Experimental Insights from OpenAirInterface 5G positioning Testbeds: Challenges and solutions

Adeel Malik; Cedric Thienot; Florian Kaltenberger; Mohsen Ahadi; Omid Esrafilian

REVIEW 4 major objections 5 minor 25 references

Experimental Insights from OpenAirInterface 5G positioning Testbeds: Challenges and solutions

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read This paper reports that a fully open-source 5G positioning pipeline, built with OpenAirInterface and uplink TDoA, achieves 1–2 meter accuracy in 90% of cases across three real-world testbeds after adding ToA/TDoA filtering and particle-swar

desk verdict Useful testbed paper with a real dataset, but the abstract's 1-2 m / 90% claim overstates what the TDoA pipeline achieves and the filters look data-fit. read the letter →

arxiv 2508.19736 v1 pith:OIZEU5AD submitted 2025-08-27 cs.NI

Mohsen Ahadi , Adeel Malik , Omid Esrafilian , Florian Kaltenberger , Cedric Thienot This is my paper · ORCID

classification cs.NI

keywords 5GpositioningUL-TDoAOpenAirInterfaceLocationManagementFunctionParticleSwarmOptimizationChannelImpulseResponsefingerprintingO-RAN

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that high-precision 5G positioning is practical with open-source network components, not just in simulation. It reports results from three real testbeds—an outdoor rooftop deployment, an indoor automotive lab, and an aircraft factory hall—using uplink time-difference-of-arrival with a newly integrated Location Management Function in the OpenAirInterface stack. To make TDoA work despite clock drift and multipath, the authors add filters that discard physically impossible measurements and a particle-swarm optimizer that estimates position. They claim 1–2 meter accuracy in 90% of test points under favorable conditions, and show a channel-impulse-response fingerprinting method that outperforms TDoA in the hardest NLoS factory environment. A public dataset of measured CIRs, timestamps, and ground-truth positions accompanies the paper.

What carries the argument

The load-bearing objects are: the strongest-peak ToA read off the channel impulse response (Eq. 1); the per-RU TDoA difference (Eq. 5) that cancels inter-RU clock drift; a TDoA bound from the triangle inequality (Theorem 5.1) used as a physical-feasibility filter; and a particle-swarm optimizer that minimizes the TDoA loss (Eqs. 8–11). For the data-driven branch, the mechanism is a TDoA-aligned, peak-normalized CIR matrix (Eq. 14) with a binary LoS/NLoS mask (Eq. 15) fed into a CNN that regresses 2D position.

What would settle it

Run the same filtered LMF pipeline in a factory hall with known ground truth while deliberately placing a strong metal reflector so that the reflected CIR peak is the strongest; if more than 10% of test points exceed 2 m error, the generalised 1–2 m/90% claim does not hold in such NLoS-dominated settings.

Watch

Extended reading notes

Core claim

The central claim is that a complete positioning pipeline implemented in the OpenAirInterface Location Management Function—empirical ToA filtering, geometry-bounded TDoA filtering, per-RU TDoA references, and a particle-swarm optimizer—makes UL-TDoA reliable enough to reach 1–2 m error in 90% of cases on three distinct live 5G networks. The paper also demonstrates that where TDoA degrades in dense multipath, an O-RAN-inspired fingerprinting framework using TDoA-aligned, peak-normalized CIRs with a learned LoS/NLoS mask achieves sub-meter accuracy, and that this data-driven method degrades gracefully when fewer antennas are available.

Load-bearing premise

The ToA is taken as the delay of the strongest multipath peak, which is a valid timing reference only when the direct path is also the strongest path; in dense NLoS environments that condition fails, as the paper's own Airbus TDoA results show.

Editorial extensions

If this is right

Operators can replicate 1–2 m UL-TDoA positioning using open-source network functions and commercial O-RAN radio units, without proprietary positioning modules.
The per-RU TDoA reference strategy mitigates clock drift in multi-RU deployments, so large outdoor coverage does not require tight cross-RU synchronization.
ToA/TDoA filtering based on empirical delay-spread bounds and the triangle-inequality limit removes physically impossible measurements, improving robustness in dense multipath.
CIR-based fingerprinting with NLoS masking can outperform TDoA accuracy in factory halls and degrades gracefully as antenna count drops.
The public dataset of CIRs with ground-truth labels enables other groups to benchmark and train data-driven positioning methods against a common reference.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The strongest-peak ToA definition ties the pipeline's success to environments where the direct path is also the strongest path; extending the filters to detect first-arrival peaks (e.g., through super-resolution or learned timing features) would likely close the gap in the hardest NLoS factory cases.
The per-RU reference strategy effectively treats each radio unit as a self-synchronized cluster, suggesting a natural extension where inter-RU clock offsets are estimated jointly with the UE position, potentially removing the need for precision time protocol altogether.
The LoS/NLoS masking threshold (gamma = 0.4) is tuned to this dataset; a self-calibrating threshold based on training CIR statistics could make the fingerprinting method transferable across sites without manual adjustment.
The MQTT-based CIR offloading used here is a stand-in for the standard E2 interface; validating the same fingerprinting model over E2 with a near-real-time RIC would test whether the approach holds under standard O-RAN latency and signaling constraints.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Here's my take. The paper gives the 5G positioning community something it actually lacks: a complete open LMF pipeline on OpenAirInterface, run on three real testbeds, with CIR data and ground truth released publicly. The per-RU TDoA reference trick for mitigating inter-RU clock drift is a sensible engineering fix, and the Airbus story is honest — they show TDoA falling apart under NLoS and then demonstrate a CIR fingerprinting CNN with NLoS masking that gets CE90 down to 0.74 m. That's a real result if it holds up.

But the abstract oversells. Table 2 shows the PSO/TDoA pipeline only hits sub-2 m CE90 in two configurations: the 8-antenna GEO-5G static case (1.96 m) and the filtered Stellantis case (1.99 m). Per-RU reference is 2.02 m, handheld 2.36 m, and Airbus TDoA is 8.94 m. The 1–2 m / 90% claim works only if you cherry-pick the fingerprinting number and ignore the TDoA failures. The conclusion says 'under favorable conditions,' which is closer to the truth.

The bigger issue is circularity. The ToA filter in Eq. 3 uses the empirical mean and standard deviation of the max-peak delay distribution from the same dataset it filters; the NLoS masking threshold gamma=0.4 is chosen from a labeled dataset; and there is no train/test split reported for either the filter calibration or the CNN. If those thresholds are fit to the same CIRs used for evaluation, the accuracy numbers, including the 0.74 m, are optimistic. The PSO estimator itself is self-contained, so the geometric TDoA results aren't as tainted, but the preprocessing steps are.

Minor concerns: no error bars, no sample sizes, and no comparison to classical estimators like Chan's method. The strongest-path ToA assumption is a known weak point, and the authors don't dwell on it, though Airbus shows the consequence.

Bottom line: this is a workshop-level experimental paper with a strong dataset component. It deserves a serious referee, not a desk reject, but the referee should push for an abstract rewrite and a proper train/test separation. I'd cite it for the dataset and the LMF implementation, and I'd bring it to a reading group to discuss where data-dependent filtering gets dangerous.

Referee Report

4 major / 5 minor

Summary. The paper reports experimental UL-TDoA positioning results from three 5G testbeds (EURECOM GEO-5G outdoor, STELLANTIS indoor, Airbus factory hall) built on OpenAirInterface with an LMF. It proposes a processing pipeline comprising statistical ToA filtering, geometry-based TDoA filtering, a PSO position estimator, and an optional CIR-based CNN fingerprinting method with NLoS antenna masking. The authors release the datasets publicly. The central claim, stated in the abstract and introduction, is that 1–2 m positioning accuracy is achieved in 90% of cases across the testbeds.

Significance. If the claims hold, this is a valuable experimental contribution: it demonstrates an open-source, 3GPP-compliant LMF positioning pipeline on real O-RAN hardware, validates a geometric TDoA bound, releases a public CIR dataset, and shows that a CNN fingerprinting method can outperform TDoA in a dense factory environment. The PSO estimator is self-contained and independent of the reported error metrics, and the geometric bound in Theorem 5.1 is correct. However, the headline accuracy claim is not supported by Table 2, and the calibration of the ToA filter and NLoS mask uses the same data on which performance is evaluated, which may make the reported numbers optimistic.

major comments (4)

[Abstract, §6, Table 2] The abstract states that results demonstrate '1–2 meter positioning accuracy in 90% of cases in different testbeds', but Table 2's CE90 values for the proposed PSO/TDoA pipeline are: GEO-5G 8 non-collinear antennas 1.96 m, per-RU reference 2.02 m, handheld mobile UE 2.36 m, Stellantis filtered 1.99 m, and Airbus 16-antenna TDoA 8.94 m. Only two of these configurations are below 2 m at CE90; the per-RU and handheld configurations exceed 2 m, and the Airbus TDoA result fails by a large margin. The sub-meter Airbus result (0.74 m) is from the CNN fingerprinting method, not the LMF TDoA pipeline. The conclusion (§7) itself is more modest ('approximately 2m accuracy in 90% of cases under favorable conditions'). The abstract and introduction should be revised to state which method and which configurations meet the 1–2 m target, and the 90%-of-cases claim should be made configuration-specific.
[§5.1, Eq. (3)] The ToA filter uses empirical mean μτ and standard deviation στ of the max-peak delay distribution, but the manuscript does not state whether this distribution is estimated on a dataset disjoint from the CIRs whose ToAs are filtered and then scored in Table 2. If the same data are used to fit the filter bounds and to compute CE90, the reported accuracy is optimistically biased. The authors should either report filter parameters calibrated on an independent training set and evaluate on a held-out test set, or quantify the sensitivity of the results to the choice of the ±1σ threshold. This is load-bearing for all filtered TDoA results, including the claimed 1.99 m at Stellantis.
[§5.3.2, Eq. (15), Table 2] The NLoS masking threshold γ=0.4 is selected 'empirically based on analysis of a labeled dataset' (Fig. 6b) and then applied in both training and inference. No train/validation/test partition is described for this threshold selection, and no training details for the CNN (loss function, optimizer, epochs, split ratio, repeated runs) are provided. The reported Airbus fingerprinting result, 0.74 m CE90, is therefore not established as a genuine test-set result. The authors should report a clear data partition, show that γ is chosen only on the training/validation portion, and give test-set metrics with confidence intervals.
[§3, Eq. (1); §4.3, Table 2] ToA is defined as the delay of the strongest multipath component, i.e., the peak of the CIR magnitude. In NLoS conditions this peak is a reflected path, not the first arrival. The Airbus TDoA result (CE90 = 8.94 m) is consistent with this limitation. The manuscript should explicitly state that the TDoA pipeline is expected to be reliable only when the strongest peak is a valid proxy for the true arrival time, and the '1–2 m' claim should be scoped accordingly. Alternatively, a first-peak detector or an LoS/NLoS classification step before TDoA estimation would be needed to support a broader claim.

minor comments (5)

[§5.3.1, Eqs. (10)-(11)] Eq. (11) uses u_p^(t) on the right-hand side, but the iteration index should be i, not t. Also, the PSO hyperparameters are partially given (w=0.9, c1=0.5, c2=0.9), but the number of particles, number of iterations, and moving-average window size are not reported; these are needed for reproducibility.
[§5.1, Eq. (7)] The notation for the speed of light is inconsistent: Eq. (6) uses v, while Eq. (7) uses c. Unify the notation and define the symbol explicitly.
[§6, Table 2] Table 2 reports MAE and CE90 without sample sizes, confidence intervals, or the number of test points. The reader cannot assess the statistical significance of differences between configurations (e.g., 1.96 m vs 2.02 m). Please add these details.
[§4.1] The conversion from RTK geographic coordinates to the local Cartesian frame is described as 'a linear transformation', but the transformation parameters (translation, rotation, scaling) and how they were estimated are not given. This is important because ground truth accuracy underpins all reported errors.
[§5.3.2] The description of the CNN input states the CIR matrix is in C^{M×C} and uses complex notation, but the layers in Table 1 appear to treat it as real. Clarify whether the absolute value/real part is used as input, and describe the preprocessing consistently.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the positioning pipeline is an empirical system evaluated against external ground truth, not a derivation that reduces to its inputs.

full rationale

The paper presents an experimental positioning system. ToA is defined via Eq. (1) as the strongest-peak delay, TDoA is a difference of ToAs, and the PSO loss (Eq. 8) compares measured TDoAs with geometric predictions from known antenna coordinates. The TDoA filter bound (Eq. 7) is an application of the triangle inequality. The ToA filter (Eq. 3) uses the empirical mean and standard deviation of the observed peak-delay distribution, and the fingerprinting threshold gamma=0.4 is explicitly selected empirically; these are transparent data-dependent calibrations, not hidden fitted parameters renamed as predictions. The reported CE90 values are measured against external laser/RTK ground truth, so the accuracy numbers are independent of the calibration choices in a way that a constructed equivalence would not be. Self-citations to prior OAI NRPPa/LMF work and to PSO comparisons are not load-bearing for the quantitative results. The main weaknesses are the lack of explicit train/test separation for the filters and CNN, which is a methodological concern about optimistic evaluation, but it does not make the central derivation circular by construction.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical entities. Its load-bearing free parameters are the data-dependent filter thresholds (ToA bounds and NLoS masking gamma), which are fit to the same environments where accuracy is reported. The PSO constants and unspecified swarm size are mild additional degrees of freedom.

free parameters (5)

ToA filter bounds mu_tau, sigma_tau = from empirical peak-delay distribution (not explicitly stated)
The filter in Eq. (3) retains CIRs whose peak lies within one standard deviation of the empirical mean of the same dataset, so the outlier threshold is fit to the data being evaluated.
NLoS masking threshold gamma = 0.4
Selected empirically from a labeled LoS/NLoS dataset in the same testbeds (Fig. 6b); used to mask antennas, and applied to data from the same environment.
PSO hyperparameters w, c1, c2 = 0.9, 0.5, 0.9
Hand-chosen constants in Eq. (10); not tuned per testbed, so less concerning, but not justified by a procedure.
Moving average window size = not specified
Averaging window in Section 5.3.1 affects latency and smoothness; no value or tuning rule given.
Particles count and iterations = not specified
PSO stopping criteria and swarm size are not reported, affecting reproducibility of the position estimate.

assumptions (5)

domain assumption UE position is always bounded within the convex frame of M antennas
Used to justify the TDoA magnitude bound in Eq. (7) and to constrain PSO; if the UE leaves the antenna hull, filtering could discard valid measurements.
domain assumption Antennas within each RU are tightly synchronized via PTP with GNSS-disciplined grandmaster, nanosecond-level
The per-RU TDoA approach assumes intra-RU timing coherence; up to 40 ns cross-RU error is acknowledged, so the per-RU reference mitigates it.
domain assumption The CIR is a sum of delayed Dirac deltas (Eq. 2) and the strongest peak corresponds to a usable ToA
Basis for Eq. (1); in dense multipath or NLoS the strongest path is not the first arrival, which the Airbus results confirm.
standard math Triangle inequality in Euclidean geometry (Theorem 5.1)
Standard result used to bound |TDoA|; not controversial.
domain assumption The measured dataset (CIRs, timestamps, labels) is ground truth accurate
RTK and laser triangulation are used for labels; any systematic error in those transfers directly to reported accuracy.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Experimental Insights from OpenAirInterface 5G positioning Testbeds: Challenges and solutions." pith.science (2026). https://pith.science/paper/OIZEU5AD

@misc{pith2026250819736,
  author       = {Pith},
  title        = {Pith review of: Experimental Insights from OpenAirInterface 5G positioning Testbeds: Challenges and solutions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OIZEU5AD}},
  note         = {Machine review of arXiv:2508.19736}
}

read the original abstract

5G New Radio (NR) is a key enabler of accurate positioning in smart cities and smart factories. This paper presents the experimental results from three 5G positioning testbeds running open-source OpenAirInterface (OAI) gNB and Core Network (CN), using Uplink Time Difference of Arrival (UL-TDoA) with the newly integrated Location Management Function (LMF). The testbeds are deployed across both indoor factories and outdoor scenarios with O-RAN Radio Units (RUs), following a 3GPP-compliant system model. The experiments highlight the impact of synchronization impairments, multipath propagation, and deployment geometry on positioning accuracy. To address these challenges, we propose tailored ToA and TDoA filtering as well as a novel position estimation method based on Particle Swarm Optimization (PSO) within the LMF pipeline. Moreover, we show a beyond-5G framework that leverages non-conventional measurements such as Channel Impulse Response (CIR) to train and test Artificial Intelligence and Machine Learning (AI/ML) models for data-driven positioning. The results demonstrate the feasibility of achieving 1-2 meter positioning accuracy in 90% of cases in different testbeds, offering practical insights for the design of robust 5G positioning systems. Moreover, we publicly release the datasets collected in this work to support the research within the 5G positioning community.

Figures

Figures reproduced from arXiv: 2508.19736 by the authors.

**Figure 1.** System model with OAI CN, multiple OAI gNBs, multiple RUs, and distributed antennas for ULSRS TDoA positioning option 7.2) with Central Unit (CU), Distributed Unit (DU), and multiple RUs per gNB. Each RU is equipped with multiple fixed antennas and synchronized via Precision Timing Protocol (PTP) using a Grandmaster clock over the Global Navigation Satellite System (GNSS), achieving nanosecondlevel timing precisi… view at source ↗

**Figure 2.** Overview of 5G positioning testbeds indoor setup includes 2 gNBs, 4 RUs and 8 antennas, covering a 20 m × 20 m area. While fewer nodes would cover the area, 2 gNBs and 4 RUs were deployed to test both single and multi-gNB configurations at scale. Despite LoS connectivity to all antennas, the small environment and indoor propagation pose challenges for positioning. Multipath components exhibit minimal delay separa… view at source ↗

**Figure 3.** Outdoor/indoor multipath CIR presence of large metallic structures and sizable airplane frames under construction further complicates the propagation environment. These obstructions severely limit the availability of LoS conditions, despite the high number of deployed antennas. As a result, the raw received signals and channel estimations are heavily affected by noise and multipath. Ensuring reliable positioning p… view at source ↗

Figures from the paper (5 more)

**Figure 5.** Figure 5: Our proposed positioning pipeline at LMF [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

**Figure 6.** Figure 6: LoS/NLoS power distributions real-world positioning testbeds. In Section 6, we present the positioning results of the proposed processing pipeline across our various testbed environments. 5.3.2 Data-driven method by Finger Printing. While TDoA is widely used for positi…

**Figure 7.** Figure 7: System model for AI/ML positioning 6 RESULTS This section presents experimental results from the three testbeds demonstrating the effectiveness of our proposed positioning pipeline at the LMF using TDoA and Finger Printing methods. Starting with the GEO-5G testbed at…

**Figure 9.** Figure 9: a demonstrates that using per-RU TDoA references improves positioning accuracy under synchronization impairments by mitigating intra-RU timing offset. Fig. 9b compares antenna placement strategies at EURECOM, showing that 2 RUs with 8 non-collinear antennas yield mor…

**Figure 10.** Figure 10: Positioning at (a)Stellantis and (b)Airbus [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 23 canonical work pages

[1]

3GPP. 2020. NG-RAN; NR Positioning Protocol A (NRPPa) . TS 38.455 v16.1.0. NG-RAN; positioning signaling between gNB and LMF

work page 2020
[2]

3GPP. 2022. 5G; System; Location Management Services . TS 29.572 v17.4.0. Defines LMF and its service-based interface (Nlmf_Location)

work page 2022
[3]

Adeel Malik and Mohsen Ahadi and Florian Kaltenberger and Klaus Warnke and Nguyen Tien Thinh and Nada Bouknana and Cedric Thienot and Godswill Onche and Sagar Arora. 2025. From Concept to Reality: 5G Positioning with Open-Source Implementation of UL- TDoA in OpenAirInterface. arXiv:2409.05217 [cs.IT] https://arxiv.org/ abs/2409.05217

arXiv 2025
[4]

Mohsen Ahadi, Adeel Malik, Florian Kaltenberger, and Cedric Thienot

work page
[5]

Ahadi, Mohsen and Kaltenberger, Florian. 2023. 5GNR indoor posi- tioning by joint DL-TDoA and DL-AoD. In2023 IEEE Wireless Commu- nications and Networking Conference (WCNC) . IEEE, 1–6

work page 2023
[6]

and Ho, K.C

Chan, Y.T. and Ho, K.C. 1994. A simple and efficient estimator for hyperbolic location. IEEE Transactions on Signal Processing 42, 8 (1994), 1905–1915. doi:10.1109/78.301830

work page doi:10.1109/78.301830 1994
[7]

and Yildirim, Ali Y

del Peral-Rosado, José A. and Yildirim, Ali Y. and Klinger, Nils C. and Nolle, Patric and Razavi, Sara M. and Parsawar, Sagar and Mundla- muri, Rakesh and Kaltenberger, Florian and Sirola, Niilo and Garlaschi, Stefano and Canzian, Luca and Talvitie, Jukka and Flachs, Detlef. 2023. Preliminary Field Results of a Dedicated 5G Positioning Network for Enhance...

work page doi:10.3390/enc2023-15432 2023
[8]

and Muruganathan, Siva and Masini, Gino and Busin, Åke and Gunnarsson, Fredrik

Dwivedi, Satyam and Shreevastav, Ritesh and Munier, Florent and Nygren, Johannes and Siomina, Iana and Lyazidi, Yazid and Shrestha, Deep and Lindmark, Gustav and Ernström, Per and Stare, Erik and Razavi, Sara M. and Muruganathan, Siva and Masini, Gino and Busin, Åke and Gunnarsson, Fredrik. 2021. Positioning in 5G Networks. IEEE Communications Magazine 59...

doi:10.1109/mcom 2021

Show all 25 references

[9]

Ettus Research. [n. d.]. USRP Software Defined Radio (SDR). https: //www.ettus.com/. Accessed: 2025-07-16

2025
[10]

FIRECELL. [n. d.]. Private 4G/5G Networks. https://www.firecell.io/. Accessed: 2025-07-16

2025
[11]

Michele Gucciardo, Ilenia Tinnirello, Gian Michele Dell’Aera, and Marco Caretti. 2019. A Flexible 4G/5G Control Platform for Fingerprint- based Indoor Localization. In IEEE INFOCOM 2019 - IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS) . 744–749. doi:10.1...

2019
[12]

and Gunnarsson, F

Gustafsson, F. and Gunnarsson, F. 2003. Positioning using time- difference of arrival measurements. In 2003 IEEE International Con- ference on Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP ’03)., Vol. 6. VI–553. doi:10.1109/ICASSP.2003.1201741

2003 arXiv
[13]

Florian Kaltenberger, Tommaso Melodia, Irfan Ghauri, Michele Polese, Raymond Knopp, Tien Thinh Nguyen, Sakthivel Velumani, Davide Villa, Leonardo Bonati, Robert Schmidt, Sagar Arora, Mikel Iraza- bal, and Navid Nikaein. 2024. Driving innovation in 6G wireless technologies: The...

2024 arXiv
[14]

Khamsi and William A

Mohamed A. Khamsi and William A. Kirk. 2001. An Introduction to Metric Spaces and Fixed Point Theory . Wiley-IEEE

2001
[15]

Pahlavan

Xinrong Li and K. Pahlavan. 2004. Super-resolution TOA estimation with diversity for indoor geolocation. IEEE Transactions on Wireless Communications 3, 1 (2004), 224–234. doi:10.1109/TWC.2003.819035

2004
[16]

Li, Dexin and Chu, Xinghe and Wang, Luhan and Lu, Zhaoming and Zhou, Shuya and Wen, Xiangming. 2022. Performance Evaluation of E-CID based Positioning on OAI 5G-NR Testbed. In 2022 IEEE/CIC International Conference on Communications in China (ICCC) . 832–837. doi:10.1109/ICCC5...

2022
[17]

Cameron and Sundaresan, Karthikeyan

Lin, Yu-Tai and Matson, N. Cameron and Sundaresan, Karthikeyan
[18]

Pratap Misra and Per Enge. 2011. Global Positioning System: Signals, Measurements and Performance (2nd ed.). Ganga-Jamuna Press

2011
[19]

Rakesh Mundlamuri, Rajeev Gangula, Florian Kaltenberger, and Ray- mond Knopp. 2025. 5G NR Positioning with OpenAirInterface: Tools and Methodologies. In 2025 20th Wireless On-Demand Network Systems and Services Conference (WONS). 1–7

2025
[20]

Palamà, Ivan and Bartoletti, Stefania and Bianchi, Giuseppe and Blefari Melazzi, Nicola. 2024. Experimental assessment of SDR-based 5G positioning: methodologies and insights.Annals of Telecommunications 79, 5 (2024), 301–313

2024
[21]

Rajeev Gangula and Tommaso Melodia and Rakesh Mundlamuri and Florian Kaltenberger. 2024. Round Trip Time Estimation Utilizing Cyclic Shift of Uplink Reference Signal. arXiv:2410.04528 [cs.IT] https: //arxiv.org/abs/2410.04528

2024 arXiv
[22]

VVDN Technologies. 2023. 5G NR Enterprise Radio Units. https://www.vvdntech.com/en-us/adaptive-compute-and- comms/enterprise-radio-unit. Low- and mid-power O -RUs pre-integrated with CU/DU stack

2023
[23]

Samuele Zanini, Luca Petrucci, Ivan Palamà, Giuseppe Bianchi, and Stefania Bartoletti. 2024. Towards End-to-end Implementation of 5G Positioning with Off-the-shelf Devices. In 2024 IEEE 100th Vehicular Technology Conference (VTC2024-Fall) . 1–6. doi:10.1109/VTC2024- Fall63153....

2024
[2023]

5G NR UL SRS TDoA Positioning by OpenAirInterface
[2024]

Bringing Collaborative Positioning to Native 5G Systems for Enhanced 2D & 3D Location Services. Proc. ACM Netw. 2, CoNEXT4, Article 23 (Nov. 2024), 21 pages. doi:10.1145/3696380

2024 doi

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

3GPP. 2020. NG-RAN; NR Positioning Protocol A (NRPPa) . TS 38.455 v16.1.0. NG-RAN; positioning signaling between gNB and LMF

work page 2020

[2] [2]

3GPP. 2022. 5G; System; Location Management Services . TS 29.572 v17.4.0. Defines LMF and its service-based interface (Nlmf_Location)

work page 2022

[3] [3]

Adeel Malik and Mohsen Ahadi and Florian Kaltenberger and Klaus Warnke and Nguyen Tien Thinh and Nada Bouknana and Cedric Thienot and Godswill Onche and Sagar Arora. 2025. From Concept to Reality: 5G Positioning with Open-Source Implementation of UL- TDoA in OpenAirInterface. arXiv:2409.05217 [cs.IT] https://arxiv.org/ abs/2409.05217

arXiv 2025

[4] [4]

Mohsen Ahadi, Adeel Malik, Florian Kaltenberger, and Cedric Thienot

work page

[5] [5]

Ahadi, Mohsen and Kaltenberger, Florian. 2023. 5GNR indoor posi- tioning by joint DL-TDoA and DL-AoD. In2023 IEEE Wireless Commu- nications and Networking Conference (WCNC) . IEEE, 1–6

work page 2023

[6] [6]

and Ho, K.C

Chan, Y.T. and Ho, K.C. 1994. A simple and efficient estimator for hyperbolic location. IEEE Transactions on Signal Processing 42, 8 (1994), 1905–1915. doi:10.1109/78.301830

work page doi:10.1109/78.301830 1994

[7] [7]

and Yildirim, Ali Y

del Peral-Rosado, José A. and Yildirim, Ali Y. and Klinger, Nils C. and Nolle, Patric and Razavi, Sara M. and Parsawar, Sagar and Mundla- muri, Rakesh and Kaltenberger, Florian and Sirola, Niilo and Garlaschi, Stefano and Canzian, Luca and Talvitie, Jukka and Flachs, Detlef. 2023. Preliminary Field Results of a Dedicated 5G Positioning Network for Enhance...

work page doi:10.3390/enc2023-15432 2023

[8] [8]

and Muruganathan, Siva and Masini, Gino and Busin, Åke and Gunnarsson, Fredrik

Dwivedi, Satyam and Shreevastav, Ritesh and Munier, Florent and Nygren, Johannes and Siomina, Iana and Lyazidi, Yazid and Shrestha, Deep and Lindmark, Gustav and Ernström, Per and Stare, Erik and Razavi, Sara M. and Muruganathan, Siva and Masini, Gino and Busin, Åke and Gunnarsson, Fredrik. 2021. Positioning in 5G Networks. IEEE Communications Magazine 59...

doi:10.1109/mcom 2021