arxiv: 2605.11037 · v1 · submitted 2026-05-11 · 💻 cs.IT · math.IT

Recognition: 1 theorem link

· Lean Theorem

Annotation-Free Indoor Radio Mapping via Physics-Informed Trajectory Inference

Zheng Xing , Mengru Wu , Yi Zhang , Guanghui Zhang , Jun Gao , Weibing Zhao , Xuhui Zhang , Jinke Ren

show 1 more author

Shuguang Cui

Authors on Pith no claims yet

Pith reviewed 2026-05-13 01:37 UTC · model grok-4.3

classification 💻 cs.IT math.IT

keywords indoor radio mappingchannel state informationtrajectory inferencephysics-informedpower-angle-delay profileannotation-freemultipath propagationBayesian inference

0 comments

The pith

CSI measurements recover user trajectories for indoor radio mapping without location labels or IMU sensors.

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

The paper shows how to construct indoor radio maps from unlabeled Channel State Information (CSI) alone by recovering the latent spatial coordinates of measurements. It does so by using the local spatial continuity of multipath propagation, specifically through a Power-Angle-Delay Profile feature distance that acts as a proxy for small displacements under stable conditions. This continuity is built into a Bayesian model that jointly infers trajectories and propagation parameters, given only known access-point positions and the walkable area. The approach avoids the cost of labeled site surveys and the hardware limits of IMU methods. Tests on real industrial data yield 0.88 m average localization error and 6.68% beam map error, beating representative baselines.

Core claim

The central claim is that the Power-Angle-Delay Profile (PADP) feature distance computed from MIMO-OFDM CSI provides a physically meaningful proxy for small spatial displacements within a local neighborhood under quasi-static multipath conditions. This proxy is incorporated as a continuity constraint in a spatially regularized Bayesian inference model that jointly recovers the latent trajectory of CSI measurements and estimates propagation parameters, enabling fully annotation-free radio mapping when access-point geometry and the walkable domain are known.

What carries the argument

The PADP feature distance serving as a local continuity constraint inside a spatially regularized Bayesian inference model for joint trajectory recovery and propagation-parameter estimation.

If this is right

Radio maps can be calibrated using only CSI data and known environment geometry, eliminating the need for location annotations.
The method achieves 0.88 m average localization error and 6.68% relative beam map construction error on real industrial CSI datasets.
Performance exceeds that of channel-embedding methods and IMU-assisted baselines in the tested settings.
Joint estimation of trajectories and propagation parameters occurs without external motion sensors.

Where Pith is reading between the lines

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

The framework could support periodic re-mapping in slowly changing environments by re-running the inference on new CSI batches.
Combining the inferred trajectories with floor-plan constraints might tighten the spatial regularization term beyond what the paper demonstrates.
If PADP continuity holds across larger scales, the same model could reduce calibration effort for multi-room deployments without additional hardware.
Crowdsourced CSI from multiple devices might be fused under the same Bayesian model to improve trajectory estimates.

Load-bearing premise

Within a local neighborhood and under quasi-static multipath conditions, the PADP feature distance reliably indicates small spatial displacements.

What would settle it

A controlled experiment in which CSI is collected at two points a known small distance apart, the PADP distance between them is computed, and the distance fails to increase monotonically with actual separation under stable multipath conditions.

Figures

Figures reproduced from arXiv: 2605.11037 by Guanghui Zhang, Jinke Ren, Jun Gao, Mengru Wu, Shuguang Cui, Weibing Zhao, Xuhui Zhang, Yi Zhang, Zheng Xing.

**Figure 2.** Figure 2: Visual comparison of trajectory recovery: (a) ground-truth trajectory, [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** Figure 3: Additional analyses of the proposed method: (a) convergence [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

read the original abstract

Constructing indoor radio maps traditionally requires extensive site surveys with precise user-location labels, making the calibration process costly and time-consuming. Existing calibration-reduction methods either depend on partial location annotations or exploit inertial measurement units (IMUs) to provide relative motion cues; however, IMU-assisted solutions are constrained by hardware availability, device-level access restrictions, and accumulated sensor drift. In this paper, we study a location-label-free indoor radio mapping problem under known access-point deployment geometry and a known walkable spatial domain. We propose a physics-informed trajectory inference framework that uses only Channel State Information (CSI), without relying on user-location labels or IMU measurements. The key idea is to recover the latent spatial coordinates of CSI measurements by exploiting the local spatial continuity of multipath propagation. To this end, we construct a Power-Angle-Delay Profile (PADP) feature distance from MIMO-OFDM CSI and show that, within a local neighborhood and under quasi-static multipath conditions, this distance provides a physically meaningful proxy for small spatial displacements. We then incorporate the PADP-based continuity constraint into a spatially regularized Bayesian inference model for joint trajectory recovery and propagation-parameter estimation. Experiments on a real-world industrial CSI dataset demonstrate that the proposed framework achieves an average localization error of 0.88 m and a relative beam map construction error of 6.68%, improving upon representative channel-embedding and IMU-assisted baselines.

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.

Desk Editor's Note private letter to a colleague

The paper gives a label-free CSI radio mapping method by joint Bayesian trajectory and parameter inference under a PADP-derived continuity constraint, with real-data numbers that beat the cited baselines, but the whole thing turns on whether that distance really tracks small displacements reliably.

read the letter

The main takeaway is that they recover both user paths and the radio map parameters from raw MIMO-OFDM CSI alone, using a power-angle-delay profile distance as a spatial regularizer inside the Bayesian model. No labels, no IMU, just the known AP geometry and walkable area. They get 0.88 m localization error and 6.68 % beam-map error on an industrial dataset, which is better than the channel-embedding and IMU baselines they compare against. That combination of the specific PADP proxy with joint inference is new relative to the prior work mentioned in the abstract. The experiments are concrete and on real measurements, which counts for something in this area. The soft spot is exactly the one the stress-test flags: the claim that PADP feature distance acts as a monotonic proxy for small Euclidean moves inside local neighborhoods under quasi-static multipath. The abstract asserts this from physics, but without seeing the full derivation or sensitivity checks for phase wrapping, array calibration drift, or non-stationary scatterers, it is hard to judge how much the reported gains depend on the dataset happening to satisfy the assumption. If those conditions slip, the regularization could pull the inference in the wrong direction. The free parameters listed (spatial weight and propagation params) also suggest some tuning is still needed. This is worth a serious referee for groups working on practical indoor wireless mapping or physics-informed sensing. The problem is real, the numbers are usable, and the framework is grounded enough that revisions could tighten the validation of the core proxy. I would send it out rather than desk-reject.

Referee Report

1 major / 0 minor

Summary. The paper proposes an annotation-free indoor radio mapping method that recovers user trajectories and constructs beam maps from MIMO-OFDM CSI alone, without location labels or IMU data. It constructs a Power-Angle-Delay Profile (PADP) feature distance asserted to serve as a proxy for small spatial displacements under quasi-static multipath conditions, incorporates this into a spatially regularized Bayesian inference model for joint trajectory and propagation-parameter estimation, and reports 0.88 m average localization error and 6.68% relative beam-map construction error on a real industrial CSI dataset, outperforming channel-embedding and IMU-assisted baselines.

Significance. If the PADP proxy holds and the joint inference is correctly grounded, the work would meaningfully advance label-free radio mapping by removing dependence on site surveys or extra sensors, with direct applicability to industrial indoor localization. The concrete real-world error metrics constitute a strength, but the absence of the supporting derivation limits the assessed impact.

major comments (1)

Abstract: the central assertion that the PADP feature distance 'provides a physically meaningful proxy for small spatial displacements' within local neighborhoods under quasi-static multipath conditions lacks any derivation or monotonicity argument; without showing that angle-delay differences remain monotonic with Euclidean position change (accounting for phase wrapping, array calibration, or non-stationary scatterers), the spatially regularized Bayesian inference model rests on an unverified assumption that directly supports the reported 0.88 m and 6.68% figures.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the thoughtful review and for highlighting the need to strengthen the physical justification of our approach. We address the major comment below and will revise the manuscript to incorporate additional supporting material.

read point-by-point responses

Referee: Abstract: the central assertion that the PADP feature distance 'provides a physically meaningful proxy for small spatial displacements' within local neighborhoods under quasi-static multipath conditions lacks any derivation or monotonicity argument; without showing that angle-delay differences remain monotonic with Euclidean position change (accounting for phase wrapping, array calibration, or non-stationary scatterers), the spatially regularized Bayesian inference model rests on an unverified assumption that directly supports the reported 0.88 m and 6.68% figures.

Authors: We acknowledge that the abstract is concise and does not contain a full derivation. The manuscript motivates the PADP distance in Section III-B via the continuous evolution of multipath components under small displacements and quasi-static conditions, supported by empirical observations on the dataset. However, we agree that an explicit monotonicity argument accounting for phase wrapping and other effects would improve rigor. In the revised version we will add a dedicated paragraph (or short subsection) providing a first-order approximation argument based on the far-field model and local linearity of angle-delay shifts, while explicitly noting the approximate nature of the proxy and its limitations under non-stationary scatterers. This clarification will not change the reported results but will better ground the Bayesian model. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation relies on independent physical modeling of multipath continuity

full rationale

The paper derives the PADP feature distance directly from MIMO-OFDM CSI measurements and grounds its use as a local spatial proxy in the physics of quasi-static multipath propagation, which is an external property independent of the target radio map or trajectory output. This constraint is then inserted into a Bayesian model for joint inference of trajectories and propagation parameters. No step reduces the claimed localization or beam-map result to a redefinition or refit of its own inputs; the experimental validation on real CSI data further separates the method from self-referential construction. The central claim therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on the physical assumption that PADP distances proxy spatial displacement and on the provision of known AP geometry and walkable domain; no new particles or forces are postulated.

free parameters (2)

spatial regularization weight
Controls the strength of the trajectory smoothness constraint in the Bayesian model; value not stated in abstract.
propagation parameters
Estimated jointly with trajectories; treated as unknowns rather than fixed inputs.

axioms (2)

domain assumption Local spatial continuity of multipath propagation holds within neighborhoods under quasi-static conditions
Invoked to justify using PADP feature distance as a proxy for spatial displacement.
domain assumption Access-point deployment geometry and walkable spatial domain are known a priori
Stated as given inputs to the inference problem.

pith-pipeline@v0.9.0 · 5569 in / 1505 out tokens · 53647 ms · 2026-05-13T01:37:09.023474+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Proposition 1 (Local PADP Continuity). For two nearby locations xi and xj with di,j = ∥xi − xj∥2 < ρ … c1,q di,j − bq ≤ E[D(Hi,q, Hj,q)|xi, xj] ≤ c2,q di,j + bq.

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

300 extracted references · 300 canonical work pages

[1]

IEEE Veh

Mogensen, Preben and Na, Wei and Kov. IEEE Veh. Technol. Conf. , pages=

work page
[2]

A novel link error prediction model for

Ikuno, Josep Colom and Mehlfuhrer, Christian and Rupp, Markus , booktitle=ICC, pages=. A novel link error prediction model for

work page
[3]

Millimeter wave channel modeling and cellular capacity evaluation , author=

work page
[4]

Outlier-Robust Matrix Completion via l_p -Minimization , author=

work page
[5]

Analysis of blockage effects on urban cellular networks , author=

work page
[6]

Cooperative diversity in wireless networks: Efficient protocols and outage behavior , author=

work page
[7]

High-performance cooperative demodulation with decode-and-forward relays , author=

work page
[8]

Fading relay channels: Performance limits and space-time signal design , author=

work page
[9]

Coverage and rate analysis for millimeter-wave cellular networks , author=

work page
[10]

Grassmannian beamforming for multiple-input multiple-output wireless systems , Volume =

Love, David J and Heath Jr, Robert W and Strohmer, Thomas , Date-Added =. Grassmannian beamforming for multiple-input multiple-output wireless systems , Volume =

work page
[11]

Two-stage beamformer design for massive

Kim, Donggun and Lee, Gilwon and Sung, Youngchul , Journal = IEEE_J_COM, Month = jun, Number =. Two-stage beamformer design for massive

work page
[12]

Hoydis, Jakob and Ten Brink, Stephan and Debbah, M. Massive

work page
[13]

Spatial channel model for

work page
[14]

Quantized multimode precoding in spatially correlated multiantenna channels , author=

work page
[15]

A new design of polar-cap differential codebook for temporally/spatially correlated

Choi, Junil and Clerckx, Bruno and Lee, Namyoon and Kim, Gil , journal=IEEE_J_WCOM, volume=. A new design of polar-cap differential codebook for temporally/spatially correlated

work page
[16]

Kronecker product correlation model and limited feedback codebook design in a

Ying, Dawei and Vook, Frederick W and Thomas, Timothy A and Love, David J and Ghosh, Amitava , booktitle=ICC, pages=. Kronecker product correlation model and limited feedback codebook design in a

work page
[17]

Partial compensation signal-level-based up-link power control to extend terminal battery duration , author=

work page
[18]

Signal-level-based dynamic power control for co-channel interference management , author=. Proc. IEEE Veh. Technol. Conf. , pages=

work page
[19]

Performance of optimum transmitter power control in cellular radio systems , author=

work page
[20]

Achieving

Huh, Hoon and Caire, Giuseppe and Papadopoulos, Haralabos C and Ramprashad, Sean A , journal=IEEE_J_WCOM, volume=. Achieving

work page
[21]

Larsson, Erik G and Edfors, Ove and Tufvesson, Fredrik and Marzetta, Thomas L , journal=IEEE_M_COM, volume=. Massive

work page
[22]

Distributed compressive

Rao, Xiongbin and Lau, Vincent KN , journal=IEEE_J_SP, volume=. Distributed compressive

work page
[23]

Pilot contamination and precoding in multi-cell

Jose, Jubin and Ashikhmin, Alexei and Marzetta, Thomas L and Vishwanath, Sriram , journal=IEEE_J_WCOM, volume=. Pilot contamination and precoding in multi-cell

work page
[24]

Noncoherent trellis coded quantization: A practical limited feedback technique for massive

Choi, Junil and Chance, Zachary and Love, David J and Madhow, Upamanyu , journal=IEEE_J_COM, volume=. Noncoherent trellis coded quantization: A practical limited feedback technique for massive

work page
[25]

Trellis-extended codebooks and successive phase adjustment:

Choi, Junil and Love, David J and Kim, Taeyoung , journal=IEEE_J_WCOM, volume=. Trellis-extended codebooks and successive phase adjustment:

work page
[26]

Downlink training techniques for

Choi, Junil and Love, David J and Bidigare, Patrick , journal=IEEE_J_JSAC, volume=. Downlink training techniques for

work page
[27]

Training sequence design for feedback assisted hybrid beamforming in massive

Noh, Song and Zoltowski, Michael D and Love, David J , journal=IEEE_J_COM, volume=. Training sequence design for feedback assisted hybrid beamforming in massive

work page
[28]

2012 IEEE Wireless Communications and Networking Conference (WCNC) , pages=

Compressive sensing based channel feedback protocols for spatially-correlated massive antenna arrays , author=. 2012 IEEE Wireless Communications and Networking Conference (WCNC) , pages=. 2012 , organization=

work page 2012
[29]

Spatially common sparsity based adaptive channel estimation and feedback for

Gao, Zhen and Dai, Linglong and Wang, Zhaocheng and Chen, Sheng , journal=IEEE_J_SP, volume=. Spatially common sparsity based adaptive channel estimation and feedback for

work page
[30]

J. C. F. Li and M. Lei and F. Gao , booktitle=. Device-to-device (. 2012 , pages=

work page 2012
[31]

Tang and C

H. Tang and C. Zhu and Z. Ding , booktitle=. Cooperative. 2013 , pages=

work page 2013
[32]

Fodor and N

G. Fodor and N. Reider , booktitle=. A Distributed Power Control Scheme for Cellular Network Assisted. 2011 , pages=

work page 2011
[33]

Lin and R

X. Lin and R. W. Heath and J. G. Andrews , journal=IEEE_J_WCOM, title=. 2015 , volume=

work page 2015
[34]

Shalmashi and E

S. Shalmashi and E. Bj\". IEEE Int. Conf. Commun. Workshop , title=. 2015 , pages=

work page 2015
[35]

Zhong and Y

W. Zhong and Y. Fang and S. Jin and K. K. Wong and S. Zhong and Z. Qian , journal=IEEE_J_JSAC, title=. 2015 , volume=

work page 2015
[36]

Jayasinghe and P

K. Jayasinghe and P. Jayasinghe and N. Rajatheva and M. Latva-Aho , journal=IEEE_J_COM, title=. 2014 , volume=

work page 2014
[37]

Ding, Minhua and Blostein, Steven D , journal=IEEE_J_SP, volume=

work page
[38]

Min and W

H. Min and W. Seo and J. Lee and S. Park and D. Hong , journal=IEEE_J_WCOM, title=. 2011 , volume=

work page 2011
[39]

Path loss models for air-to-ground radio channels in urban environments , author=

work page
[40]

Optimal positioning of flying relays for wireless networks , year=

Junting Chen and David Gesbert , journal=. Optimal positioning of flying relays for wireless networks , year=

work page
[41]

Learning air-to-ground channels in dense urban areas for

Junting Chen and Omid Esrafilian and David Gesbert and Urbashi Mitra , journal=. Learning air-to-ground channels in dense urban areas for

work page
[42]

An overview of limited feedback in wireless communication systems , author=

work page
[43]

Rebollo-Monedero, David and Rane, Shantanu and Girod, Bernd , booktitle=Asilomar, volume=. Wyner-

work page
[44]

A cooperative channel estimation approach for coordinated multipoint transmission networks , author=. Proc. IEEE Int. Conf. Commun. Workshop , pages=

work page
[45]

Rate analysis of two-receiver

Hao, Chenxi and Wu, Yueping and Clerckx, Bruno , journal=IEEE_J_COM, volume=. Rate analysis of two-receiver

work page
[46]

A rate splitting strategy for massive

Dai, Mingbo and Clerckx, Bruno and Gesbert, David and Caire, Giuseppe , journal=IEEE_J_WCOM, vol =. A rate splitting strategy for massive

work page
[47]

arXiv preprint arXiv:1601.07254 , year=

Active Target Localization using Low-Rank Matrix Completion and Unimodal Regression , author=. arXiv preprint arXiv:1601.07254 , year=

work page arXiv
[48]

Analysis of target detection via matrix completion , author=

work page
[49]

Power-Distortion metrics for path planning over

Akyol, Emrah and Mitra, Urbashi , journal=IEEE_J_COM, volume=. Power-Distortion metrics for path planning over

work page
[50]

Computer Communications , volume=

Path planning of mobile landmarks for localization in wireless sensor networks , author=. Computer Communications , volume=

work page
[51]

Huang and G

R. Huang and G. V. Zaruba , booktitle=. Static Path Planning for Mobile Beacons to Localize Sensor Networks , year=

work page
[52]

Real-time path planning of mobile anchor node in localization for wireless sensor networks , year=

Hongjun Li and Jianwen Wang and Xun Li and Hongxu Ma , booktitle=. Real-time path planning of mobile anchor node in localization for wireless sensor networks , year=

work page
[53]

C. H. Ou and W. L. He , journal=. Path Planning Algorithm for Mobile Anchor-Based Localization in Wireless Sensor Networks , year=

work page
[54]

Rezazadeh and M

J. Rezazadeh and M. Moradi and A. S. Ismail and E. Dutkiewicz , journal=. Superior Path Planning Mechanism for Mobile Beacon-Assisted Localization in Wireless Sensor Networks , year=

work page
[55]

Yardibi and J

T. Yardibi and J. Li and P. Stoica and M. Xue and A. B. Baggeroer , journal=. Source Localization and Sensing: A Nonparametric Iterative Adaptive Approach Based on Weighted Least Squares , year=

work page
[56]

Distributed, robust acoustic source localization in a wireless sensor network , author=

work page
[57]

Applied Acoustics , volume=

Direct regressions for underwater acoustic source localization in fluctuating oceans , author=. Applied Acoustics , volume=

work page
[58]

The Journal of the Acoustical Society of America , volume=

Matched field processing: Source localization in correlated noise as an optimum parameter estimation problem , author=. The Journal of the Acoustical Society of America , volume=

work page
[59]

The Journal of the Acoustical Society of America , volume=

An artificial neural network for sound localization using binaural cues , author=. The Journal of the Acoustical Society of America , volume=

work page
[60]

Deleforge and R

A. Deleforge and R. Horaud and Y. Y. Schechner and L. Girin , journal=. Co-Localization of Audio Sources in Images Using Binaural Features and Locally-Linear Regression , year=

work page
[61]

ACM Trans

A kernel-based learning approach to ad hoc sensor network localization , author=. ACM Trans. on Sensor Networks , volume=

work page
[62]

Indoor localization with channel impulse response based fingerprint and nonparametric regression , author=

work page
[63]

IEEE Trans

Target localization using ensemble support vector regression in wireless sensor networks , author=. IEEE Trans. on Cybernetics , volume=

work page
[64]

Exact and approximate solutions of source localization problems , author=

work page
[65]

A lagrangian dual approach to the single-source localization problem , author=

work page
[66]

Maximum likelihood multiple-source localization using acoustic energy measurements with wireless sensor networks , author=

work page
[67]

On energy-based acoustic source localization for sensor networks , author=

work page
[68]

Focalization: Environmental focusing and source localization , author=. J. of the Acoustical Society of America , volume=

work page
[69]

Underwater acoustic communication channels: Propagation models and statistical characterization , author=

work page
[70]

Underwater acoustic networks , author=. IEEE J. of Oceanic Engineering , volume=

work page
[71]

Journal of Chemometrics , volume=

Least squares algorithms under unimodality and non-negativity constraints , author=. Journal of Chemometrics , volume=

work page
[72]

arXiv preprint arXiv:1512.02956 , year=

Adaptive Risk Bounds in Unimodal Regression , author=. arXiv preprint arXiv:1512.02956 , year=

work page arXiv
[73]

Broadband underwater localization of multiple sources using basis pursuit de-noising , author=

work page
[74]

Maximum likelihood localization of multiple sources by alternating projection , author=

work page
[75]

2016 , pages =

Laura Cottatellucci , booktitle =. 2016 , pages =

work page 2016
[76]

Exact matrix completion via convex optimization , author=. Commun. of the ACM , volume=

work page
[77]

The power of convex relaxation: Near-optimal matrix completion , author=

work page
[78]

Proceedings of the IEEE , volume=

Matrix completion with noise , author=. Proceedings of the IEEE , volume=

work page
[79]

1995 , publisher=

Introduction to spread spectrum systems , author=. 1995 , publisher=

work page 1995
[80]

arXiv preprint, arXiv:1701.08174 , year =

Rotated Eigenstructure Analysis for Source Localization without Energy-decay Models (full version) , author =. arXiv preprint, arXiv:1701.08174 , year =

work page arXiv

Showing first 80 references.