pith. machine review for the scientific record. sign in

arxiv: 2605.01777 · v1 · submitted 2026-05-03 · 📡 eess.SP · cs.AI

Recognition: unknown

Data driven approach for Outdoor Channel Prediction in 5G and Beyond

A. Sathi Babu , V. Udaya Sankar , Vishnu Ram OV
Authors on Pith no claims yet

Pith reviewed 2026-05-09 17:02 UTC · model grok-4.3

classification 📡 eess.SP cs.AI
keywords channel predictionmachine learningray tracing5Glinear regressionchannel estimationoutdoor propagation
0
0 comments X

The pith

Linear regression predicts 5G outdoor channel coefficients from transmitter and user locations alone

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

The paper investigates a data-driven method to estimate channel information in wireless networks without relying on periodic pilot signals. It generates synthetic data through ray-tracing simulations at 7 GHz, using transmitter and user positions as input features and the channel coefficient as the target. Three regression models are trained and tested on this data. Linear regression achieves the lowest errors, with mean absolute error of 7.5155 times 10 to the minus 5 and root mean square error of 9.2861 times 10 to the minus 5, outperforming support vector regression and decision tree regression. The approach is presented as a potential digital twin for 5G and beyond networks to reduce estimation overhead.

Core claim

For a given user location, channel estimation at 7GHz frequency band is done using data generated by ray tracing mechanism where feature variables are transmitter location and user location and target variable is channel coefficient. Among Linear Regression, Support Vector Regression and Decision Tree Regression, Linear Regression performs better with MAE of 7.5155×10^{-5} and RMSE of 9.2861×10^{-5}.

What carries the argument

Linear regression model that maps ray-traced transmitter and user location coordinates directly to the complex channel coefficient value.

Load-bearing premise

Ray-tracing simulations at 7 GHz produce channel coefficients that are sufficiently representative of real outdoor propagation, and transmitter and user locations alone contain enough information to predict the channel coefficient accurately.

What would settle it

Compare the model's predicted channel coefficients against actual field measurements collected at 7 GHz in an outdoor environment with the same known transmitter and user locations.

Figures

Figures reproduced from arXiv: 2605.01777 by A. Sathi Babu, Vishnu Ram OV, V. Udaya Sankar.

Figure 1
Figure 1. Figure 1: SRM University AP, India, an outdoor campus scenario,with view at source ↗
Figure 2
Figure 2. Figure 2: , receiver coordinate points from outdoor envi￾ronment is first converted into UTM coordinates. This UTM coordnates is converted into local coordinates using Blender/Sionna as Sionna Ray Tracing (RT) model works only with local coordinates view at source ↗
Figure 3
Figure 3. Figure 3: Machine learning workflow for channel prediction view at source ↗
Figure 4
Figure 4. Figure 4: SRM University-AP, India, with fixed Tx position and 15,000 view at source ↗
Figure 5
Figure 5. Figure 5: Distributions Re{hi}, Im{hi} for indoor and outdoor channel environments Hence, we go for different regression models for predic￾tion of channel coeffcients. The spread of distribution indicates that has reduced multipath richness which is consistent with dominated line-of-sight propagation and a limited number of scattering paths. Regression models are widely used for channel mod￾eling and channel predict… view at source ↗
Figure 7
Figure 7. Figure 7: eCDF performance comparison of theoretical and machine view at source ↗
Figure 6
Figure 6. Figure 6: Comparison of RMSE prediction errors for outdoor channel view at source ↗
read the original abstract

An evolution of Wireless Communications towards 5G and beyond provides improved user experience in terms of quality of services. Understanding and estimating Channel information plays crucial role in providing better user experience. Traditional methods of channel estimation involves periodically sending pilots (known signals), estimating channel and send back estimated channel information to the BS which increases computational complexity and communication complexity. Hence, we focus on data driven approach for channel estimation. This work can be deployed as Digital twin in 5G and beyond wireless networks. In this work, we explore a channel estimation mechanism at 7GHz frequency band for a given user location. This work involves data generation using Ray tracing mechanism and Machine learning model training that contains feature variables such as transmitter location, user location and target variable as channel coefficient . We explored Linear Regression, Support Vector Regression and Decision Tree Regression. We found via simulations that Linear Regression performs (with MAE of $\mathbf{7.5155\times10^{-5}}$ and RMSE of $\mathbf{9.2861\times10^{-5}}$) better than Support Vector Regression and Decision Tree Regression.

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 / 1 minor

Summary. The paper proposes a data-driven channel prediction method for outdoor 5G scenarios at 7 GHz. Ray-tracing simulations generate data with transmitter and user locations as input features and channel coefficients as the target. Linear Regression, Support Vector Regression, and Decision Tree Regression are trained and compared on this data, with the abstract claiming Linear Regression achieves the lowest errors (MAE 7.5155×10^{-5}, RMSE 9.2861×10^{-5}) and suggesting the approach can serve as a digital twin to reduce pilot overhead.

Significance. If the performance claims are substantiated with proper validation protocols, the work could contribute to low-overhead channel estimation in 5G networks by exploiting location-based prediction. The reported error levels on deterministic ray-tracing data highlight that simple regressors may capture propagation effects in controlled simulations, supporting the digital-twin concept in principle. However, the simulation-only scope restricts immediate significance for real deployments.

major comments (3)
  1. [Data generation and model training] The data generation and model training description supplies no information on dataset size, location sampling ranges or distributions, train/test split, hyperparameter selection for SVR and DTR, or cross-validation. These omissions are load-bearing because the central claim that Linear Regression outperforms the others with the stated MAE/RMSE rests on the empirical results; without them the metrics cannot be assessed for overfitting, generalization, or statistical reliability.
  2. [Results and validation] No real-world channel measurements, sensitivity analysis on ray-tracing parameters (e.g., material properties, diffuse scattering), or out-of-distribution location tests are presented. The weakest assumption—that ray-tracing at 7 GHz with location inputs alone produces representative outdoor channels—is therefore untested, directly affecting whether the low errors reflect predictive power or the simplicity of the deterministic forward model.
  3. [Results] The abstract and results report specific error values for the three models but contain no variance estimates, confidence intervals, or formal significance test for the performance gap. This leaves the headline comparison unsupported by visible evidence of robustness.
minor comments (1)
  1. [Abstract] The abstract contains minor grammatical and phrasing issues (e.g., 'provides improved user experience' and 'contains feature variables such as...') that reduce readability.

Simulated Author's Rebuttal

3 responses · 1 unresolved

We thank the referee for the constructive and detailed comments. These have highlighted important aspects of reproducibility, validation scope, and statistical presentation that we will address in the revision. Below we respond point-by-point to the major comments.

read point-by-point responses

  1. Referee: The data generation and model training description supplies no information on dataset size, location sampling ranges or distributions, train/test split, hyperparameter selection for SVR and DTR, or cross-validation. These omissions are load-bearing because the central claim that Linear Regression outperforms the others with the stated MAE/RMSE rests on the empirical results; without them the metrics cannot be assessed for overfitting, generalization, or statistical reliability.

    Authors: We agree that these methodological details are essential for evaluating the reliability of the reported results. The original manuscript omitted them primarily due to brevity. In the revised version we will add a dedicated subsection under 'Data Generation and Model Training' that specifies the total number of generated samples, the spatial ranges and sampling distribution for transmitter and user locations, the train/test split ratio employed, the hyperparameter tuning procedure (including any grid or random search) used for SVR and DTR, and whether cross-validation was applied. These additions will allow readers to assess potential overfitting and generalization. revision: yes

  2. Referee: No real-world channel measurements, sensitivity analysis on ray-tracing parameters (e.g., material properties, diffuse scattering), or out-of-distribution location tests are presented. The weakest assumption—that ray-tracing at 7 GHz with location inputs alone produces representative outdoor channels—is therefore untested, directly affecting whether the low errors reflect predictive power or the simplicity of the deterministic forward model.

    Authors: We acknowledge that the study is confined to deterministic ray-tracing simulations and does not include real-world measurements or sensitivity analyses. This limits direct claims about real deployments. In the revision we will expand the 'Discussion' and 'Limitations' sections to explicitly articulate the assumptions of the ray-tracing model, note that real outdoor channels contain additional stochastic and site-specific effects not captured here, and outline future work involving field measurements and parameter sensitivity studies. However, acquiring and incorporating actual measurement data or performing new sensitivity experiments lies outside the scope of the present revision. revision: partial

  3. Referee: The abstract and results report specific error values for the three models but contain no variance estimates, confidence intervals, or formal significance test for the performance gap. This leaves the headline comparison unsupported by visible evidence of robustness.

    Authors: We agree that reporting variability strengthens the comparison. In the revised manuscript we will re-run the three regression models across multiple independent random seeds (or data shuffles) to obtain mean and standard-deviation values for MAE and RMSE. These statistics, together with the original point estimates, will be presented in the results tables and text. We will also add a short discussion of the observed consistency of the performance ordering across runs. revision: yes

standing simulated objections not resolved
  • Incorporating real-world channel measurements or conducting sensitivity analyses on ray-tracing parameters, as these require new experimental campaigns and data collection that are not available within the current study.

Circularity Check

0 steps flagged

No significant circularity; empirical ML performance metrics on simulated data

full rationale

The paper generates channel coefficient data via ray-tracing at 7 GHz using transmitter and user locations as inputs, then trains and evaluates Linear Regression, SVR, and Decision Tree models to predict those coefficients. The reported MAE (7.5155×10^{-5}) and RMSE (9.2861×10^{-5}) are direct empirical outputs of standard model fitting and error computation on the generated dataset. No derivation step reduces these metrics to inputs by construction, no self-citations or uniqueness theorems are load-bearing, and the chain is self-contained as a simulation-based regression study without circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim depends on the unverified premise that ray-tracing data faithfully represents real channels and that location coordinates are sufficient features.

axioms (1)
  • domain assumption Ray-tracing simulation at 7 GHz produces channel coefficients representative of real outdoor propagation
    All training and test data are generated by ray tracing; no measurement validation is mentioned.

pith-pipeline@v0.9.0 · 5493 in / 1218 out tokens · 56461 ms · 2026-05-09T17:02:33.469640+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

19 extracted references · 2 canonical work pages

  1. [1]

    A survey of artificial in- telligence enabled channel estimation methods: Recent advance, performance, and outlook,

    B. Li, Q. Zheng, X. Tian, M. Yang, G. Gui, W. Jiang, H. Lei, J. Jiang, F. Shu, A. Elhanashiet al., “A survey of artificial in- telligence enabled channel estimation methods: Recent advance, performance, and outlook,”Artificial Intelligence Review, vol. 58, no. 6, p. 187, 2025

    2025

  2. [2]

    6g wireless communications in 7–24 ghz band: Opportunities, techniques, and challenges,

    Z. Cui, P. Zhang, and S. Pollin, “6g wireless communications in 7–24 ghz band: Opportunities, techniques, and challenges,” in2025 IEEE International Symposium on Dynamic Spectrum Access Networks (DySPAN). IEEE, 2025, pp. 1–8

    2025

  3. [3]

    Review on ray tracing channel simulation accuracy in sub-6 ghz outdoor deployment scenarios,

    A. W. Mbugua, Y . Chen, L. Raschkowski, L. Thiele, S. Jaeckel, and W. Fan, “Review on ray tracing channel simulation accuracy in sub-6 ghz outdoor deployment scenarios,”IEEE Open Journal of Antennas and Propagation, vol. 2, pp. 22–37, 2020

    2020

  4. [4]

    Ray tracing for radio propagation modeling: Principles and applications,

    Z. Yun and M. F. Iskander, “Ray tracing for radio propagation modeling: Principles and applications,”IEEE Access, vol. 3, pp. 1089–1100, 2015

    2015

  5. [5]

    Deeprt: A hybrid framework combining large model architectures and ray tracing principles for 6g digital twin channels,

    M. Li, T. Wu, Z. Dong, X. Liu, Y . Lu, S. Zhang, Z. Wu, Y . Zhang, L. Yu, and J. Zhang, “Deeprt: A hybrid framework combining large model architectures and ray tracing principles for 6g digital twin channels,”Electronics, vol. 14, no. 9, p. 1849, 2025

    2025

  6. [6]

    A light-weighted machine learning approach to channel estimation for new-radio systems,

    H. W. Lee and S. W. Choi, “A light-weighted machine learning approach to channel estimation for new-radio systems,”Electron- ics, vol. 12, no. 23, p. 4740, November 2023

    2023

  7. [7]

    DeepMIMO: A Generic Deep Learning Dataset for Millimeter Wave and Massive MIMO Applications

    A. Alkhateeb, “Deepmimo: A generic deep learning dataset for millimeter wave and massive mimo applications,”arXiv preprint arXiv:1902.06435, 2019

    work page Pith review arXiv 1902

  8. [8]

    Ai-assisted agile propagation modeling for real-time digital twin wireless networks,

    A. Saeizadeh, M. Tehrani Moayyed, D. Villa, J. G. Beattie, I. C. Wong, P. Johari, E. W. Anderson, S. Basagni, and T. Melodia, “Ai-assisted agile propagation modeling for real-time digital twin wireless networks,” inProc. IEEE 29th Int. Workshop on Computer Aided Modeling and Design of Communication Links and Networks (CAMAD), 2024, pp. 1–6

    2024

  9. [9]

    Deep learning for physical-layer wireless communications,

    J. Hoydis, S. ten Brink, and M. Debbah, “Deep learning for physical-layer wireless communications,”IEEE Signal Process- ing Magazine, vol. 35, no. 5, pp. 24–45, 2018

    2018

  10. [10]

    Machine learning- based path loss prediction in suburban environment in the sub-6 ghz band,

    F. Tarhouni, M. Al-Zubi, and M.-S. Alouini, “Machine learning- based path loss prediction in suburban environment in the sub-6 ghz band,”IEEE Transactions on Vehicular Technology, 2025, early Access

    2025

  11. [11]

    Path loss prediction in urban environments using ray tracing and machine learning: A comparative evaluation,

    Y . Egi, “Path loss prediction in urban environments using ray tracing and machine learning: A comparative evaluation,” 10 2025

    2025

  12. [12]

    Predicting wireless channel features using neural networks,

    S. Navabi, C. Wang, O. Y . Bursalioglu, and H. Papadopoulos, “Predicting wireless channel features using neural networks,” in 2018 IEEE International Conference on Communications (ICC), 2018, pp. 1–6

    2018

  13. [13]

    A study on path loss modeling using resnet and pre-training with free space path loss,

    T. Nagao and T. Hayashi, “A study on path loss modeling using resnet and pre-training with free space path loss,”2022 IEEE 33rd Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), 2022

    2022

  14. [14]

    Millimeter wave wireless communications (rappaport, t., et al; 2014)[book review],

    S. Misra, “Millimeter wave wireless communications (rappaport, t., et al; 2014)[book review],”IEEE Wireless Communications, vol. 22, no. 5, pp. 6–7, 2015

    2014

  15. [15]

    Creating rf scenarios for large-scale, real-time wire- less channel emulators,

    M. Tehrani-Moayyed, L. Bonati, P. Johari, T. Melodia, and S. Basagni, “Creating rf scenarios for large-scale, real-time wire- less channel emulators,” in2021 19th Mediterranean Commu- nication and Computer Networking Conference (MedComNet), 2021, pp. 1–8

    2021

  16. [16]

    K. P. Murphy,Probabilistic machine learning: an introduction. MIT press, 2022

    2022

  17. [17]

    Sionna RT: Technical report,

    F. A ¨ıt Aoudia, J. Hoydis, M. Nimier-David, B. Nicolet, S. Cam- merer, and A. Keller, “Sionna rt: Technical report,”arXiv preprint, vol. arXiv:2504.21719, 2025

    work page arXiv 2025

  18. [18]

    Study on channel model for frequencies from 0.5 to 100 ghz (release 17),

    3rd Generation Partnership Project (3GPP), “Study on channel model for frequencies from 0.5 to 100 ghz (release 17),” 3GPP Technical Specification Group Radio Access Network (TSG RAN), Tech. Rep. TR 38.901, 2022

    2022

  19. [19]

    A light-weighted machine learning approach to channel estimation for new-radio systems,

    H. W. Lee and S. W. Choi, “A light-weighted machine learning approach to channel estimation for new-radio systems,” Electronics, vol. 12, no. 23, 2023. [Online]. Available: https://www.mdpi.com/2079-9292/12/23/4740

    2023