Pith. sign in

REVIEW 3 major objections 4 minor 23 references

An Experimental Study of Passive UAV Tracking with Digital Arrays and Cellular Downlink Signals

T0 review · 3 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read A passive LTE-downlink sensing system can track a UAV's trajectory to 1.49 m RMSE.

desk verdict A credible single-flight proof of concept for passive LTE-based UAV tracking, but the 2D measurement model and steel-sphere target make the meter-level claim conditional on unreported geometry. read the letter →

arxiv 2412.20788 v1 pith:5JN7AYIP submitted 2024-12-30 eess.SP cs.SYeess.SY

Yifei Sun , Chao Yu , Yan Luo , Tony Xiao Han , Haisheng Tan , Rui Wang , Francis C. M. Lau This is my paper · ORCID
classification eess.SPcs.SYeess.SY
keywords passivesensingbistaticradarUAVtrackingLTEdownlinkdigitalbeamformingmulti-targetcrossambiguityfunctionlow-altitudeeconomy
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 reports a fully passive system, LIPASE, that tracks the trajectory of a flying UAV using only the downlink signal of an ordinary LTE base station as an illuminator of opportunity. The receiver carries two small digital uniform linear arrays and extracts bistatic range, Doppler, and angle of arrival from the cross ambiguity function of the direct and target-scattered signals. A multi-target tracking framework then turns those detections into a continuous trajectory. In a field experiment with a 255 m baseline, the Cartesian tracking variant reached 1.49 m root-mean-square localization error, which the authors claim is the first meter-level result for bistatic UAV tracking.

What carries the argument

The load-bearing mechanism is the cross ambiguity function computed between the beamformed LTE reference signal and clutter-cancelled surveillance beams: it produces per-slot detections of bistatic range, bistatic Doppler, and, through phase interferometry across the eight-element array, angle of arrival. These noisy and gappy observations are fed to a multi-target tracking framework with two state definitions. Bistatic tracking keeps the raw bistatic range, range rate, and AoA as state; Cartesian tracking instead models position and velocity directly, with a nonlinear measurement model, and achieves the better result. The experimental parameters are a 2132.5 MHz carrier, 5 MHz bandwidth, 0.2 s coherent integration time, and a 255 m transmitter-receiver baseline.

What would settle it

Repeat the same flight with the steel sphere removed and an ordinary quadcopter; observe whether the CFAR detection rate and Cartesian-tracking RMSE stay near 71.9% and 1.49 m, or degrade sharply.

Watch

Extended reading notes

Core claim

The central claim is that bistatic passive sensing with LTE downlink signals provides sufficient range, Doppler, and angular resolution for meter-level UAV trajectory tracking, provided the receiver uses a digital antenna array and a tracking layer that tolerates missed detections and false alarms. The evidence is a field experiment in which LIPASE, with a four-element reference array and an eight-element surveillance array, followed a quadcopter flying a J-shaped path at a range of about 100 m from the receiver. Despite a coarse 30 m range resolution and a detection rate of only 71.9%, the multi-target tracking framework with a Cartesian state model produced a localization RMSE of 1.49 m, with 1.33 m MAE. This is stated as the first experimental demonstration of bistatic UAV trajectory tracking at meter-level accuracy.

Load-bearing premise

The drone was fitted with a steel sphere to imitate a delivery drone, so the experiment's detection rate and meter-level accuracy rely on a stronger radar echo than a typical small consumer drone would return.

Editorial extensions

If this is right

  • If correct, low-altitude surveillance can reuse existing LTE base stations as transmitters without any upgrade or cooperation on their side.
  • A single bistatic receiver pair is enough for meter-level tracking, so multi-static deployments or out-of-band sensors are not required for tracking accuracy.
  • The MTT framework makes the system robust to realistic detector imperfections: the reported 71.9% detection rate with 28.1% missed detections and 18.2% false alarms still yields sub-meter-level-per-axis errors in Cartesian tracking.
  • The same receiver architecture should transfer to other cellular downlink waveforms with similar bandwidth, potentially improving resolution as bandwidth grows.

Reading between the lines

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

  • We infer the meter-level result is specific to the enhanced radar cross section of the tested payload; a bare consumer drone would likely need multi-static illumination or longer coherent integration to hold this accuracy.
  • We infer that the accuracy comes largely from temporal smoothing: the raw range resolution is 30 m, yet tracking error is 1.49 m, suggesting the trajectory is recovered by integrating many low-resolution detections along a kinematic model rather than by a single high-resolution snapshot.
  • We infer that extending the system to 5G NR downlink with wider bandwidth could push the same architecture below meter-level error, and that deploying several such receivers could localize drones in the full volume rather than on a bistatic contour.
  • We infer the bistatic-tracking-versus-Cartesian comparison indicates the physical model of the state transition, not the measurement resolution, is the dominant factor in final accuracy; any extension should invest in accurate target dynamics.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. This paper reports LIPASE, a passive bistatic sensing system that uses LTE downlink signals from a commercial eNB to track a UAV. The receiver uses a 4-element reference ULA and an 8-element surveillance ULA, and the signal-processing chain includes digital beamforming, least-squares clutter cancellation, cross-ambiguity range-Doppler processing, CA-CFAR detection, and AoA estimation, followed by a multi-target tracking (MTT) framework with two extended/linear Kalman filter variants whose states are defined in bistatic coordinates and in Cartesian coordinates. In one 12 s experiment with a quadcopter carrying a steel sphere and a differential-GPS reference, the Cartesian tracker is reported to achieve a 1.49 m RMSE localization error in the horizontal plane, with a 71.9% detection rate and an 18.2% false-alarm rate. The paper claims that this is the first experimental demonstration of bistatic UAV tracking with meter-level accuracy.

Significance. If the claims are accepted, the result is significant: it would show that a single LTE base station used as an illuminator, together with a compact digital array at the receiver, can provide sufficient range and angle information for meter-level trajectory tracking without out-of-band sensors. The experiment is carefully instrumented with differential GPS ground truth, and the paper honestly reports detection, false-alarm, and missed-detection statistics as well as MAE/RMSE for both tracking methods. The limitations below, however, concern whether the reported 1.49 m RMSE can be interpreted as a three-dimensional UAV-localization result and whether it extends beyond the single flight and the artificially enhanced target used in the experiment.

major comments (3)
  1. [Section IV, Eq. (35) and Section V-C] The Cartesian measurement model is strictly two-dimensional: the state in Eq. (31) has no altitude component, the position is written as l_m=[x_m,y_m]^T, and the AoA is modeled as azimuth arctan2(y_m-y_r, x_m-x_r). The manuscript never reports the drone altitude, the array height, or the eNB elevation. Because a ULA measures only a cone angle, the estimated angle is arcsin(cos ε sin ψ) rather than the azimuth ψ when the target is at elevation ε, and the true bistatic range contains a slant-range height term. At roughly 100 m horizontal range, a 10 m height difference produces about 0.5 m of range bias and a 30 m height difference produces about 4.4 m, the latter being comparable to the claimed localization RMSE. The EKF could absorb a slowly varying bias, but then the reported accuracy is specific to an unreported geometry. The headline meter-level claim is therefore not yet established for typical 3D UAV flights.
  2. [Section V-A and Table II] The reported result rests on a single 12-second 'J' flight with no repeated trials, no confidence intervals, and no cross-validation. The MTT parameters in Table II—including the measurement noise covariance σ²_R, σ²_θ, the process noise covariances, the gating threshold γ, and the CFAR threshold α—are presented as fixed values without a sensitivity analysis or a statement of how they were chosen. Since the 1.49 m RMSE is one output of one hand-tuned pipeline on one trajectory, the repeatability and robustness of the central claim are not established.
  3. [Section V-A and Section V-C] The target is described as 'a quadcopter hung with a steel sphere to imitate a delivery drone with a payload.' A steel sphere is a high-radar-cross-section reflector, and the experiment already shows a 28.1% missed-detection rate, including a gap when the trajectory follows the zero-Doppler bistatic contour. Without the sphere, the scattered signal from a typical consumer drone would be much weaker, so the reported detection statistics and tracking accuracy may not transfer to the UAVs mentioned in the abstract and introduction. The paper should either provide a control experiment without the sphere, quantify the sphere's RCS contribution, or explicitly restrict the claim to delivery drones with large metallic payloads.
minor comments (4)
  1. [Section III-A, Eq. (9)] The surveillance steering vector in Eq. (9) uses N_ref−1 in the phase exponent, but it should use N_sur−1 to match the 8-element surveillance array described in Section V-A.
  2. [Section V-C] The sentence 'the corresponding MAEs and RMSEs are compared in Table 9' appears to refer to the data displayed in Fig. 9; the cross-reference should be corrected.
  3. [Throughout] The manuscript repeatedly types 'UA V' with a spurious space in the abstract, headers, and body text; this should be normalized to 'UAV'.
  4. [Section V-A] The phrase 'the intrinsic global positioning system (GPS) system' contains a redundant 'system' and should be rewritten.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the meter-level accuracy is a measured comparison against DGPS ground truth, not a fitted or self-cited output.

full rationale

The paper's central claim is that LIPASE achieves meter-level localization accuracy, with a best Cartesian-tracking RMSE of 1.49 m (Fig. 9). This number is an experimental result computed by Eq. (75) between the estimated trajectory and the DGPS ground truth, which the paper describes as providing 'sub-meter localization accuracy' (Section V-A). The tracking filters use measurement and process noise covariances listed as fixed constants in Table II, and the paper does not report fitting these values against the ground-truth trajectory; no equation in the derivation reduces to a fitted output. The measurement model in Eq. (35) is a geometric mapping from target state to bistatic range, range rate, and AoA, and the track initialization in Eqs. (53) and the Kalman correction in Eqs. (64)-(67) are standard estimation operations rather than definitions of the claimed accuracy. The only self-citation is Ref. [19], used for 'detection clustering' in the CFAR processing step (Section III-C); this is a routine grouping procedure that is not load-bearing for the central accuracy claim, and no uniqueness or modeling assumption is imported from it. Concerns about the artificially inflated radar cross section of the steel-sphere payload and the two-dimensional measurement model that ignores UAV altitude are legitimate threats to external validity, but they do not make the reported measurement circular. Accordingly, there is no identified circular step; the score reflects only the minor, non-load-bearing self-citation.

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

The accuracy result is a measured output, so the ledger mostly records the hand-tuned detector and tracker parameters that affect it, plus the scenario assumptions of static clutter, a point-scatterer target, and a high-RCS reflector. No new physical entities are postulated.

free parameters (5)
  • Measurement noise covariance R = diag(7^2 m^2, 1^2 m^2/s^2, 3^2 deg^2)
    Table II. Hand-set for the Kalman filter; directly controls the smoothing versus accuracy trade-off.
  • Process noise covariance for Cartesian tracking = σ²_̈x = σ²_̈y = 4^2 m^2/s^4
    Table II. Hand-set; if too small the filter trusts the constant-velocity model too much, if too large it follows noisy measurements.
  • Process noise covariance for bistatic tracking = σ²_¨R = 10^2, σ²_¨θ = 3^2 (units as in Table II)
    Table II. Hand-set for the bistatic Kalman filter.
  • CA-CFAR threshold factor α = 15 dB
    Table II. Detection threshold; chosen to trade detection rate (71.9%) against false alarms (18.2%).
  • Gating threshold γ and track management windows = γ=20, N_cnf=5, N_del=14
    Table II. Hand-set MTT parameters; determine which tentative tracks become confirmed and how quickly tracks are deleted.
assumptions (5)
  • domain assumption The LoS reference signal is an undistorted copy of the eNB transmission x(t).
    Invoked in Eq. (1) and used throughout; if multipath or fading corrupts the reference, the cross-ambiguity and clutter cancellation degrade.
  • domain assumption All undesired scatterers are static, so clutter has zero/low Doppler and lies within the cancellation band (P=0).
    Section III-B; the LS clutter cancellation projects out the clutter subspace built from the reference signal. Moving scatterers would not be fully cancelled and could create false alarms.
  • domain assumption The UAV behaves as a point scatterer with a constant-velocity motion model over each 0.2 s coherent integration time; accelerations are absorbed into process noise.
    Section IV-C, Eqs. (57)-(62); the constant-velocity model with tuned Q is the basis for both tracking methods.
  • standard math Measurement noises are zero-mean Gaussian with known covariance R, independent across range, range-rate, and AoA.
    Eq. (32); needed for the Kalman update and Mahalanobis gating.
  • ad hoc to paper The surveillance array's field of view covers the target throughout the flight.
    Section II; patch antennas have limited beamwidth, and the experiment's geometry was chosen to keep the target in view. The paper does not quantify the FoV.

how reviews work

0 comments
Cite this review

Pith. "Pith review of An Experimental Study of Passive UAV Tracking with Digital Arrays and Cellular Downlink Signals." pith.science (2026). https://pith.science/paper/5JN7AYIP

@misc{pith2026241220788,
  author       = {Pith},
  title        = {Pith review of: An Experimental Study of Passive UAV Tracking with Digital Arrays and Cellular Downlink Signals},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5JN7AYIP}},
  note         = {Machine review of arXiv:2412.20788}
}
read the original abstract

Given the prospects of the low-altitude economy (LAE) and the popularity of unmanned aerial vehicles (UAVs), there are increasing demands on monitoring flying objects at low altitude in wide urban areas. In this work, the widely deployed long-term evolution (LTE) base station (BS) is exploited to illuminate UAVs in bistatic trajectory tracking. Specifically, a passive sensing receiver with two digital antenna arrays is proposed and developed to capture both the line-of-sight (LoS) signal and the scattered signal off a target UAV. From their cross ambiguity function, the bistatic range, Doppler shift and angle-of-arrival (AoA) of the target UAV can be detected in a sequence of time slots. In order to address missed detections and false alarms of passive sensing, a multi-target tracking framework is adopted to track the trajectory of the target UAV. It is demonstrated by experiments that the proposed UAV tracking system can achieve a meter-level accuracy.

Figures

Figures reproduced from arXiv: 2412.20788 by the authors.

Figure 1
Figure 1. An example scenario of LIPASE. Particularly, let y ref i (t) be the received signal at the i￾th antenna of the reference array, i = 1, 2, . . . , Nref, the aggregation of the received signal at the reference array, namely the reference signal vector, can be written as y ref(t) ≜ [y ref 1 (t), yref 2 (t), . . . , yref Nref (t)]T. It consists of the desired signal from the reference channel, Lref undesired scattered s… view at source ↗
Figure 2
Figure 2. The proposed signal processing scheme of LIPASE. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Geometry of passive sensing. In the following, we first introduce the procedure of a multi-target tracking algorithm with general operators of track initialization, prediction, and update. Then, these operators are specified for both definitions of the UAV state. A. The Multi-Target Tracking Framework The MTT algorithm initializes and maintains a set of tracks, where each track consists of a sequence of estimated UA… view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The experimental platform of LIPASE. positioning system (GPS) system, which is not sufficiently accurate as the experiment ground truth, the drone is equipped with an additional differential GPS (DGPS) module. The DGPS reference station is deployed on the ground, such …
Figure 5
Figure 5. Figure 5: Overview of experiment scenario. B. Range, Velocity and Angular Resolutions The range resolution ∆R is the minimum required range to distinguish two different targets, which is defined by [22] ∆R = c 2B cos(β bist/2), (70) where B is the bandwidth and β bist is the bis…
Figure 6
Figure 6. Figure 6: Observations and tracking results versus time in terms of (a) The bistatic range, (b) bistatic Doppler shift, and (c) AoA. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: A sample of the RD response (a) before and (b) after CFAR [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]
Figure 8
Figure 8. Figure 8: Observations and tracking results versus time in terms of (a) x-axis position, (b) y-axis position, (c) x-axis velocity and (d) y-axis [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]
Figure 10
Figure 10. Figure 10: The estimated trajectories of the drone. [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 20 canonical work pages

  1. [1]

    Integrated sensing and communications: Toward dual-functional wire- less networks for 6g and beyond,

    F. Liu, Y . Cui, C. Masouros, J. Xu, T. X. Han, Y . C. Eldar, and S. Buzzi, “Integrated sensing and communications: Toward dual-functional wire- less networks for 6g and beyond,” IEEE Journal on Selected Areas in Communications, vol. 40, no. 6, pp. 1728–1767, 2022

    2022

  2. [2]

    Networked ISAC for Low-Altitude Economy: Transmit Beamforming and UAV Trajectory Design

    G. Cheng, X. Song, Z. Lyu, and J. Xu, “Networked isac for low-altitude economy: Transmit beamforming and uav trajectory design,” 2024. [Online]. Available: https://arxiv.org/abs/2405.07568

    work page Pith review arXiv 2024

  3. [3]

    A Survey on Detection, Classification, and Tracking of UAVs using Radar and Communications Systems

    W. Khawaja, M. Ezuma, V . Semkin, F. Erden, O. Ozdemir, and I. Guvenc, “A survey on detection, classification, and tracking of aerial threats using radar and communications systems,” 2024. [Online]. Available: https://arxiv.org/abs/2402.05909

    work page Pith review arXiv 2024

  4. [4]

    Design, development and experimental validation of multi -target tracking framework for passive radar,

    R. Umar, M. Abdulhadi, M. Shoaib, M. Bilal, A. Aldowesh, and K. Jamil, “Design, development and experimental validation of multi -target tracking framework for passive radar,” in 2019 IEEE Radar Conference (RadarConf), 2019, pp. 1–6

    work page 2019

  5. [5]

    Fm radio based bistatic radar,

    P. Howland, D. Maksimiuk, and G. Reitsma, “Fm radio based bistatic radar,” IEE Proceedings - Radar, Sonar and Navigation , vol. 152, pp. 107–115(8), June 2005. [Online]. Available: https: //digital-library.theiet.org/content/journals/10.1049/ip-rsn 20045077

    work page doi:10.1049/ip-rsn 2005

  6. [6]

    Micro- uav detection using dab-based passive radar,

    C. Sch ¨upbach, C. Patry, F. Maasdorp, U. B¨oniger, and P. Wellig, “Micro- uav detection using dab-based passive radar,” in 2017 IEEE Radar Conference (RadarConf), 2017, pp. 1037–1040

    work page 2017

  7. [7]

    A practical bistatic passive radar system for use with dab and drm illuminators,

    C. Coleman, R. Watson, and H. Yardley, “A practical bistatic passive radar system for use with dab and drm illuminators,” in 2008 IEEE Radar Conference, 2008, pp. 1–6

    work page 2008

  8. [8]

    On the potentials of passive, multistatic, low frequency radars to counter stealth and detect low flying targets,

    H. Kuschel, J. Heckenbach, S. Muller, and R. Appel, “On the potentials of passive, multistatic, low frequency radars to counter stealth and detect low flying targets,” in 2008 IEEE Radar Conference , 2008, pp. 1–6

    work page 2008

Show all 23 references
  1. [9]

    Target tracking in time-division- multifrequency-based passive radar,

    M. L ¨u, J. Yi, X. Wan, and K. Shu, “Target tracking in time-division- multifrequency-based passive radar,” IEEE Sensors Journal , vol. 20, no. 8, pp. 4382–4394, 2020

    2020

  2. [10]

    Gsm passive radar for medium range surveillance,

    R. Zemmari, U. Nickel, and W.-D. Wirth, “Gsm passive radar for medium range surveillance,” in 2009 European Radar Conference (Eu- RAD), 2009, pp. 49–52

    2009

  3. [11]

    A passive bistatic radar experiment for very low radar cross-section target detection,

    A. Aldowesh, M. Shoaib, K. Jamil, S. Alhumaidi, and M. Alam, “A passive bistatic radar experiment for very low radar cross-section target detection,” in 2015 IEEE Radar Conference , 2015, pp. 406–410

    2015

  4. [12]

    Detecting and tracking a small uav in gsm passive radar using track-before-detect,

    B. Knoedler, C. Steffes, and W. Koch, “Detecting and tracking a small uav in gsm passive radar using track-before-detect,” in 2020 IEEE Radar Conference (RadarConf20), 2020, pp. 1–6

    2020

  5. [13]

    On the detection of small uav using a gsm passive coherent location system,

    B. Knoedler, R. Zemmari, and W. Koch, “On the detection of small uav using a gsm passive coherent location system,” in 2016 17th International Radar Symposium (IRS) , 2016, pp. 1–4

    2016

  6. [14]

    Lte-based passive radar for drone detection and its experimental results,

    Y . Dan, J. Yi, X. Wan, Y . Rao, and B. Wang, “Lte-based passive radar for drone detection and its experimental results,” The Journal of Engineering , vol. 2019, no. 20, pp. 6910–6913, 2019. [Online]. Available: https://ietresearch.onlinelibrary.wiley.com/doi/abs/ 10.1049/joe....

    2019

  7. [15]

    Passive drone localization using lte signals,

    M. Sun, Z. Guo, M. Li, and R. Gerdes, “Passive drone localization using lte signals,” in Proceedings of the 15th ACM Conference on Security and Privacy in Wireless and Mobile Networks , ser. WiSec ’22. 13 New York, NY , USA: Association for Computing Machinery, 2022, p. 295–29...

    2022

  8. [16]

    Passive radar using mobile communication: A discussion of use cases and feasibility,

    C. Steffes, B. Demissie, B. Knoedler, M. Broetje, M. Mandt, and W. Koch, “Passive radar using mobile communication: A discussion of use cases and feasibility,” in 2022 IEEE Radar Conference (Radar- Conf22), 2022, pp. 1–6

    2022

  9. [17]

    R ´enyi entropy-based adaptive integration method for 5g-based passive radar drone detection,

    R. Maksymiuk, K. Abratkiewicz, P. Samczy ´nski, and M. Płotka, “R ´enyi entropy-based adaptive integration method for 5g-based passive radar drone detection,” Remote Sensing , vol. 14, no. 23, 2022. [Online]. Available: https://www.mdpi.com/2072-4292/14/23/6146

    2022

  10. [18]

    Cancellation of clutter and multipath in passive radar using a sequential approach,

    F. Colone, R. Cardinali, and P. Lombardo, “Cancellation of clutter and multipath in passive radar using a sequential approach,” in 2006 IEEE Conference on Radar , 2006, pp. 1–7–

    2006

  11. [19]

    An indoor environment sensing and localization system via mmwave phased array,

    Y . Sun, J. Li, T. Zhang, R. Wang, X. Peng, X. Han, and H. Tan, “An indoor environment sensing and localization system via mmwave phased array,” Journal of Communications and Information Networks , vol. 7, no. 4, pp. 383–393, 2022

    2022

  12. [20]

    Design and analysis of modern tracking systems(book),

    S. Blackman and R. Popoli, “Design and analysis of modern tracking systems(book),” Norwood, MA: Artech House, 1999. , 1999

    1999

  13. [21]

    An introduction to the kalman filter,

    G. Bishop, G. Welch et al., “An introduction to the kalman filter,” Proc of SIGGRAPH, Course , vol. 8, no. 27599-23175, p. 41, 2001

    2001

  14. [22]

    H. D. Griffiths and C. J. Baker, An introduction to passive radar. Artech House, 2022

    2022

  15. [23]

    Tse and P

    D. Tse and P. Viswanath, Fundamentals of wireless communication . Cambridge university press, 2005

    2005

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.