol. 14, no. 3, pp. 28–41, Sep. 2019, doi: 10.1109/MVT.2019.2921208

    work page doi:10.1109/mvt.2019.2921208 2019

  4. [4]

    Spectral Efficiency Comparison Between Analog and Digital RoF for Mobile Fronthaul Transmission Link,

    H Ji, C Sun, and W Shieh, “Spectral Efficiency Comparison Between Analog and Digital RoF for Mobile Fronthaul Transmission Link,” J. Lightwave Technol., vol. 38, no. 20, pp. 5617–5623, Oct. 2020, doi: 10.1109/JLT.2020.3003123

    work page doi:10.1109/jlt.2020.3003123 2020

  5. [5]

    Analog vs Digital Radio-Over-Fiber: A Spectral Efficiency Debate From the SNR Perspective,

    D Che, “Analog vs Digital Radio-Over-Fiber: A Spectral Efficiency Debate From the SNR Perspective,” J. Lightwave Technol., vol. 39, no. 16, pp. 5325–5335, Aug. 2021, doi: 10.1109/JLT.2021.3102220

    work page doi:10.1109/jlt.2021.3102220 2021

  6. [6]

    Things You Should Know About Fronthaul,

    A. Pizzinat, P. Chanclou, F. Saliou, and T. Diallo, “Things You Should Know About Fronthaul,” J. Lightwave Technol., vol. 33, no. 5, pp. 1077–1083, Mar. 2015, doi: 10.1109/JLT.2014.2382872

    work page doi:10.1109/jlt.2014.2382872 2015

  7. [7]

    CPRI Specification V7 0: Interface Specification,

    CPRI Cooperation, “CPRI Specification V7 0: Interface Specification,” CPRI Cooperation, Interface Specification, Oct. 2015. [Online]. Available: https://www.cpri.info/downloads/CPRI_v_7_0_2015-10- 09.pdf

    2015

  8. [8]

    eCPRI Specification V2 0,

    CPRI Cooperation, “eCPRI Specification V2 0,” May

  9. [9]

    Available: https://www.cpri.info/downloads/eCPRI_v_2.0_2019_05_10c

    [Online]. Available: https://www.cpri.info/downloads/eCPRI_v_2.0_2019_05_10c. pdf

  10. [10]

    Efficient Mobile Fronthaul via DSP-Based Channel Aggregation,

    X. Liu, H. Zeng, N. Chand, and F. Effenberger, “Efficient Mobile Fronthaul via DSP-Based Channel Aggregation,” J. Lightwave Technol., vol. 34, no. 6, pp. 1556– 1564, Mar. 2016, doi: 10.1109/JLT.2015.2508451

    work page doi:10.1109/jlt.2015.2508451 2016

  11. [11]

    314-Tbit/s (576 × 380.16-MHz 5G NR OFDM Signals) SDM/WDM/SCM-Based IF-over-Fiber Transmission for Analog Mobile Fronthaul,

    K. Tanaka et al., “314-Tbit/s (576 × 380.16-MHz 5G NR OFDM Signals) SDM/WDM/SCM-Based IF-over-Fiber Transmission for Analog Mobile Fronthaul,” in Optical Fiber Communication Conference (OFC) 2022, San Diego, California: Optica Publishing Group, 2022, p. W4C.2. doi: 10.1364/OFC.2022.W4C.2

    work page doi:10.1364/ofc.2022.w4c.2 2022

  12. [12]

    10 51-Tbit/s IF-over-Fibre Mobile Fronthaul Link Using SDM/WDM/SCM for Accommodating Ultra High-Density Antennas in Beyond-5G Mobile Communication Systems,

    K. Tanaka et al., “10 51-Tbit/s IF-over-Fibre Mobile Fronthaul Link Using SDM/WDM/SCM for Accommodating Ultra High-Density Antennas in Beyond-5G Mobile Communication Systems,” 2022

    2022

  13. [13]

    1 032-Tb/s CPRI-Equivalent Rate IF-Over- Fiber Transmission Using a Parallel IM/PM Transmitter for High-Capacity Mobile Fronthaul Links,

    S. Ishimura, A. Bekkali, K. Tanaka, K. Nishimura, and M Suzuki, “1 032-Tb/s CPRI-Equivalent Rate IF-Over- Fiber Transmission Using a Parallel IM/PM Transmitter for High-Capacity Mobile Fronthaul Links,” Journal of Lightwave Technology, vol. 36, no. 8, pp. 1478–1484, Apr. 2018, doi: 10.1109/JLT.2017.2787151

    work page doi:10.1109/jlt.2017.2787151 2018

  14. [14]

    1 53-Tbps CPRI-Equivalent Data Rate Transmission with Kramers-Kronig Receiver for Mobile Fronthaul Links,

    S. T. Le, K. Schuh, M. Chagnon, F. Buchali, and H. Buelow, “1 53-Tbps CPRI-Equivalent Data Rate Transmission with Kramers-Kronig Receiver for Mobile Fronthaul Links,” in 2018 European Conference on Optical Communication (ECOC), Rome: IEEE, Sep. 2018, pp. 1–3. doi: 10.1109/ECOC.2018.8535539

    work page doi:10.1109/ecoc.2018.8535539 2018

  15. [15]

    5G; NR; Base Station (BS) radio transmission and reception,

    3GPP, “5G; NR; Base Station (BS) radio transmission and reception,” ETSI, Technical Specification, Jul. 2018. [Online]. Available: https://www.etsi.org/deliver/etsi_ts/138100_138199/138104/1 5.02.00_60/ts_138104v150200p.pdf

    2018

  16. [16]

    300 GHz OFDM electronic terahertz wireless transmission based on PS and DFT-S,

    L Jiang, “300 GHz OFDM electronic terahertz wireless transmission based on PS and DFT-S,” J. Infrared Millim. Waves, vol. 43, no. 5, p. 634, 2024, doi: 10.11972/j.issn.1001-9014.2024.05.008

    work page doi:10.11972/j.issn.1001-9014.2024.05.008 2024

  17. [17]

    Enabling Optical Network Technologies for 5G and Beyond,

    X Liu, “Enabling Optical Network Technologies for 5G and Beyond,” J. Lightwave Technol., vol. 40, no. 2, pp. 358–367, Jan. 2022, doi: 10.1109/JLT.2021.3099726

    work page doi:10.1109/jlt.2021.3099726 2022

  18. [18]

    Hybrid Digital-Analog Radio-over-Fiber (DA-RoF) Modulation and Demodulation Achieving a SNR Gain over Analog RoF of >10 dB at Halved Spectral Efficiency,

    X Liu, “Hybrid Digital-Analog Radio-over-Fiber (DA-RoF) Modulation and Demodulation Achieving a SNR Gain over Analog RoF of >10 dB at Halved Spectral Efficiency,” in Optical Fiber Communication Conference (OFC) 2021, Washington, DC: Optica Publishing Group, 2021, p. Tu5D.4. doi: 10.1364/OFC.2021.Tu5D.4

    work page doi:10.1364/ofc.2021.tu5d.4 2021

  19. [19]

    Cascaded digital–analog radio-over-fiber for efficient SNR scaling at >10 dB per extra bandwidth,

    Y Zhu, Y Xu, W Hu, and Q Zhuge, “Cascaded digital–analog radio-over-fiber for efficient SNR scaling at >10 dB per extra bandwidth,” Opt. Lett., vol. 47, no. 15, p. 3836, Aug. 2022, doi: 10.1364/OL.462631

    work page doi:10.1364/ol.462631 2022

  20. [20]

    1.92-Tb/s CPRI-Equivalent Rate Direct Detection Transmission based on ANN Pre- Equalization for Digital-Analog Radio-over-Fiber Mobile Fronthaul,

    J. Zhao et al., “1.92-Tb/s CPRI-Equivalent Rate Direct Detection Transmission based on ANN Pre- Equalization for Digital-Analog Radio-over-Fiber Mobile Fronthaul,” in Optical Fiber Communication Conference (OFC) 2024, San Diego California: Optica Publishing Group, 2024, p. Tu3K.3. doi: 10.1364/OFC.2024.Tu3K.3

    work page doi:10.1364/ofc.2024.tu3k.3 2024

  21. [21]

    Sensitivity-Improved and Dispersion- Tolerant Lite-Coherent Hybrid Receiver for Digital-Analog Radio-over-Fiber Mobile Fronthaul,

    J. Zhao et al., “Sensitivity-Improved and Dispersion- Tolerant Lite-Coherent Hybrid Receiver for Digital-Analog Radio-over-Fiber Mobile Fronthaul,” J. Lightwave Technol., pp. 1–9, 2025, doi: 10.1109/JLT.2025.3550176

    work page doi:10.1109/jlt.2025.3550176 2025

  22. [22]

    Cloned-Comb Enabled Communication & Clock Distribution Integrated Fronthaul Architecture,

    J. Lin et al., “Cloned-Comb Enabled Communication & Clock Distribution Integrated Fronthaul Architecture,” Optical Fiber Communication Conference, 2025

    2025

  23. [23]

    1λ 10 5Tb/s CPRI-Equivalent Rate 1024-QAM Transmission via Self-Homodyne Digital-Analog Radio-over-Fiber Architecture

    Y. Zhu et al., “1λ 10 5Tb/s CPRI-Equivalent Rate 1024-QAM Transmission via Self-Homodyne Digital-Analog Radio-over-Fiber Architecture”

  24. [24]

    High-fidelity digital–analog hybrid RoF fronthaul link enabled by nonlinear radio signal shaping,

    C. Cheng et al., “High-fidelity digital–analog hybrid RoF fronthaul link enabled by nonlinear radio signal shaping,” Opt. Lett., vol. 49, no. 23, p. 6876, Dec. 2024, doi: 10.1364/OL.541887

    work page doi:10.1364/ol.541887 2024

  25. [25]

    592 Gbps/$\uplambda$ Capacity of Equivalent Fronthaul Channel Based on Time-Interleaved Digital-Analog Radio-Over-Fiber,

    C. Cheng et al., “592 Gbps/$\uplambda$ Capacity of Equivalent Fronthaul Channel Based on Time-Interleaved Digital-Analog Radio-Over-Fiber,” J. Lightwave Technol., vol. 42, no. 5, pp. 1340–1346, Mar. 2024, doi: 10.1109/JLT.2023.3322900

    work page doi:10.1109/jlt.2023.3322900 2024

  26. [26]

    End-to-End Learning for OFDM: From Neural Receivers to Pilotless Communication,

    F Ait Aoudia and J Hoydis, “End-to-End Learning for OFDM: From Neural Receivers to Pilotless Communication,” IEEE Trans. Wireless Commun., vol. 21, no. 2, Art. no. 2, 2022, doi: 10.1109/TWC.2021.3101364

    work page doi:10.1109/twc.2021.3101364 2022

  27. [27]

    End-to-end learning of communications systems without a channel model,

    F A Aoudia and J Hoydis, “End-to-end learning of communications systems without a channel model,” arXiv, pp. 298–303, 2018. 15 > REPLACE THIS LINE WITH YOUR MANUSCRIPT ID NUMBER (DOUBLE -CLICK HERE TO EDIT) <

    2018

  28. [28]

    Intelligent End-to-End Nonlinear Constellation Auto-Optimization in W-band Fiber-MMW Integrated Transmission for 6G Access,

    J. Jia et al., “Intelligent End-to-End Nonlinear Constellation Auto-Optimization in W-band Fiber-MMW Integrated Transmission for 6G Access,” in 2022 Optical Fiber Communications Conference and Exhibition (OFC), Mar. 2022, pp. 1–3

    2022

  29. [29]

    Model-Driven Deep-Learning for End- to-End Optimization in Fiber-Terahertz Communication Systems

    Z. Li et al., “Model-Driven Deep-Learning for End- to-End Optimization in Fiber-Terahertz Communication Systems”

  30. [30]

    Model-Free End-to-End Deep Learning of Joint Geometric and Probabilistic Shaping for Optical Fiber Communication in IM/DD System

    Z. Li et al., “Model-Free End-to-End Deep Learning of Joint Geometric and Probabilistic Shaping for Optical Fiber Communication in IM/DD System”

  31. [31]

    Neural Architecture Search with Reinforcement Learning

    B. Zoph and Q V Le, “Neural Architecture Search with Reinforcement Learning,” Feb 15, 2017, arXiv: arXiv:1611.01578. doi: 10.48550/arXiv.1611.01578

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1611.01578 2017

  32. [32]

    Learning Dexterous In-Hand Manipulation

    OpenAI et al., “Learning Dexterous In-Hand Manipulation,” Jan 18, 2019, arXiv: arXiv:1808.00177. doi: 10.48550/arXiv.1808.00177

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1808.00177 2019

  33. [33]

    Reinforcement Learning: An Introduction,

    R S Sutton and A G Barto, “Reinforcement Learning: An Introduction,” 2018

    2018

  34. [34]

    Over-the-fiber Digital Predistortion Using Reinforcement Learning,

    J. Song et al., “Over-the-fiber Digital Predistortion Using Reinforcement Learning,” in 2021 European Conference on Optical Communication (ECOC), Bordeaux, France: IEEE, Sep. 2021, pp. 1–4. doi: 10.1109/ECOC52684.2021.9605972

    work page doi:10.1109/ecoc52684.2021.9605972 2021

  35. [35]

    Reinforcement learning-based complex-valued space-time MIMO 2D-LSTM nonlinear equalizer for photonics-assisted THz indoor optical wireless access networks,

    S. Wang et al., “Reinforcement learning-based complex-valued space-time MIMO 2D-LSTM nonlinear equalizer for photonics-assisted THz indoor optical wireless access networks,” J. Opt. Commun. Netw., vol. 17, no. 9, p. D144, Sep. 2025, doi: 10.1364/JOCN.558913

    work page doi:10.1364/jocn.558913 2025

  36. [36]

    Optimization of Delta-Sigma Modulator Based on Reinforcement Learning for Mobile Fronthaul,

    Z Yan, Y Zhu, G Yang, and W Hu, “Optimization of Delta-Sigma Modulator Based on Reinforcement Learning for Mobile Fronthaul,” IEEE Photon. Technol. Lett., vol. 37, no. 7, pp. 397–400, Apr. 2025, doi: 10.1109/LPT.2025.3546995

    work page doi:10.1109/lpt.2025.3546995 2025

  37. [37]

    Deep Reinforcement Learning Based Joint Allocation Scheme in a TWDM-PON-Based mMIMO Fronthaul Network,

    Y. Cheng, Y. Shao, S. Ding, and C.-K Chan, “Deep Reinforcement Learning Based Joint Allocation Scheme in a TWDM-PON-Based mMIMO Fronthaul Network,” IEEE Photonics J., vol. 16, no. 3, pp. 1–11, Jun. 2024, doi: 10.1109/JPHOT.2024.3388571

    work page doi:10.1109/jphot.2024.3388571 2024

  38. [38]

    Coherent digital-analog radio-over- fiber (DA-RoF) system with a CPRI-equivalent data rate beyond 1 Tb/s for fronthaul,

    Y. Xu et al., “Coherent digital-analog radio-over- fiber (DA-RoF) system with a CPRI-equivalent data rate beyond 1 Tb/s for fronthaul,” Opt. Express, vol. 30, no. 16, p. 29409, Aug. 2022, doi: 10.1364/OE.457586

    work page doi:10.1364/oe.457586 2022

  39. [39]

    Power Amplifier Modeling Framework for Front-End-Aware Next- Generation Wireless Networks,

    K Kostrzewska and P Kryszkiewicz, “Power Amplifier Modeling Framework for Front-End-Aware Next- Generation Wireless Networks,” Electronics, vol. 13, no. 9, p. 1643, Apr. 2024, doi: 10.3390/electronics13091643

    work page doi:10.3390/electronics13091643 2024

  40. [40]

    J. G. Proakis and M. Salehi, Digital communications, 5th ed. Boston: McGraw-Hill, 2008

    2008

  41. [41]

    5G; NR; Base Station (BS) conformance testing; Part 1: Conducted conformance testing,

    3GPP, “5G; NR; Base Station (BS) conformance testing; Part 1: Conducted conformance testing,” ETSI, Technical Specification, Jan. 2023

    2023

  42. [42]

    1 02Tb/s CPRI- Equivalent Rate Direct Detection Transmission Supporting 1024-QAM Using IQ Interleaved Digital-Analog Radio-over- Fiber for Mobile Fronthaul,

    Y Zhu, Q Zhuge, and W Hu, “1 02Tb/s CPRI- Equivalent Rate Direct Detection Transmission Supporting 1024-QAM Using IQ Interleaved Digital-Analog Radio-over- Fiber for Mobile Fronthaul,” in 2022 Asia Communications and Photonics Conference (ACP), Shenzhen, China: IEEE, Nov. 2022, pp. 688–692. doi: 10.1109/ACP55869.2022.10088671

    work page doi:10.1109/acp55869.2022.10088671 2022