Pith. sign in

REVIEW 4 major objections 5 minor 1 cited by

The paper reports a public dataset that pairs 3D LiDAR room geometry with Wi-Fi RSSI measurements from 20 access-point layouts, giving geometry-aware radio mapping real ground truth.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-04 00:29 UTC pith:ID2IXNMQ

load-bearing objection A real, openly documented indoor radio mapping dataset pairing LiDAR point clouds with RSSI; the manual registration needs validation, but the resource is genuine. the 4 major comments →

arxiv 2511.00494 v3 pith:ID2IXNMQ submitted 2025-11-01 eess.SP cs.AI

An Indoor Radio Mapping Dataset Combining 3D Point Clouds and RSSI

Ljupcho Milosheski , Kuon Akiyama , Bla\v{z} Bertalani\v{c} , Jernej Hribar , Ryoichi Shinkuma This is my paper
classification eess.SP cs.AI
keywords Indoor radio environment mapLiDAR point cloudRSSI measurement datasetWi-Fi signal mappinghuman presence attenuation3D geometry-aware wireless modelingindoor localizationaccess point placement
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper's central claim is that a new public dataset exists, combining high-resolution 3D LiDAR point clouds of a multi-room indoor space with Wi-Fi RSSI measurements collected at 20 different single-access-point setups. The dataset includes two scenarios: an empty office and the same office with 7–10 people present, so models can learn how physical geometry and human presence shape indoor signal coverage. The stated purpose is to give data-driven radio environment map (REM) estimation a realistic alternative to simplified floor plans or purely synthetic ray-tracing data, with the 3D geometry directly aligned to real signal measurements. If the dataset is usable as claimed, it would let machine-learning models train and validate against actual propagation conditions, supporting applications like XR that need reliable high-capacity indoor wireless links.

Core claim

The paper reports the collection and public release of a dataset in which high-resolution 3D point clouds of an indoor environment — an open office and an adjacent corridor/elevator hall — are paired with Wi-Fi RSSI measurements taken at 53 grid locations spaced 2 meters apart. Signal strength was recorded for 20 different single-access-point setups: 12 with the office empty and 8 with 7–10 people present, including four setups that reuse the same AP positions to isolate human-body attenuation. The stated aim is to provide real-world ground truth that lets learning models map physical geometry to radio signal, overcoming the reliance on simplified floor plans or fully synthetic data.

What carries the argument

The load-bearing artifact is the aligned dataset itself: two LiDAR scans (office and corridor) merged by manual point-correspondence registration into one point cloud, plus a coordinate-referenced RSSI grid. The reported 0.2 m measurement-point deviation and the observation that RSSI changes little below about 1 m displacement define the spatial coupling between geometry and signal. The 20 AP placements and the occupied-versus-empty contrast supply the variability needed to train and validate REM models.

Load-bearing premise

The manual registration of the office and corridor point clouds aligns the 3D geometry with the RSSI measurement grid accurately enough that the geometry actually explains the measured signal.

What would settle it

Re-merge the two point clouds with an automatic point-cloud alignment algorithm and measure the discrepancy along the shared office-corridor wall; if the seam shifts by more than about a meter relative to the claimed RSSI grid coordinates, then geometry-aware models trained on this dataset would not be learning the intended geometry-signal coupling.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Models can train and validate on real geometry plus real RSSI rather than synthetic-only data, improving realism in REM estimation.
  • Direct comparisons of the same AP locations with and without people isolate the signal attenuation caused by human bodies in a working office.
  • The dataset supports progressive validation: pre-deployment testing, deployment fine-tuning, and evaluation on unseen AP placements.
  • The 3D point clouds can feed geometry-aware simulation frameworks for hybrid real-plus-synthetic training approaches.
  • Missing RSSI values at distant, wall-separated points provide a natural test case for interpolation and robust prediction methods.

Where Pith is reading between the lines

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

  • If the manual registration between the office and corridor point clouds is off by more than about a meter relative to the RSSI grid, the geometry-signal coupling that makes this dataset valuable would break; an automatic re-registration test could quantify this.
  • Because the measurements use a single 2.4 GHz access point, the dataset's relevance to higher-frequency Wi-Fi (where geometry matters more due to weaker wall penetration) remains an untested extension.
  • The reported 0.2 m grid deviation being negligible suggests the dataset is robust to small positional noise, making it a possible benchmark for studying how sensitive geometry-aware signal models are to alignment error.
  • The four reused AP positions across empty and occupied scenarios offer a controlled paired comparison, but human activity was uncontrolled, so occupancy effects are averages over varied movements rather than precisely quantified body positions.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This data descriptor introduces a publicly archived dataset (Zenodo DOI 10.5281/zenodo.15791300) that pairs high-resolution 3D LiDAR point clouds of an indoor office/corridor environment with Wi-Fi RSSI measurements collected at 53 grid points across 20 single-AP setups. Twelve setups were recorded in an empty office and eight with 7–10 people present. The paper describes the LiDAR acquisition pipeline, the RSSI measurement procedure, the manual registration used to combine the office and corridor point clouds, the file formats (.csv, .h5, .ply), and a toolbox for point-cloud and RSSI processing. The intended contribution is a real-world dataset for training and validating geometry-aware radio environment map (REM) estimation models.

Significance. If the dataset is spatially reliable, it fills a genuine gap: most public indoor REM datasets are either 2D floor-plan based or purely simulated, whereas this one provides dense 3D geometry with co-located RSSI measurements, including an empty/occupied comparison. The paper's delivery strengths are real: the dataset is on Zenodo with a DOI, the code/toolbox is public, 1027 actual RSSI samples are reported, and the two-scenario design supports occupancy-robustness studies. However, the scientific value depends on the accuracy of the geometry–RSSI coupling, which currently rests on an unquantified manual registration and single, unrepeated RSSI samples. These issues are fixable with additional validation and metadata, but they are central enough that the descriptor should not be accepted in its present form.

major comments (4)
  1. [Data Record — 3D data; RSSI Measurement Process] The combined point cloud is produced by manually selecting corresponding points on the common wall between the office and corridor scans, with substantial overlap through glass and with outlier points. No quantitative validation of the registration error is reported, and the stated RSSI measurement-point deviation of 0.2 m is an empirical claim that is not supported by a measurement procedure. Because the dataset's primary use case is learning geometry–RSSI relationships, a systematic misalignment of even 0.5 m can assign an RSSI sample to the wrong wall/obstacle configuration. Please report a registration-error estimate (e.g., distances between corresponding planar surfaces, ICP residual after manual initialization, or survey checkpoints) and a sensitivity analysis showing that downstream REM learning is robust to 0.2–0.5 m misalignment.
  2. [RSSI Measurement Process; Technical Validation] Each grid point appears to be measured once per setup; no repeated-measurement statistics (mean, variance, temporal fading) are provided. The text states that RSSI values do not change for displacements below 0.2 m and that the first observable changes occur at about 1 m, but no data or dedicated experiment in this paper substantiates that claim, and reference [27] does not directly establish it. In the occupied scenario, human activity is uncontrolled, so a single sample cannot separate human-induced fading from other temporal variation. Add repeated measurements at a subset of points, quantify temporal/device variability, and report the distribution of missing values across setups and locations.
  3. [Scenario 2; Usage Notes] The occupied-scenario comparison is load-bearing for the claim that the dataset captures human-presence effects, but the number, positions, and activities of occupants are recorded only as '7–10 individuals engaged in typical activities.' The Usage Notes themselves concede that these factors were not controlled. Please provide per-setup occupancy metadata (count, approximate positions, activity log) or explicitly mark the eight occupied setups as uncontrolled and unsuitable for quantitative human-effect analysis. Without such metadata, the empty/occupied comparison in Figure 7 is anecdotal rather than a validated dataset feature.
  4. [Data Record — HDF5 File Structure] It is unclear whether the `data` array in the .h5 file contains raw RSSI values with placeholders for missing entries or values already filled by interpolation. Since downstream users may train directly on this matrix, the paper must specify the exact semantics of missing entries and state which interpolation method (if any) was applied to the distributed file. The raw .csv should be clearly identified as the authoritative source for unfiltered measurements, and the .h5 processing pipeline should be reproducible.
minor comments (5)
  1. [Table 2] The caption says 'Similar to Table 2' but should refer to Table 1. Also, the coordinate origin for the hallway LiDAR positions is not defined relative to the combined point cloud; please state the reference frame.
  2. [RSSI Measurement Process] The sentence beginning 'as well as distribution of values of WiFi RSSI in the 2.4GHzof around 2dBm' is grammatically unclear and the connection to reference [26] is not obvious. Please rephrase and verify that the reference supports the claimed 2 dBm variability.
  3. [Figure 7] Panels (c) and (d) are described in the caption as observations about specific setups, but the figure itself does not label the setup numbers. Adding 'setup 2' and 'setup 19' directly on the panels would improve readability.
  4. [Table 3] The 'Fast Roaming' column in the sample .csv table shows the value '\n', which appears to be a rendering artifact. Use an actual representative value or indicate that this column is not used in the dataset.
  5. [Background & Summary] In the final sentence of the Background section, 'SEM estimation systems' should be 'REM estimation systems'.

Circularity Check

0 steps flagged

No circularity: the paper reports a measurement dataset with no predictive derivations or parameter fitting.

full rationale

This paper is a data descriptor. Its central claim is that a dataset was collected combining 3D LiDAR point clouds with Wi-Fi RSSI measurements across 20 setups in an indoor environment. There is no derivation chain to inspect: no equations map inputs to predicted outputs, no fitted parameters are renamed as predictions, and no uniqueness theorem or ansatz is imported from prior work. The only self-references are citation [18] (a LiDAR sensor network paper co-authored by two of this paper's authors) and citation [29] (the Zenodo dataset DOI itself). Neither is load-bearing in a circular sense: the sensor network paper is used as background for sensor hardware, and the dataset DOI is the artifact being described, not evidence invoked to prove the dataset's validity. The manual registration of point clouds, while a potential accuracy limitation, is a measurement procedure and not a self-defined result. The technical validation consists of qualitative observations of RSSI plots, not predictions derived from the data. Therefore, the paper is self-contained as a dataset report and exhibits no circular reasoning.

Axiom & Free-Parameter Ledger

0 free parameters · 3 axioms · 0 invented entities

The paper is a data descriptor; it relies on domain assumptions about interference, temporal stationarity, and spatial alignment rather than on fitted parameters or invented entities. These assumptions are load-bearing for the dataset's utility.

axioms (3)
  • domain assumption Interference from co-existing Wi-Fi and Bluetooth networks is negligible in the measurement environment
    Stated in 'Measurement System for the RSSI' with a general citation [25], but no in-situ measurement of interference is provided.
  • domain assumption RSSI values at a grid point are time-invariant; the temporal aspect is not relevant
    Stated in 'RSSI Measurement Process': 'the temporal aspect is not considered relevant.' This underlies the use of single samples per point, including in the human-present scenario where people move.
  • domain assumption The manually registered point cloud aligns with the RSSI measurement grid within 0.2 m
    Registration is manual via corresponding points ('Data Record'), and 'RSSI measurement points deviation is 0.2m.' This alignment is essential for geometry-RSSI correspondence.

pith-pipeline@v1.3.0-alltime-deepseek · 9833 in / 8080 out tokens · 78243 ms · 2026-08-04T00:29:47.548255+00:00 · methodology

0 comments
read the original abstract

The growing number of smart devices supporting bandwidth-intensive and latency-sensitive applications, such as real-time video analytics, smart sensing, Extended Reality (XR), etc., necessitates reliable wireless connectivity in indoor environments. In such environments, accurate design of Radio Environment Maps (REMs) enables adaptive wireless network planning and optimization of Access Point (AP) placement. However, generating realistic REMs remains difficult due to the variability of indoor environments and the limitations of existing modeling approaches, which often rely on simplified layouts or fully synthetic data. These challenges are further amplified by the adoption of next-generation Wi-Fi standards, which operate at higher frequencies and suffer from limited range and wall penetration. To support the efforts in addressing these challenges, we collected a dataset that combines high-resolution 3D LiDAR scans with Wi-Fi RSSI measurements collected across 20 setups in a multi-room indoor environment. The dataset includes two measurement scenarios, the first without human presence in the environment, and the second with human presence, enabling the development and validation of REM estimation models that incorporate physical geometry and environmental dynamics. The described dataset supports research in data-driven wireless modeling and the development of high-capacity indoor communication networks.

Figures

Figures reproduced from arXiv: 2511.00494 by Bla\v{z} Bertalani\v{c}, Jernej Hribar, Kuon Akiyama, Ljupcho Milosheski, Ryoichi Shinkuma.

Figure 1
Figure 1. Figure 1: Alignment of the office room and the hallways, including the RSSI measurement locations. in transitional areas like doorways and corners. This sensor setup was chosen to leverage each device’s strengths: the Avia for range and density, and the VLP-16 for wide angular coverage. In [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Room with position of LiDAR system. 3/12 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Hallway with position of LiDAR system [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 2
Figure 2. Figure 2: Each row in the table corresponds to a single sensor, identified by its name in the “LiDAR” column. The X- and [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: presents the proposed LiDAR data acquisition and processing pipeline, highlighting both the hardware components and their interactions. In this system, each sensor device integrates a LiDAR unit (either Avia or VLP), which generates raw point clouds that are fed into the edge device in the form of NVIDIA Jetson Nano Developer Kit23. Within the edge device, a Grabber module continuously retrieves frames fro… view at source ↗
Figure 5
Figure 5. Figure 5: RSSI acquisition hardware and software. We collect the RSSI measurements using a commercial Android-based smartphone (Samsung model SM-A556B/DS, running Android 14) with the open-source application Wi-Fi Analyzer24. The use of widely available, off-the-shelf hardware for both the AP and the measurement device is a deliberate choice aimed at replicating realistic deployment conditions and capturing signal b… view at source ↗
Figure 6
Figure 6. Figure 6: Indoor environment and measurements [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Observations of the contour plots of RSSI measurements. visualization purposes. Other methods could also work. To enable reproducibility and efficient storage, the processed datasets, including signal maps, setup identifiers, and spatial indices, are compiled into compressed HDF532 files with timestamped filenames. Code availability The code supporting the findings of this study is publicly available on Ze… view at source ↗

discussion (0)

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

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Radio Environment Mapping with World Models for Active Measurement Control: Should Networks Dream of Optimal Control?

    eess.SP 2026-05 unverdicted novelty 7.0

    A world-model-inspired active learning framework for RSSI map reconstruction outperforms Gaussian process interpolation by up to 5x lower RMSE in the few-shot regime on real indoor data.

Reference graph

Works this paper leans on

28 extracted references · 2 canonical work pages · cited by 1 Pith paper

  1. [1]

    Akyildiz, I. F. & Guo, H. Wireless communication research challenges for extended reality (xr).ITU J. on Futur. Evol. Technol.3, 1–15 (2022)

    2022

  2. [2]

    The metaverse and extended reality – implications for wireless communications

    Schwarz, R. . The metaverse and extended reality – implications for wireless communications. https://www. rohde-schwarz.com/us/solutions/wireless-communications-testing/wireless-standards/5g-nr/extended-reality-xr-testing/ white-paper-the-metaverse-and-extended-reality_257899.html (2023). Accessed: 2025-05-14

    2023

  3. [3]

    & Mohorcic, M

    Pesko, M., Javornik, T., Kosir, A., Stular, M. & Mohorcic, M. Radio environment maps: The survey of construction methods.KSII Transactions on Internet Inf. Syst. (TIIS)8, 3789–3809 (2014)

    2014

  4. [4]

    Wireless InSite Propagation Software

    Remcom Inc. Wireless InSite Propagation Software. https://www.remcom.com/wireless-insite-propagation-software (2025). Accessed: June 17, 2025. 5.Hoydis, J.et al.Sionna (2022). Https://nvlabs.github.io/sionna/

    2025

  5. [6]

    I., Leu, J.-S., Su, K.-W., Haniz, A

    Rufaida, S. I., Leu, J.-S., Su, K.-W., Haniz, A. & Takada, J.-I. Construction of an indoor radio environment map using gradient boosting decision tree.Wirel. Networks26, 6215–6236 (2020)

    2020

  6. [7]

    IEEE Transactions on Cogn

    Wang, X.et al.Radiodiff: An effective generative diffusion model for sampling-free dynamic radio map construction. IEEE Transactions on Cogn. Commun. Netw.(2024)

    2024

  7. [8]

    Quan, H.et al.Large language model agents for radio map generation and wireless network planning.IEEE Netw. Lett. (2025)

    2025

  8. [9]

    & Cathabard, S

    Bufort, A., Lebocq, L. & Cathabard, S. Data-driven radio propagation modeling using graph neural networks. TechRxiv, DOI: 10.36227/techrxiv.22820273.v2 (2023)

    work page doi:10.36227/techrxiv.22820273.v2 2023

  9. [10]

    In 2025 IEEE Wireless Communications and Networking Conference (WCNC), 1–6 (IEEE, 2025)

    Li, X.et al.Radiogat: A model-based learning framework for radio map reconstruction via graph attention networks. In 2025 IEEE Wireless Communications and Networking Conference (WCNC), 1–6 (IEEE, 2025)

    2025

  10. [11]

    & Zhou, Q

    Chen, Q., Yang, J., Huang, M. & Zhou, Q. Act-gan: Radio map construction based on generative adversarial networks with act blocks.IET Commun.18, 1541–1550 (2024)

    2024

  11. [12]

    & Demir, B

    Jaensch, F., Caire, G. & Demir, B. Radio map estimation–an open dataset with directive transmitter antennas and initial experiments.arXiv preprint arXiv:2402.00878(2024)

    Pith/arXiv arXiv 2024

  12. [13]

    & Wassell, I

    Bakirtzis, S., Chen, J., Qiu, K., Zhang, J. & Wassell, I. Em deepray: An expedient, generalizable, and realistic data-driven indoor propagation model.IEEE Transactions on Antennas Propag.70, 4140–4154 (2022)

    2022

  13. [14]

    on Multiscale Multiphysics Comput

    Bakirtzis, S.et al.Rigorous indoor wireless communication system simulations with deep learning-based radio propagation models.IEEE J. on Multiscale Multiphysics Comput. Tech.(2024)

    2024

  14. [15]

    T., Guillet, V ., Baala, O., Spies, F

    Cisse, C. T., Guillet, V ., Baala, O., Spies, F. & Caminada, A. Fine tuning an ai-based indoor radio propagation model with crowd-sourced data. In2024 18th European Conference on Antennas and Propagation (EuCAP), 1–5 (IEEE, 2024)

    2024

  15. [16]

    T., Baala, O., Guillet, V ., Spies, F

    Cisse, C. T., Baala, O., Guillet, V ., Spies, F. & Caminada, A. Irgan: cgan-based indoor radio map prediction. In2023 IFIP Networking Conference (IFIP Networking), 1–9 (IEEE, 2023)

    2023

  16. [17]

    & Zhang, J

    Bakirtzis, S., Çagkan Yapar, Qui, K., Wassell, I. & Zhang, J. Indoor radio map dataset, DOI: 10.21227/c0ec-cw74 (2024)

    doi:10.21227/c0ec-cw74 2024

  17. [18]

    & Shiomi, J

    Akiyama, K., Azuma, K., Shinkuma, R. & Shiomi, J. Real-time adaptive data transmission against various traffic load in multi-lidar sensor network for indoor monitoring.IEEE Sensors J.23, 17676–17689 (2023)

    2023

  18. [19]

    & Jakob, W

    Nimier-David, M., Vicini, D., Zeltner, T. & Jakob, W. Mitsuba 2: A retargetable forward and inverse renderer.ACM Transactions on Graph. (ToG)38, 1–17 (2019). 20.Ouster. Vlp-16 lidar. https://ouster.com/products/hardware/vlp-16. 21.Livox. Avia lidar. https://www.livoxtech.com/avia/specs

    2019

  19. [22]

    & Ghabcheloo, R

    Garigipati, B., Strokina, N. & Ghabcheloo, R. Evaluation and comparison of eight popular lidar and visual slam algorithms. In2022 25th International Conference on Information Fusion (FUSION), 1–8 (IEEE, 2022). 23.NVIDIA. Jetson nano (2025). Accessed: 21 October 2025. 11/12

    2022

  20. [24]

    Wifi analyzer

    VREM Software Development. Wifi analyzer. https://github.com/VREMSoftwareDevelopment/WiFiAnalyzer. Accessed: 16 Jun 2025

    2025

  21. [25]

    G., Gazzarrini, L., Giordano, S

    Garroppo, R. G., Gazzarrini, L., Giordano, S. & Tavanti, L. Experimental assessment of the coexistence of wi-fi, zigbee, and bluetooth devices. In2011 IEEE International Symposium on a World of Wireless, Mobile and Multimedia Networks, 1–9 (IEEE, 2011)

    2011

  22. [26]

    & Ledlie, J

    Park, J.-g., Curtis, D., Teller, S. & Ledlie, J. Implications of device diversity for organic localization. In2011 Proceedings IEEE INFOCOM, 3182–3190 (IEEE, 2011)

    2011

  23. [27]

    & Spachos, P

    Sadowski, S. & Spachos, P. Rssi-based indoor localization with the internet of things.IEEE access6, 30149–30161 (2018)

    2018

  24. [28]

    Methods17, 261–272, DOI: 10.1038/s41592-019-0686-2 (2020)

    Virtanen, P.et al.SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python.Nat. Methods17, 261–272, DOI: 10.1038/s41592-019-0686-2 (2020)

    doi:10.1038/s41592-019-0686-2 2020

  25. [29]

    & Shinkuma, R

    Milosheski, L., Akiyama, K., Bertalanic, B., Hribar, J. & Shinkuma, R. An indoor radio mapping dataset combining 3D point clouds and rssi. https://doi.org/10.5281/zenodo.15791300 (2025)

    work page doi:10.5281/zenodo.15791300 2025

  26. [30]

    & Koltun, V

    Zhou, Q.-Y ., Park, J. & Koltun, V . Open3d: A modern library for 3d data processing.arXiv preprint arXiv:1801.09847 (2018)

    Pith/arXiv arXiv 2018

  27. [31]

    Hunter, J. D. Matplotlib: A 2d graphics environment.Comput. Sci. & Eng.9, 90–95, DOI: 10.1109/MCSE.2007.55 (2007)

    doi:10.1109/mcse.2007.55 2007

  28. [32]

    Hierarchical Data Format, version 5

    The HDF Group. Hierarchical Data Format, version 5. https://www.hdfgroup.org/HDF5/ (1997). Accessed: Jun. 16, 2025. 12/12

    1997