Pith. sign in

REVIEW 4 major objections 5 minor 39 references

3-D Positioning and Environment Mapping for mmWave Communication Systems

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A closed-form weighted least-squares estimator jointly locates and tracks a moving mmWave user from hybrid TDoA/FDoA/AoA measurements and maps surrounding walls from single-bounce reflections, with a neural-network variant reaching…

desk verdict A solid closed-form WLS estimator for 3D mmWave localization and mapping, but the SLAM validation is circular and the velocity results rely on synthetic FDoA. read the letter →

arxiv 1908.04142 v1 pith:SUY3ELIB submitted 2019-08-12 cs.IT eess.SPmath.IT

Jie Yang , Shi Jin , Chao-Kai Wen , Jiajia Guo , Michail Matthaiou This is my paper · ORCID
classification cs.ITeess.SPmath.IT
keywords millimeter-wavelocalizationcooperativepositioningweightedleastsquareshybridtimeandanglemeasurementssimultaneousmappingenvironmentCramér-Raolowerboundneuralnetworkresiduallearning
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 sets out to make positioning and environment mapping a by-product of millimeter-wave communication in a cloud radio access network. It derives a closed-form weighted least-squares (WLS) estimator that jointly recovers a moving user's 3-D position and velocity from hybrid time-difference-of-arrival, frequency-difference-of-arrival, and angle-of-arrival measurements, and a companion WLS estimator that locates the scatterers producing single-bounce non-line-of-sight paths. The estimators are claimed to be asymptotically unbiased and to reach the Cramér-Rao lower bound under small Gaussian measurement noise, giving decimeter-level user accuracy. A neural-network-assisted version, WLS-Net, learns the residual between the linearized model and the true measurement error instead of learning the position directly; on a public ray-tracing urban dataset, WLS-Net and its ensemble variant eWLS-Net reach centimeter-level accuracy. The interest is the combination: localization and mapping are embedded in the communication process itself, and multipath reflections are turned into information about the environment rather than treated as interference.

What carries the argument

The load-bearing mechanism is pseudo-linearization plus weighted least squares. By squaring the TDoA range equations and rewriting the angle equations with the unit vectors $a_n=(\cos\theta_n\cos\phi_n,\cos\theta_n\sin\phi_n,\sin\theta_n)^T$, $c_n=(-\sin\phi_n,\cos\phi_n,0)^T$, and $d_n=(-\sin\theta_n\cos\phi_n,-\sin\theta_n\sin\phi_n,\cos\theta_n)^T$, the unknown position and velocity enter linearly, giving pseudo-linear equations $h=Gx^\circ$. A weighting matrix $W=(B Q B^T)^{-1}$ is derived from the first-order propagation of measurement noise through the equations and is updated iteratively; this choice is what makes the estimator approach the Cramér-Rao lower bound. In WLS-Net, a fully connected network learns the residual $e$ itself, so the weighting matrix can be built from the learned residual plus a small regularization term, removing the need for a known $Q$ and for iterations.

What would settle it

Run the environment-mapping algorithm on ray-tracing scenes that include known double-bounce reflectors and check whether estimated scatterer positions land on real walls; or independently classify the second- and third-strongest paths in the public ray-tracing dataset from ground-truth ray-interaction data and count how many are actually single-bounce rather than assumed to be.

Watch

Extended reading notes

Core claim

The paper's central claim is that a single-stage closed-form WLS solution can solve the joint position/velocity and environment-mapping problems in 3-D mmWave CRANs. From the nonlinear geometry, the authors build pseudo-linear systems $h=Gx^\circ$ and $h_n^s=G_n^s s_n^\circ$ by squaring range equations and exploiting the unit direction vectors of the angle measurements. The estimator $x=(\tilde G^T W \tilde G)^{-1}\tilde G^T W \tilde h$, with $W=(B Q B^T)^{-1}$ where $B$ linearizes the measurement-error propagation, is proven asymptotically unbiased (Theorem 1) and is shown to attain the CRLB under low Gaussian noise (Theorem 2); the same construction is applied to scatterer mapping. The paper further claims that replacing the linear approximation with a neural network that learns the residual vector $e$, then using $W=(\hat e \hat e^T + a I)^{-1}$, improves accuracy when measurement errors are correlated, and that ensembling $L$ independently trained WLS-Nets with a subtractive-clustering selector gives the best results. On the ray-tracing dataset, the reported numbers are centimeter-level user-position RMSE (0.0104 m for eWLS-Net) while the pure WLS gives 0.02 m.

Load-bearing premise

Every mapped reflection is a single bounce off one point-like surface in the environment, and the second and third strongest paths in the ray-tracing data really are such single-bounce paths; if a measured path bounces twice, spreads diffusely, or has higher-order interactions, the reconstructed wall locations stop being meaningful.

Editorial extensions

If this is right

  • With three or more LoS-connected RRHs, the one-stage WLS outperforms the downlink single-bounce reference scheme, and using two RRHs already matches its accuracy.
  • In the low-noise regime the joint estimator reaches the CRLB for position and velocity; as measurement noise grows, the performance deviates only slowly.
  • The environment-mapping WLS reaches the CRLB for scatterer position, so walls and other reflectors can be reconstructed from single-bounce NLoS paths without specular-reflection assumptions or prior UE position and direction.
  • WLS-Net and eWLS-Net run without iterations and without knowing the noise covariance, cutting runtime to about 18-21% of the iterative WLS while improving accuracy when measurement errors are correlated.

Reading between the lines

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

  • Editorial inference: residual learning should also help when measurement errors carry systematic, position-dependent biases (for example antenna misalignment or clock-drift leftovers), because those are the conditions under which the paper's correlated-error experiments show the largest gains.
  • Editorial inference: since WLS-Net does not need the noise covariance matrix $Q$, a natural next step is training on simulated residual statistics and deploying in a new environment without per-site noise calibration; the paper does not test cross-scene transfer.
  • Editorial inference: the single-bounce assumption is the fragile point, so a practical SLAM system built on this method would need a separate path-order classifier to reject double-bounce, diffuse, and higher-order multipath before mapping; the paper does not provide such a classifier.
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

4 major / 5 minor

Summary. The paper considers 3-D joint position and velocity estimation and environment mapping in a millimeter-wave cloud radio access network. It formulates hybrid TDoA/FDoA/AoA measurements, derives a closed-form weighted least squares (WLS) estimator for the user position/velocity, proves asymptotic unbiasedness and CRLB attainment under a first-order noise approximation, and extends the approach to single-bounce NLoS scatterer mapping with a second WLS estimator. It then introduces WLS-Net and ensemble eWLS-Net, which learn the residual error vectors instead of the position directly, and evaluates them on a public ray-tracing dataset (DeepMIMO/Remcom). The numerical sections report decimeter-level accuracy for the WLS estimator, CRLB-consistent Monte Carlo curves, and centimeter-level position accuracy for eWLS-Net on the ray-tracing dataset.

Significance. If the results hold, the paper makes a useful contribution by providing a one-stage closed-form WLS solution for 3-D joint position/velocity estimation from hybrid measurements, with an explicit CRLB comparison and Monte Carlo validation. The neural-network-assisted extension is a sensible way to combine geometric models with learned residuals, and the use of a public ray-tracing dataset is a strength. The derivations in Appendices A–D are detailed and the CRLB-matching behavior at low noise is convincingly demonstrated for the joint position/velocity estimator. However, the environment-mapping and SLAM claims are currently supported by a CRLB that ignores the propagation of UE position error, and by a single-bounce assumption whose validation is circular; these issues must be addressed before the environment-mapping claims can be accepted.

major comments (4)
  1. [Section IV and Fig. 8] The environment-mapping CRLB comparison treats the UE position u° as exactly known when constructing h_s^n and G_s^n in (41), but in practice the scatterer estimator (44) is evaluated with the UE position estimate from Algorithm 1, and Algorithm 2 explicitly states that it uses "the UE position u° obtained in Algorithm 1." The error in u propagates into the scatterer estimate, and the RMSE of the mapping is therefore bounded by a joint CRLB that includes the UE position uncertainty, not by the conditional CRLB plotted in Fig. 8. The text in Section VI-B acknowledges that "the position estimation error of the UE will then proceed to the proposed environment mapping algorithm," but the CRLB curve does not incorporate this effect. The claim that the mapping algorithm "can achieve the CRLB" is therefore overstated; the authors should either derive the joint CRLB or explicitly present the current curve as a conditional CRLB for a known UE position.
  2. [Section VI-B, Fig. 7] The validation of the single-bounce NLoS assumption is circular. The paper states that "most of the second and third strongest paths in the ray-tracing dataset are verified to be the single-bounce NLoS paths" based on the observation that the scatterer positions estimated by the proposed algorithm "match the position of the walls." However, the environment-mapping estimator in Section IV is constrained to place each scatterer on the measured AoA ray at a distance determined by the round-trip delay under the single-bounce model. A double-bounce or higher-order path can also produce a point on a building facade that satisfies these geometric constraints, so wall matching does not distinguish single-bounce paths from higher-order paths. An independent path-order audit using the ray-tracing ground truth (e.g., labeling paths by number of interactions in Remcom) or a quantitative comparison of predicted single-bounce delays against the measured delays is needed. This issue is load-bearing because the environment-mapping estimator, WLS-Net sub-Net 2, and the SLAM results in Table II all inherit the single-bounce assumption.
  3. [Section VI-C, Table II] The FDoA measurements used in the ray-tracing experiments are not part of the public dataset; the paper states that "There are no FDoA measurements given in the ray-tracing dataset. For each user, we generate its speed in a random way, and then calculate its corresponding FDoA measurements." Consequently, the velocity RMSE entries in Table II (v = 0.0143, 0.0054, 0.0039 m/s) are not validated against the ray-tracing data but against a synthetic measurement model with randomly generated speeds. The "centimeter-level accuracy" claim is a position-only claim, and the velocity performance should be presented as a synthetic-data result, not as a property of the ray-tracing dataset. This limitation should be stated explicitly wherever the Table II results are summarized.
  4. [Section III-C, Theorem 1 and Remark 1] Theorem 1 states that the estimator is "asymptotically unbiased," and Remark 1 explains this as meaning that "the proposed algorithm will become increasingly accurate as the number of measurements increases." The proof, however, shows only that E{Δx} ≈ 0 under the first-order approximation E{e} ≈ B E{Δm} = 0, which is an approximate unbiasedness for small Gaussian noise, not an asymptotic property in the number of measurements. If consistency as N_a grows is intended, a different argument is required; as written, the theorem and the remark do not match the proof. The statement should be revised to "approximately unbiased under small measurement noise" or a proper asymptotic analysis should be provided.
minor comments (5)
  1. [Algorithm 2, line 2] Algorithm 2 says "the UE position u° obtained in Algorithm 1," but u° denotes the true value elsewhere in the paper; the algorithm actually uses the estimate u from Algorithm 1. Rename the variable to u or u_hat to avoid confusion.
  2. [Section VI-B, Fig. 8] The RMSE curves in Figs. 5–9 are plotted in dB, but the axis labels do not specify the reference value (e.g., dB relative to 1 m or 1 m/s). Adding the reference would make the figures self-contained.
  3. [Notations] In the Notations section, "|c| denotes the module of c" should read "modulus of c."
  4. [Section VI-C] The paper introduces a "dominant fixed error" component in the D0–D4 scenarios, but the theoretical analysis in Section III-C assumes a zero-mean Gaussian noise vector with covariance Q. Clarify that the WLS optimality and CRLB analysis apply to the Gaussian random part, while the neural-network methods are designed to also learn the fixed part.
  5. [Section V-A, Eq. (46)] The regularization constant a in W = (ê ê^T + aI)^{-1} is described only as "a very small disturbance value," but no value is reported in the numerical section. Please state the chosen value and its sensitivity, since it affects the WLS-Net results.

Circularity Check

1 steps flagged · score 3.0 of 10

Closed-loop validation of the single-bounce assumption; the WLS/CRLB core derivation is independent.

  1. other [Section VI-B, Fig. 7 (environment mapping with the ray-tracing dataset); Algorithm 2 and Eq. (33)]
    "The estimated locations of the scatterers match the position of the walls of the buildings ... Second, most of the second and third strongest paths in the ray-tracing dataset are verified to be the single-bounce NLoS paths, which also corroborates the assumption of the proposed environment mapping algorithm."

    The verification uses scatterer positions estimated by the proposed environment-mapping WLS solver (Algorithm 2), whose model (33)-(39) is derived from the single-bounce relation d1 + d2 = r_s + r_1. For a double-bounce or higher-order path, this solver still returns the point on the measured AoA ray whose round-trip distance equals the measured delay, and that point can lie on a building wall behind the first reflection point. Therefore, wall-matching of the estimated points is a consistency check of the single-bounce model with itself, not an independent path-order audit.

full rationale

The load-bearing statistical derivations are not circular. The WLS joint position/velocity estimator (Section III) is derived from noise-free pseudo-linear equations, and its asymptotic unbiasedness and CRLB attainment (Theorems 1 and 2) are proven against an externally defined CRLB (Eq. 22) using first-order perturbation of the measurement noise; the Monte Carlo comparisons in Section VI-A provide an independent numerical check. The WLS-Net/eWLS-Net results are evaluated on held-out samples from an external ray-tracing dataset, so the learned residual vectors are not fitted to the test targets. The one genuine circular element is the validation of the single-bounce NLoS assumption in Section VI-B: the environment-mapping algorithm that is used to produce the scatterer positions already assumes single-bounce, and those estimated positions are then used to claim that the dataset paths are indeed single-bounce. This does not invalidate the WLS estimation theory, but it means the environment-mapping/SLAM credibility currently rests on a partially self-confirming consistency check rather than on independent evidence of path order.

Assumptions & free parameters 3 free parameters · 7 assumptions · 0 invented entities

The central derivation is a pseudo-linearization of standard geometric relations, so the free-parameter count is small. The load-bearing assumptions are the sparse single-bounce channel, known Gaussian noise covariances, and first-order error linearization; the neural-network stage adds untabulated hyperparameters. No invented physical entities are introduced.

free parameters (3)
  • a (WLS-Net regularization constant) = not specified
    Added to the learned residual outer product in Eq. (46) to keep the weighting matrix invertible; its value changes the effective weighting and is never reported.
  • ra and rb (subtractive clustering radii) = not specified
    Radii in Eqs. (49) and (50) control which ensemble predictions are clustered and which is selected; their values are not given in the paper.
  • L (ensemble size) = 10
    Ten independently trained WLS-Nets are used in eWLS-Net; this hand-chosen engineering choice affects both accuracy and runtime.
assumptions (7)
  • domain assumption Only LoS and single-bounce NLoS paths exist in the mmWave channel, and each RRH-UE pair has at most one scatterer for the model.
    Stated in Section II-D and used to build the mapping model (40) and the geometry in Figure 1; it is load-bearing for environment mapping.
  • domain assumption The LoS path is stronger than NLoS paths and can be selected via a threshold, so Na LoS RRHs are known.
    Section II-C assumes a threshold selects Na LoS paths; no NLoS identification or path-association algorithm is given.
  • domain assumption Measurement noise is a zero-mean Gaussian vector with known covariance Q, and similarly Qs_n for mapping.
    Used in Lemma 1, in W = (B Q B^T)^-1, and in the CRLB expression (22); the WLS-Net later relaxes the need to know Q.
  • domain assumption The error vector e is well approximated by the first-order term B Delta m.
    Appendix B uses a Taylor expansion to obtain e approximately equal to B Delta m; the CRLB-equivalence proof in Appendix D relies on this linearization being accurate.
  • domain assumption The UE position estimate from Algorithm 1 is treated as exact when mapping scatterers and when computing the environment-mapping CRLB.
    Algorithm 2 and Section IV feed u from Algorithm 1 into the mapping model without modeling its uncertainty; Section VI-B acknowledges error propagation only qualitatively.
  • domain assumption Scatterers are static point reflectors and the environment is stationary during estimation.
    The mapping model in Section IV estimates a fixed s_n; moving scatterers and extended surfaces are not modeled.
  • ad hoc to paper The second and third strongest paths in the ray-tracing dataset correspond to single-bounce NLoS paths.
    Section VI-B claims these paths are mostly single-bounce and uses them to validate the assumption, but no independent path-type labels are shown.

how reviews work

0 comments
Cite this review

Pith. "Pith review of 3-D Positioning and Environment Mapping for mmWave Communication Systems." pith.science (2026). https://pith.science/paper/SUY3ELIB

@misc{pith2026190804142,
  author       = {Pith},
  title        = {Pith review of: 3-D Positioning and Environment Mapping for mmWave Communication Systems},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SUY3ELIB}},
  note         = {Machine review of arXiv:1908.04142}
}
read the original abstract

Millimeter-wave (mmWave) cloud radio access networks (CRANs) provide new opportunities for accurate cooperative localization, in which large bandwidths, large antenna arrays, and increased densities of base stations allow for their unparalleled delay and angular resolution. Combining localization into communications and designing simultaneous localization and mapping (SLAM) algorithms are challenging problems. This study considers the joint position and velocity estimation and environment mapping problem in a three-dimensional mmWave CRAN architecture. We first embed cooperative localization into communications and establish the joint estimation and mapping model with hybrid delay and angle measurements. Then, we propose a closed-form weighted least square (WLS) solution for the joint estimation and mapping problems. The proposed WLS estimator is proven asymptotically unbiased and confirmed by simulations as effective in achieving the Cramer-Rao lower bound (CRLB) and the desired decimeter-level accuracy. Furthermore, we propose a WLS-Net-based SLAM algorithm by embedding neural networks into the proposed WLS estimators to replace the linear approximations. The solution possesses both powerful learning ability of the neural network and robustness of the proposed geometric model, and the ensemble learning is applied to further improve positioning accuracy. A public ray-tracing dataset is used in the simulations to test the performance of the WLS-Net-based SLAM algorithm, which is proven fast and effective in attaining the centimeter-level accuracy.

Figures

Figures reproduced from arXiv: 1908.04142 by the authors.

Figure 1
Figure 1. Illustration of the mmWave CRAN system model in which RRHs are connected with the CU. [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Structure of the neural network-assisted WLS algorithm with two sub-nets. [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Structure of the ensemble learning-based WLS-Net. [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗
Figures from the paper (3 more)
Figure 5
Figure 5. Figure 5: Performance analysis of the presented positioning algo [PITH_FULL_IMAGE:figures/full_fig_p019_5.png]
Figure 6
Figure 6. Figure 6: RMSE as a function of σd with a given σa for (a) position of UE and (b) velocity of UE. ;ĂͿ ;ďͿ [PITH_FULL_IMAGE:figures/full_fig_p020_6.png]
Figure 8
Figure 8. Figure 8: Comparison of RMSEs of the proposed environment [PITH_FULL_IMAGE:figures/full_fig_p021_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 38 canonical work pages

  1. [1]

    Massive MIMO is a Reality -- What is Next? Five Promising Research Directions for Antenna Arrays

    E. Bj ¨ornson, L. Sanguinetti, H. Wymeersch, J. Hoydis, and T. L. Marzetta, “Massive MIMO is a reality-what is next? Five promising research directions for antenna arrays,” https://arxiv.org/abs/1902.07678

    work page Pith review arXiv 1902

  2. [2]

    Localization as a feature of mmWave communication,

    F. Lemic, J. Martin, C. Yarp, D. Chan, V . Handziski, R. Brodersen, G. Fettweis, A. Wolisz, and J. Wawrzynek, “Localization as a feature of mmWave communication,” in Proc. IEEE IWCMC , Sep. 2016, pp. 1033-1038

    work page 2016

  3. [3]

    Survey of cellular mobile radio localization methods: From 1G to 5G,

    J. A. del Peral-Rosado, R. Raulefs, J. A. L ´opez-Salcedo, and G. Seco-Granados, “Survey of cellular mobile radio localization methods: From 1G to 5G,” IEEE Commun. Surv. Tutor., vol. 20, no. 2, pp. 1124-1148, May 2018

    work page 2018

  4. [4]

    5G mmWave positioning for vehicular networks,

    H. Wymeersch, G. Seco-Granados, G. Destino, D. Dardari, and F. Tufvesson, “5G mmWave positioning for vehicular networks,” IEEE Wireless Commun. Mag. , vol. 24, no. 6, pp. 80-86, Dec. 2017

    work page 2017

  5. [5]

    An overview of signal processing techniques for millimeter wave MIMO systems,

    R. W. Heath, Jr., N. G. Prelcic, S. Rangan, W. Roh, and A. Sayeed, “An overview of signal processing techniques for millimeter wave MIMO systems,” IEEE J. Sel. Top. Signal Process. , vol. 10, no. 3, pp. 436-453, Apr. 2016

    work page 2016

  6. [6]

    Investigation of transmission schemes for millimeter-wave massive MU-MIMO systems,

    Y . Han, H. Zhang, S. Jin, X. Li, R. Yu, and Y . Zhang, “Investigation of transmission schemes for millimeter-wave massive MU-MIMO systems,” IEEE Syst. J. , vol. 11, no. 1, pp. 72-83, Mar. 2017

    work page 2017

  7. [7]

    Mm-wave MIMO channel modeling and user localization using sparse beamspace signatures,

    H. Deng and A. M. Sayeed, “Mm-wave MIMO channel modeling and user localization using sparse beamspace signatures,” in Proc. IEEE SPAWC, Jun. 2014, pp. 130-134

    work page 2014

  8. [8]

    Beamspace channel estimation in mmWave systems via cosparse image reconstruction technique,

    J. Yang, C. K. Wen, S. Jin, and F. Gao, “Beamspace channel estimation in mmWave systems via cosparse image reconstruction technique,” IEEE Trans. Commun. , vol. 66, no. 10, pp. 4767-4782, Oct. 2018

    work page 2018

Show all 39 references
  1. [9]

    Uplink channel estimation and data transmission in millimeter-wave CRAN with lens antenna arrays,

    R. G. Stephen and R. Zhang, “Uplink channel estimation and data transmission in millimeter-wave CRAN with lens antenna arrays,” IEEE Trans. Commun. , vol. 66, no. 12, pp. 6542-6555, Jul. 2018

    2018

  2. [10]

    Location-aware communications for 5G networks: How location information can improve scalability, latency, and robustness of 5G,

    R. D. Taranto, S. Muppirisetty, R. Raulefs, D. Slock, T. Svensson, and H. Wymeersch, “Location-aware communications for 5G networks: How location information can improve scalability, latency, and robustness of 5G,” IEEE Signal Process. Mag. , vol. 31, no. 6, pp. 102-112, Nov. 2014

    2014

  3. [11]

    A survey on 5G massive MIMO localization,

    F. Wen, H. Wymeersch, B. Peng, W. P. Tay, H. C. So, and D. Yang, “A survey on 5G massive MIMO localization,” Digit. Signal Process, 2019, https://doi.org/10.1016/j.dsp.2019.05.005

    2019 doi

  4. [12]

    Positioning via direct localization in C-RAN systems,

    S. Jeong, O. Simeone, A. Haimovich, and J. Kang, “Positioning via direct localization in C-RAN systems,” IET Commun. , vol. 10, no. 16, pp. 2238-2244, Jun. 2016

    2016

  5. [13]

    Direct localization for massive MIMO,

    N. Garcia, H. Wymeersch, E. G. Larsson, A. M. Haimovich, and M. Coulon, “Direct localization for massive MIMO,” IEEE Trans. Signal Process., vol. 65, no. 10, pp. 2475-2487, May 2017

    2017

  6. [14]

    Weighted least squares algorithm for target localization in distributed MIMO radar,

    M. Einemo and H. C. So, “Weighted least squares algorithm for target localization in distributed MIMO radar,” Signal Process., vol. 115, no. C, pp. 144-150, Oct. 2015

    2015

  7. [15]

    Asymptotically efficient target localization from bistatic range measurements in distributed MIMO radars,

    R. Amiri, F. Behnia, and H. Zamani, “Asymptotically efficient target localization from bistatic range measurements in distributed MIMO radars,” IEEE Signal Process. Lett. , vol. 24, no. 3, pp. 299-303, Jan. 2017

    2017

  8. [16]

    An asymptotically efficient estimator in closed-form for 3-D AOA localization using a sensor network,

    Y . Wang and K. C. Ho, “An asymptotically efficient estimator in closed-form for 3-D AOA localization using a sensor network,” IEEE Trans. Wireless Commun. , vol. 14, no. 12, pp. 6524-6535, Jul. 2015

    2015

  9. [17]

    Efficient 3-D positioning using time-delay and AoA measurements in MIMO radar systems,

    R. Amiri, F. Behnia, and H. Zamani, “Efficient 3-D positioning using time-delay and AoA measurements in MIMO radar systems,” IEEE Commun. Lett. , vol. 21, no. 12, pp. 2614-2617, Aug. 2017

    2017

  10. [18]

    An accurate algebraic solution for moving source location using TDoA and FDoA measurements,

    K. C. Ho and W. W. Xu, “An accurate algebraic solution for moving source location using TDoA and FDoA measurements,” IEEE Trans. Signal Process. , vol. 52, no. 9, pp. 2453-2463, Aug. 2004

    2004

  11. [19]

    Multistatic pseudolinear target motion analysis using hybrid measurements,

    N. H. Nguyen and K. Dogancay, “Multistatic pseudolinear target motion analysis using hybrid measurements,” Signal Process., vol. 130, no. C, pp. 22-36, Jan. 2017

    2017

  12. [20]

    5G mmWave downlink vehicular positioning,

    H. Wymeersch, N. Garcia, H. Kim, G. Seco-Granados, S. Kim, F. Wen, and M. Fr ¨ohle, “5G mmWave downlink vehicular positioning,” in Proc. IEEE GLOBECOM , Dec. 2018, pp. 1-6

    2018

  13. [21]

    Efficient multisensor localization for the internet of things: Exploring a new class of scalable localization algorithms,

    M. Z. Win, F. Meyer, Z. Liu, W. Dai, S. Bartoletti, and A. Conti, “Efficient multisensor localization for the internet of things: Exploring a new class of scalable localization algorithms,” IEEE Signal Process. Mag. vol. 35, no. 5, pp. 153-167, Sept. 2018

    2018

  14. [22]

    Distributed localization: A linear theory,

    S. Safavi, U. A. Khan, S. Kar, and J. M. F. Moura, “Distributed localization: A linear theory,” Proc. IEEE, vol. 106, no. 7, pp. 1204-1223, Jul. 2018. 30

    2018

  15. [23]

    Nonline-of-sight error mitigation in mobile location,

    L. Cong and W. Zhuang, “Nonline-of-sight error mitigation in mobile location,” IEEE Trans. Wireless Commun. , vol. 4, no. 2, pp. 560-573, Mar. 2005

    2005

  16. [24]

    A survey on TOA based wireless localization and NLOS mitigation techniques,

    I. Guvenc and C.-C. Chong, “A survey on TOA based wireless localization and NLOS mitigation techniques,” IEEE Commun. Surv. Tutor., vol. 11, no. 3, pp. 107-124, Aug. 2009

    2009

  17. [25]

    High-accuracy localization for assisted living: 5G systems will turn multipath channels from foe to friend,

    K. Witrisal et al., “High-accuracy localization for assisted living: 5G systems will turn multipath channels from foe to friend,” IEEE Signal Process. Mag. , vol. 33, no. 2, pp. 59-70, Mar. 2016

    2016

  18. [26]

    Simultaneous localization and mapping: Part I,

    H. Durrant-Whyte and T. Bailey, “Simultaneous localization and mapping: Part I,” IEEE Robot. Autom. Mag. , vol. 13, no. 2, pp. 99-110, Jun. 2006

    2006

  19. [27]

    Leveraging mmWave imaging and communications for simultaneous localization and mapping,

    M. Aladsani, A. Alkhateeb, and G. C. Trichopoulos, “Leveraging mmWave imaging and communications for simultaneous localization and mapping,” in Proc. IEEE ICASSP , May. 2019, pp. 4539-4543

    2019

  20. [28]

    Multipath assisted positioning with simultaneous localization and mapping,

    C. Gentner, T. Jost, W. Wang, S. Zhang, A. Dammann, and U. Fiebig, “Multipath assisted positioning with simultaneous localization and mapping,” IEEE Trans. Wireless Commun. , vol. 15, no. 9, pp. 6104-6117, Sept. 2016

    2016

  21. [29]

    Millimeter wave channel modeling and cellular capacity evaluation,

    M. R. Akdeniz et al., “Millimeter wave channel modeling and cellular capacity evaluation,” IEEE J. Sel. Areas Commun. , vol. 32, no. 6, pp. 1164-1179, Jun. 2014

    2014

  22. [30]

    38GHz and 60GHz angle-dependent propagation for cellular & peer-to-peer wireless communications,

    T. S. Rappaport, E. Ben-Dor, J. N. Murdock, and Y . Qiao, “38GHz and 60GHz angle-dependent propagation for cellular & peer-to-peer wireless communications,” in Proc. IEEE ICC , Jun. 2012, pp. 4568-4573

    2012

  23. [31]

    DeepFi: Deep learning for indoor fingerprinting using channel state information,

    X. Wang, L. Gao, S. Mao, and S. Pandey, “DeepFi: Deep learning for indoor fingerprinting using channel state information,” in Proc. IEEE WCNC , Mar. 2015, pp. 1666-1671

    2015

  24. [32]

    CSI-based outdoor localization for massive MIMO: Experiments with a learning approach,

    A. Decurninge, L. G. Ord ´o˜nez, P. Ferrand, G. He, B. Li, W. Zhang, and M. Guillaud, “CSI-based outdoor localization for massive MIMO: Experiments with a learning approach,” in Proc. IEEE ISWCS , Aug. 2018, pp. 1-6

    2018

  25. [33]

    DeepMIMO: A generic deep learning dataset for millimeter wave and massive MIMO applications,

    A. Alkhateeb, “DeepMIMO: A generic deep learning dataset for millimeter wave and massive MIMO applications,” in Proc. IEEE ITA, Feb. 2019, pp. 1-8

    2019

  26. [34]

    Wireless insite,

    Remcom, “Wireless insite,” http://www.remcom.com/wireless-insite

  27. [35]

    Channel estimation in broadband millimeter wave MIMO systems with few-bit ADCs,

    J. Mo, P. Schniter, and R. W. Heath. Jr, “Channel estimation in broadband millimeter wave MIMO systems with few-bit ADCs,” IEEE Trans. Signal Process. , vol. 66, no. 5, pp. 1141-1154, Jan. 2018

    2018

  28. [36]

    Initial access in milllimeter wave cellular systems,

    C. N. Barati et al., “Initial access in milllimeter wave cellular systems,” IEEE Trans. Wireless Commun. , vol. 15, no. 12, pp. 7926-7940, Dec. 2016

    2016

  29. [37]

    S. M. Kay, Fundamentals of statistical signal processing, Estimation Theory. Englewood Cliffs, NJ, USA: Prentice-Hall, 1993

    1993

  30. [38]

    T. G. Dietterich, Ensemble learning, The handbook of brain theory and neural networks. Cambridge, MA: MIT Press, 2002

    2002

  31. [39]

    Towards optimal power control via ensembling deep learning neural networks,

    F. Liang, C. Shen, W. Yu, and F. Wu, “Towards optimal power control via ensembling deep learning neural networks,” https://arxiv.org/abs/1807.10025

    arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.