pith. sign in

arxiv: 2604.26080 · v1 · submitted 2026-04-28 · 💻 cs.NI · cs.LG

NeuralEmu: in situ Measurement-Driven, ML-based, High-Fidelity 5G Network Emulation

Haoran Wan , Yaxiong Xie , Kyle Jamieson This is my paper

Pith reviewed 2026-05-07 14:31 UTC · model grok-4.3

classification 💻 cs.NI cs.LG
keywords 5G network emulationmachine learning scheduler modeltelemetry-driven emulationmulti-user contentionhigh-fidelity network testbedresource block predictiontraffic reconstruction
0
0 comments X

The pith

NeuralEmu learns real 5G scheduler decisions from telemetry to emulate volatile networks with far lower error than existing tools.

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

Current emulators either sever the feedback between applications and real 5G base stations or rely on oversimplified scheduling logic, leaving a gap when testing latency-sensitive apps. NeuralEmu closes that gap by training machine learning models directly on high-resolution measurements of resource block allocations and modulation choices. The system predicts these allocations for multiple clients at once while reconstructing the hidden traffic patterns of background users to capture realistic contention. When evaluated on web, video, and gaming workloads, it cuts emulation error by roughly half compared with prior approaches. A reader cares because this gives developers a standardized, high-fidelity testbed for protocols that must survive actual cellular dynamics.

Core claim

NeuralEmu is an in-situ, measurement-driven emulator that trains models on live 5G telemetry to predict instantaneous resource block allocations and modulation schemes from user buffer occupancy and channel state, while a separate traffic reconstruction model recovers background user demands so that cross-client contention remains realistic; the resulting Linux middlebox therefore reproduces the packet dynamics seen in commercial 5G deployments for multiple simultaneous applications.

What carries the argument

ML models that map instantaneous per-user buffer occupancy and channel conditions to resource block allocations and modulation schemes, plus an inversion model that recovers background traffic from observed scheduling outcomes.

If this is right

  • Provides a repeatable test environment for real-time interactive protocols that must tolerate 5G scheduler volatility.
  • Achieves 55% lower error on web-page load time, 57% on WebRTC encoder bit rate, and 51% on cloud-gaming one-way delay.
  • Supports multiple simultaneous clients with accurate contention modeling rather than isolated single-user traces.
  • Runs as a high-performance Linux middlebox, allowing unmodified applications to interact with the emulated 5G link.

Where Pith is reading between the lines

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

  • Developers could iterate on latency-critical applications without repeated expensive field trials or access to proprietary base-station code.
  • The same telemetry-to-model pipeline might be adapted to other cellular technologies if comparable high-resolution traces become available.
  • Future protocol designers could use the emulator to stress-test edge cases that are rare in controlled lab setups but common in live networks.

Load-bearing premise

Models trained on telemetry from particular 5G sites and workloads will continue to predict accurate allocations when faced with new networks, traffic mixes, or radio conditions.

What would settle it

Run the same application workloads on a live commercial 5G testbed and on NeuralEmu trained only on earlier traces from different cells; if measured packet delays, bit rates, and load times diverge beyond the reported error reductions, the generalization claim fails.

read the original abstract

Current and future applications demand ultra-low latency and consistent throughput, yet frequently traverse 5G cellular networks, so cope with volatile packet dynamics, as 5G base station schedulers dynamically react to user workloads and wireless channel conditions. The task of evaluating network algorithms in these environments is hamstrung by current tools: record-and-replay emulators sever the feedback interaction that exists between application end points and a commercial operator's proprietary 5G scheduler, while full-stack simulators rely on overly simplistic scheduling logic. To bridge this reality gap, we present NeuralEmu, a high-fidelity, machine learning-based emulation framework that learns complex 5G scheduler resource allocation behaviors directly from extremely high-resolution network telemetry tools. The first emulator to handle multiple clients, NeuralEmu utilizes machine learning to dynamically predict resource block allocations and modulation schemes based on instantaneous user buffer occupancy and channel states. To capture realistic cross-user contention, a traffic reconstruction model inverts cellular network scheduling results to recover the underlying traffic patterns of uncontrolled background users. Implemented as an high-performance Linux middlebox emulator, NeuralEmu reduces emulation error relative to the state of the art for various network applications including but not limited to 55% for web-page load time, 57% for WebRTC encoder bit rate, and 51% for cloud gaming packet one-way delay, providing an accurate, standardized testing ground for tomorrow's real-time interactive network protocols and applications.

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

3 major / 2 minor

Summary. NeuralEmu is a high-fidelity 5G network emulator that trains ML models on high-resolution telemetry to predict resource block allocations and modulation schemes from instantaneous buffer occupancy and channel states, while using a traffic reconstruction model to recover background user patterns for realistic cross-user contention. Implemented as a Linux middlebox, it claims to reduce emulation error relative to the state of the art by 55% for web-page load time, 57% for WebRTC encoder bit rate, and 51% for cloud gaming packet one-way delay across multiple clients.

Significance. If the generalization and fidelity claims hold, NeuralEmu would offer a valuable standardized testing platform for latency-sensitive applications over real 5G schedulers, addressing limitations of both record-and-replay emulators and simplistic simulators. The measurement-driven ML approach is a strength, but its significance is tempered by the absence of demonstrated out-of-distribution performance.

major comments (3)
  1. Abstract: the central claims of 55%/57%/51% error reductions for specific applications are stated without any reference to experimental setup, datasets, baselines (e.g., ns-3 or other emulators), statistical tests, number of runs, or validation procedures, preventing assessment of whether the data support the claims.
  2. Evaluation section (presumed §5): the reported gains appear based on held-out traces from the same deployments and workloads used for training; no cross-operator, cross-location, temporal, or explicit OOD splits are described, which is load-bearing for the claim that the models deliver high-fidelity results on arbitrary unseen networks and channel conditions.
  3. ML model sections (presumed §3–4): the models predicting RB allocations/MCS and performing traffic reconstruction are trained on telemetry from particular 5G deployments, yet no details on architecture, regularization, or explicit tests for extrapolation to different schedulers or traffic patterns are provided, leaving the generalization assumption unverified.
minor comments (2)
  1. Abstract: the phrase 'including but not limited to' before the three examples is redundant and could be removed for conciseness.
  2. Implementation description: the claim that NeuralEmu is 'the first emulator to handle multiple clients' should be supported by a brief comparison table or citation to prior work that handled only single clients.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for the constructive and detailed feedback on our manuscript. We address each major comment below and have made revisions to improve clarity, add missing details, and strengthen the evaluation where possible.

read point-by-point responses

  1. Referee: Abstract: the central claims of 55%/57%/51% error reductions for specific applications are stated without any reference to experimental setup, datasets, baselines (e.g., ns-3 or other emulators), statistical tests, number of runs, or validation procedures, preventing assessment of whether the data support the claims.

    Authors: We agree that the abstract would benefit from additional context. In the revised manuscript, we have updated the abstract to briefly reference the experimental setup (high-resolution telemetry from commercial 5G deployments), baselines (ns-3 and record-and-replay emulators), validation (held-out traces with multiple runs), and statistical reporting of the error reductions. revision: yes

  2. Referee: Evaluation section (presumed §5): the reported gains appear based on held-out traces from the same deployments and workloads used for training; no cross-operator, cross-location, temporal, or explicit OOD splits are described, which is load-bearing for the claim that the models deliver high-fidelity results on arbitrary unseen networks and channel conditions.

    Authors: The primary evaluation uses held-out traces from the same deployments, which is standard for validating ML models on realistic data before broader claims. We have added results on temporal and cross-location splits within the operator to the revised evaluation section. Full cross-operator OOD testing is not possible without additional proprietary datasets, which we now explicitly note as a limitation and direction for future work. revision: partial

  3. Referee: ML model sections (presumed §3–4): the models predicting RB allocations/MCS and performing traffic reconstruction are trained on telemetry from particular 5G deployments, yet no details on architecture, regularization, or explicit tests for extrapolation to different schedulers or traffic patterns are provided, leaving the generalization assumption unverified.

    Authors: We have expanded Sections 3 and 4 in the revision to include complete details on the neural network architectures, input features, regularization techniques (dropout and L2), and training procedures. We have also added explicit extrapolation tests on traces with unseen traffic patterns and channel conditions to better support the generalization claims. revision: yes

standing simulated objections not resolved
  • Cross-operator OOD evaluation, as it requires proprietary telemetry datasets from other 5G operators that were not accessible for this study.

Circularity Check

0 steps flagged

No significant circularity; empirical ML training and evaluation are self-contained

full rationale

The paper describes a measurement-driven ML emulator that trains models on 5G telemetry to predict resource block allocations, MCS, and background traffic patterns, then measures empirical error reductions (e.g., 55% web load time) against baselines on held-out traces. No derivation chain, equations, or first-principles results are presented that reduce any claimed prediction or gain to the fitted parameters or inputs by construction. No self-citations are invoked as load-bearing uniqueness theorems or ansatzes, and the central claims rest on observable performance deltas rather than tautological redefinitions or fitted-input renamings. This is the expected non-finding for a data-driven systems paper whose results are externally falsifiable via new deployments.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Only the abstract is available, so the ledger is limited to the core assumptions stated or implied there.

free parameters (1)
  • ML model parameters
    Neural network weights and hyperparameters are fitted to the collected telemetry to predict allocations and modulation.
axioms (1)
  • domain assumption High-resolution telemetry faithfully captures the proprietary scheduler's allocation logic under the observed conditions
    The entire learning pipeline rests on this fidelity assumption.

pith-pipeline@v0.9.0 · 5567 in / 1330 out tokens · 98568 ms · 2026-05-07T14:31:39.877781+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

57 extracted references · 32 canonical work pages

  1. [1]

    Alomar, P

    A. Alomar, P. Hamadanian, A. Nasr-Esfahany, A. Agar- wal, M. Alizadeh, D. Shah. CausalSim: A causal framework for unbiased Trace-Driven simulation.20th USENIX Symposium on Networked Systems Design and Implementation (NSDI 23), 1115–1147. USENIX Asso- ciation, Boston, MA, 2023. ISBN 978-1-939133-33-5

    2023

  2. [2]

    Z. An, L. Shangguan, J. Kaewell, P. Pietraski, K. Jamieson. Radiotwin: A digital building mate- rial twin for wideband, cross-link, cross-band wireless channel prediction. IEEE, London, UK, 2025. doi: 10.1109/DySPAN64764.2025.11115919

    work page doi:10.1109/dyspan64764.2025.11115919 2025

  3. [3]

    Ananthanarayanan, V

    G. Ananthanarayanan, V . Bahl, L. Cox, A. Crown, S. Nogbahi, Y . Shu. Video analytics - killer app for edge computing.Proceedings of the 17th Annual Interna- tional Conference on Mobile Systems, Applications, and Services, MobiSys ’19, 695–696. Association for Com- puting Machinery, New York, NY , USA, 2019. ISBN 9781450366618. doi:10.1145/3307334.3328589

    work page doi:10.1145/3307334.3328589 2019

  4. [4]

    R. H. Arpaci-Dusseau, A. C. Arpaci-Dusseau.Operat- ing Systems: Three Easy Pieces. Arpaci-Dusseau Books, 1.00 edn., 2018

    2018

  5. [5]

    V . Arun, H. Balakrishnan. Copa: Practical Delay-Based Congestion Control for the Internet.Proceedings of the Applied Networking Research Workshop, 19–19. ACM, Montreal QC Canada, 2018. ISBN 978-1-4503-5585-8. doi:10.1145/3232755.3232783

    work page doi:10.1145/3232755.3232783 2018

  6. [6]

    N. Blum, S. Lachapelle, H. Alvestrand. WebRTC: real- time communication for the open web platform.Com- munications of the ACM,64(8), 50–54, 2021. ISSN 0001-0782, 1557-7317. doi:10.1145/3453182

    work page doi:10.1145/3453182 2021

  7. [7]

    Cardwell, Y

    N. Cardwell, Y . Cheng, C. S. Gunn, S. H. Yeganeh, V . Jacobson. BBR: Congestion-Based Congestion Con- trol: Measuring bottleneck bandwidth and round-trip propagation time.Queue,14(5), 20–53, 2016. ISSN 1542-7730. doi:10.1145/3012426.3022184

    work page doi:10.1145/3012426.3022184 2016

  8. [8]

    Chang, M

    H. Chang, M. Varvello, F. Hao, S. Mukherjee. Can you see me now? a measurement study of zoom, we- bex, and meet.ACM Internet Measurement Conference, 216–228. ACM, New York, NY , USA, 2021. ISBN 9781450391290

    2021

  9. [9]

    K. Cho, B. van Merriënboer, C. Gulcehre, D. Bahdanau, F. Bougares, H. Schwenk, Y . Bengio. Learning Phrase Representations using RNN Encoder–Decoder for Sta- tistical Machine Translation. A. Moschitti, B. Pang, W. Daelemans, eds.,Proceedings of the 2014 Confer- ence on Empirical Methods in Natural Language Pro- cessing (EMNLP), 1724–1734. Association for ...

    work page doi:10.3115/v1/ 2014

  10. [10]

    Network features reference

    Chrome Developers. Network features reference. devel- oper.chrome.com. Accessed: 2026-04-22

    2026

  11. [11]

    reference.dashif.org

    DASH Industry Forum. reference.dashif.org. 13

  12. [12]

    Fouladi, J

    S. Fouladi, J. Emmons, E. Orbay, C. Wu, R. S. Wahby, K. Winstein. Salsify: Low-Latency network video through tighter integration between a video codec and a transport protocol.15th USENIX Symposium on Networked Systems Design and Implementation (NSDI 18), 267–282. USENIX Association, Renton, W A, 2018. ISBN 978-1-939133-01-4

    2018

  13. [13]

    Ghasemi, Y

    M. Ghasemi, Y . Fu, X. Ouyang, P. Wang, M. K. Turk- can, J. Tavori, S. Kleisarchaki, T. Calmant, L. Gürgen, Z. Kostic, X. S. Di, G. Zussman, J. Ghaderi. Real-time video analytics for urban safety: Deployment over edge and end devices.Proceedings of the Tenth ACM/IEEE Symposium on Edge Computing, SEC ’25. Association for Computing Machinery, New York, NY ,...

    work page doi:10.1145/3769102.3770618 2025

  14. [14]

    I. J. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, Y . Bengio. Gen- erative adversarial nets.Advances in neural information processing systems,27, 2014

    2014

  15. [15]

    docs.cloud.google.com

    Google Compute Engine: Configure NTP on a com- pute instance. docs.cloud.google.com. Google Cloud Documentation

  16. [16]

    developers.google.com

    Google public NTP. developers.google.com. Google for Developers

  17. [17]

    Goyal, A

    P. Goyal, A. Agarwal, R. Netravali, M. Alizadeh, H. Bal- akrishnan. ABC: A simple explicit congestion controller for wireless networks.17th USENIX Symposium on Net- worked Systems Design and Implementation (NSDI 20), 353–372. USENIX Association, Santa Clara, CA, 2020. ISBN 978-1-939133-13-7

    2020

  18. [18]

    Gutman, D

    C. Gutman, D. Waxemberg, A. Neyer, M. Bergeron, A. Hennessy, A. Campbell. Moonlight Mobile Cloud Gaming Client, 2023

    2023

  19. [19]

    S. Ha, I. Rhee, L. Xu. CUBIC: a new TCP-friendly high- speed TCP variant.ACM SIGOPS Operating Systems Review,42(5), 64–74, 2008. ISSN 0163-5980. doi: 10.1145/1400097.1400105

    work page doi:10.1145/1400097.1400105 2008

  20. [20]

    Hemminger

    S. Hemminger. Network emulation with netem

  21. [21]

    G. Judd, P. Steenkiste. Repeatable and realistic wire- less experimentation through physical emulation.SIG- COMM Comput. Commun. Rev.,34(1), 63–68, 2004. ISSN 0146-4833. doi:10.1145/972374.972386

    work page doi:10.1145/972374.972386 2004

  22. [22]

    Karagkioules, G

    T. Karagkioules, G. S. Paschos, N. Liakopoulos, A. Fian- drotti, D. Tsilimantos, M. Cagnazzo. Online learning for adaptive video streaming in mobile networks.ACM Trans. Multimedia Comput. Commun. Appl.,18(1), 2022. ISSN 1551-6857. doi:10.1145/3460819

    work page doi:10.1145/3460819 2022

  23. [23]

    Kushner, P

    H. Kushner, P. Whiting. Convergence of proportional- fair sharing algorithms under general conditions.IEEE Transactions on Wireless Communications,3(4), 1250– 1259, 2004. ISSN 1558-2248. doi:10.1109/TWC.2004. 830826

    work page doi:10.1109/twc.2004 2004

  24. [24]

    X. Li, S. Zhang, Y . Huang, X. Ma, Z. Wang, H. Luo. Towards timely video analytics services at the network edge.IEEE Transactions on Mobile Computing,23(11), 10,443–10,459, 2024. doi:10.1109/TMC.2024.3376769

    work page doi:10.1109/tmc.2024.3376769 2024

  25. [25]

    Y . Li, Z. Li, Z. Gao, T. Chen. Geo2sigmap: High-fidelity rf signal mapping using geographic databases.2024 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN), 277–285. IEEE, 2024

    2024

  26. [26]

    Y . Li, J. Qiu, H. Wang, Z. Li, F. Qian, J. Yang, H. Lin, Y . Liu, B. Xiao, X. Qin, T. Xu. Dissecting and Stream- lining the Interactive Loop of Mobile Cloud Gaming. 595–611, 2025. ISBN 978-1-939133-46-5

    2025

  27. [27]

    Z. Li, Z. Gao, X. Liu, Z. Wang, X. Zhou, L. Liu, Y . Wu, W. Feng, Y . Huang. Large Model Enabled Embod- ied Intelligence for 6G Integrated Perception, Com- munication, and Computation Network, 2025. doi: 10.48550/arXiv.2512.15109. ArXiv:2512.15109 [eess] version: 1

    work page doi:10.48550/arxiv.2512.15109 2025

  28. [28]

    Liang, Y

    H. Liang, Y . Diao, Y . Wu, F. Zhou, Q. Wu. Synergetic Empowerment: Wireless Communications Meets Em- bodied Intelligence.arXiv, (2509.10481), 2025

    work page arXiv 2025

  29. [29]

    M. Lim, M. N. Akcay, A. Bentaleb, A. C. Begen, R. Zim- mermann. When they go high, we go low: low-latency live streaming in dash.js with lol.Proceedings of the 11th ACM Multimedia Systems Conference, MMSys ’20, 321–326. Association for Computing Machinery, New York, NY , USA, 2020. ISBN 9781450368452. doi: 10.1145/3339825.3397043

    work page doi:10.1145/3339825.3397043 2020

  30. [30]

    Y . Liu, T. Zhang, K. Jamieson, Y . Xie. Seeing Through the Fog: Empowering Mobile Devices to Expose and Mitigate RAN Buffer Effects on Delay-Sensitive Proto- cols, 2025

    2025

  31. [31]

    Sunshine: a self-hosted game stream host for Moonlight., 2023

    LizardByte. Sunshine: a self-hosted game stream host for Moonlight., 2023

    2023

  32. [32]

    M. N. Mahfoudi, T. Turletti, T. Parmentelat, W. Dab- bous. Lessons learned while trying to reproduce the openrf experiment.Proceedings of the Reproducibil- ity Workshop, Reproducibility ’17, 21–23. Association for Computing Machinery, New York, NY , USA, 2017. ISBN 9781450350600. doi:10.1145/3097766.3097772

    work page doi:10.1145/3097766.3097772 2017

  33. [33]

    Medbo, K

    J. Medbo, K. Börner, K. Haneda, V . Hovinen, T. Imai, J. Järvelainen, T. Jämsä, A. Karttunen, K. Kusume, J. Kyröläinen, P. Kyösti, J. Meinilä, V . Nurmela, L. Raschkowski, A. Roivainen, J. Ylitalo. Channel modelling for the fifth generation mobile communi- cations.The 8th European Conference on Antennas and Propagation (EuCAP 2014), 219–223, 2014. doi: 10...

    work page doi:10.1109/eucap.2014.6901730 2014

  34. [34]

    A. k. Mishra, S. Ayoubi, G. Grassi, R. Teixeira. Nemfi: record-and-replay to emulate wifi.SIGCOMM Comput. Commun. Rev.,51(3), 2–8, 2021. ISSN 0146-4833. doi: 10.1145/3477482.3477484

    work page doi:10.1145/3477482.3477484 2021

  35. [35]

    Netravali, A

    R. Netravali, A. Sivaraman, S. Das, A. Goyal, K. Win- stein, J. Mickens, H. Balakrishnan. Mahimahi: Accurate Record-and-Replay for {HTTP}.Proc. of the USENIX Annual Technical Conference (ATC), 417–429, 2015. 14

    2015

  36. [36]

    ns-3: a discrete-event network simulator for internet systems

    nsnam. ns-3: a discrete-event network simulator for internet systems. nsnam.org

  37. [37]

    eurecom.fr, 2026

    OAI / openairinterface5G 5G Wireless Implementation. eurecom.fr, 2026. GitLab

    2026

  38. [38]

    Pultarova

    T. Pultarova. 5G Networks Offer Secure GPS Alterna- tives. spectrum.ieee.org. IEEE Spectrum

  39. [39]

    coling-main.89/

    Y . Rubner, C. Tomasi, L. Guibas. A metric for distri- butions with applications to image databases.Sixth International Conference on Computer Vision (IEEE Cat. No.98CH36271), 59–66, 1998. doi:10.1109/ICCV . 1998.710701

    work page doi:10.1109/iccv 1998

  40. [40]

    Sentosa, B

    W. Sentosa, B. Chandrasekaran, P. B. Godfrey, H. Has- sanieh. CellReplay: Towards accurate record-and-replay for cellular networks.USENIX NSDI, 1169–1186, 2025. ISBN 978-1-939133-46-5

    2025

  41. [41]

    Spiteri, R

    K. Spiteri, R. Urgaonkar, R. K. Sitaraman. BOLA: Near-Optimal Bitrate Adaptation for Online Videos. IEEE/ACM Transactions on Networking,28(4), 1698– 1711, 2020. ISSN 1063-6692. doi:10.1109/TNET.2020. 2996964

    work page doi:10.1109/tnet.2020 2020

  42. [42]

    S. R. System. srsRAN Project: Open Source RAN, 2023

    2023

  43. [43]

    Varvello, H

    M. Varvello, H. Chang, Y . Zaki. Performance characteri- zation of videoconferencing in the wild.Proceedings of the 22nd ACM Internet Measurement Conference, IMC ’22, 261–273. Association for Computing Machinery, New York, NY , USA, 2022. ISBN 9781450392594. doi: 10.1145/3517745.3561442

    work page doi:10.1145/3517745.3561442 2022

  44. [44]

    SmoothGrad: Removing noise by adding noise.arXiv preprint arXiv:1706.03825, 2017

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, I. Polosukhin. Atten- tion Is All You Need, 2023. doi:10.48550/arXiv.1706. 03762. ArXiv:1706.03762 [cs]

    work page doi:10.48550/arxiv.1706 2023

  45. [45]

    H. Wan, X. Cao, A. Marder, K. Jamieson. NR-Scope: A Practical 5G Standalone Telemetry Tool.Proceed- ings of the 20th International Conference on emerging Networking EXperiments and Technologies, CoNEXT ’24, 73–80. Association for Computing Machinery, New York, NY , USA, 2024. ISBN 9798400711084. doi: 10.1145/3680121.3697808

    work page doi:10.1145/3680121.3697808 2024

  46. [46]

    H. Wan, K. Jamieson. Evolving Mobile Cloud Gaming with 5G Standalone Network Telemetry, 2024. doi: 10.48550/arXiv.2402.04454. ArXiv:2402.04454 [cs]

    work page doi:10.48550/arxiv.2402.04454 2024

  47. [47]

    Winstein, A

    K. Winstein, A. Sivaraman, H. Balakrishnan. Stochas- tic Forecasts Achieve High Throughput and Low Delay over Cellular Networks.Proc. of the USENIX Sympo- sium on Networked Systems Design and Implementation (NSDI), 2013

    2013

  48. [48]

    H. Xiao, W. Tian, W. Liu, J. Shen. Channelgan: Deep learning-based channel modeling and generating.IEEE Wireless Communications Letters,11(3), 650–654, 2022. doi:10.1109/LWC.2021.3140102

    work page doi:10.1109/lwc.2021.3140102 2022

  49. [49]

    Y . Xie, K. Jamieson. NG-Scope: Fine-Grained Teleme- try for NextG Cellular Networks.Proceedings of the ACM on Measurement and Analysis of Comput- ing Systems,6(1), 1–26, 2022. ISSN 2476-1249. doi: 10.1145/3508032

    work page doi:10.1145/3508032 2022

  50. [50]

    F. Y . Yan, J. Ma, G. D. Hill, D. Raghavan, R. S. Wahby, P. Levis, K. Winstein. Pantheon: the training ground for Internet congestion-control research.USENIX ATC, 2018

    2018

  51. [51]

    W. Ye, X. Hu, S. Sleder, A. Zhang, U. K. Dayalan, A. Hassan, R. A. K. Fezeu, A. Jajoo, M. Lee, E. Ra- madan, F. Qian, Z.-L. Zhang. Dissecting Carrier Ag- gregation in 5G Networks: Measurement, QoE Im- plications and Prediction.Proceedings of the ACM SIGCOMM 2024 Conference, ACM SIGCOMM ’24, 340–357. Association for Computing Machinery, New York, NY , USA,...

    work page doi:10.1145/3651890.3672250 2024

  52. [52]

    F. Yi, H. Wan, K. Jamieson, O. Michel. Auto- mated, Cross-Layer Root Cause Analysis of 5G Video- Conferencing Quality Degradation.Proceedings of the 2025 ACM Internet Measurement Conference, IMC ’25, 835–850. Association for Computing Machinery, New York, NY , USA, 2025. ISBN 979-8-4007-1860-1. doi: 10.1145/3730567.3764434

    work page doi:10.1145/3730567.3764434 2025

  53. [53]

    F. Yi, H. Wan, K. Jamieson, J. Rexford, Y . Xie, O. Michel. Athena: Seeing and Mitigating Wireless Impact on Video Conferencing and Beyond.Proceed- ings of the 23rd ACM Workshop on Hot Topics in Net- works, HotNets ’24, 103–110. Association for Com- puting Machinery, New York, NY , USA, 2024. ISBN 9798400712722. doi:10.1145/3696348.3696889

    work page doi:10.1145/3696348.3696889 2024

  54. [54]

    E. Yu, J. Zhou, Z. Li, G. Tyson, W. Li, X. Zhang, Z. Xu, G. Xie. Mustang: Improving qoe for real-time video in cellular networks by masking jitter.ACM Trans. Mul- timedia Comput. Commun. Appl.,20(9), 2024. ISSN 1551-6857. doi:10.1145/3672399

    work page doi:10.1145/3672399 2024

  55. [55]

    H. Yu, H. Wang, C. Tian, S. D. Sathyanarayana, S. Shi, Z. Xue, S. Yu, H. Li, Y . Peng, X. Pang, R. Zhang. Mag- pie: Improving the Efficiency of A/B Tests for Large Scale Video-on-Demand Systems.Proceedings of the 2024 ACM on Internet Measurement Conference, IMC ’24, 588–594. Association for Computing Machinery, New York, NY , USA, 2024. ISBN 979-8-4007-05...

    work page doi:10.1145/3646547.3689019 2024

  56. [56]

    Y . Zaki, T. Pötsch, J. Chen, L. Subramanian, C. Görg. Adaptive Congestion Control for Unpredictable Cellular Networks.Proc. of the Conference of the ACM Special Interest Group on Data Communication, SIGCOMM, 2015

    2015

  57. [57]

    Zhang, A

    H. Zhang, A. Zhou, R. Ma, J. Lu, H. Ma. Arsenal: Un- derstanding Learning-Based Wireless Video Transport via In-Depth Evaluation.IEEE Transactions on Vehic- ular Technology,70(10), 10,832–10,844, 2021. ISSN 1939-9359. doi:10.1109/TVT.2021.3105479. 15

    work page doi:10.1109/tvt.2021.3105479 2021