Pith. sign in

REVIEW 3 major objections 10 minor 42 references

PC-DeepNet: A GNSS Positioning Error Minimization Framework Using Permutation-Invariant Deep Neural Network

T0 review · 3 major / 10 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read PC-DeepNet claims that learning position corrections from permutation-invariant satellite features cuts average horizontal GNSS error to 7.77 m, below WLS, r-WLS, and Kalman-filter baselines.

desk verdict A plausible, modest ML-for-GNSS contribution whose headline result is not yet verifiable because the train/test split is under-documented and may leak temporally correlated epochs. read the letter →

arxiv 2504.13990 v1 pith:HG5462FE submitted 2025-04-18 cs.LG cs.AIcs.SYeess.SY

M. Humayun Kabir , Md. Ali Hasan , Md. Shafiqul Islam , Kyeongjun Ko , Wonjae Shin This is my paper · ORCID
classification cs.LGcs.AIcs.SYeess.SY
keywords GNSSpositioningpermutation-invariantneuralnetworkNLOSandmultipathmitigationpositioncorrectionlearningAndroidrawmeasurementsurbanaccuracyr-WLSinitialguessIoTlocalization
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

PC-DeepNet is a learning-based framework for correcting GNSS positions in urban and suburban environments. The paper tries to establish that a permutation-invariant deep network, fed per-satellite features such as pseudorange residual, line-of-sight vector, GDOP, elevation angle, and $C/N_0$, can estimate a three-dimensional correction to an r-WLS initial fix and thereby reduce positioning error under NLOS and multipath conditions. Using two public Android raw-GNSS datasets and three driving scenarios, the paper reports an aggregate horizontal score of 7.77 m versus 12.02 m for WLS, 10.26 m for r-WLS, and 8.97 m for a Kalman filter, with 88,033 parameters and about 340 KB of memory compared with 151,107 parameters and 611 KB for the prior set-transformer baseline [17]. The interest is that a small, infrastructure-free network could bring meter-level urban positioning to IoT devices.

What carries the argument

The load-bearing object is the permutation-invariant deep network (PI-DNN), whose mapping is $\Psi(\{v_i\})=\psi(\sum_{i=1}^M \phi(v_i))$, where $\phi$ is a four-layer Leaky ReLU encoder, the sum is the permutation-invariant aggregation, and $\psi$ is a four-layer Leaky ReLU decoder. Each input $v_i$ is the seven-dimensional feature vector of one satellite: pseudorange residual, line-of-sight unit vector components ($g^x, g^y, g^z$), GDOP, elevation angle, and $C/N_0$. Because the aggregation is a sum, the network output is invariant to satellite ordering and tolerant of a changing satellite count. The decoder outputs the predicted 3D correction $\delta x$, which is added to the r-WLS initial guess to give $\hat{x}=x_{\text{initial}}+\delta x$; training targets are the differences between ground truth and r-WLS positions.

What would settle it

Run the same training and evaluation with each entire driving route assigned wholly to training, validation, or testing, and compare the aggregate horizontal score; if the 7.77 m result rises to near or above the 10.26 m r-WLS baseline, route memorization rather than learned error correction explains the gain.

Watch

Extended reading notes

Core claim

The paper's central claim is that the non-linear, non-Gaussian error remaining after a robust weighted least-squares fix can be learned by a permutation-invariant network that treats visible satellites as a set. The network maps a set of seven-dimensional satellite feature vectors to the position correction $\delta x = x_{\text{truth}} - x_{\text{initial}}$, and the final position is $x_{\text{initial}} + \delta x$. On the paper's score metric (the mean of per-trace 50th and 95th percentile horizontal errors), the method scores 7.77 m overall, with scenario-level scores of 11.98 m, 6.76 m, and 4.56 m in dense urban San Francisco, suburban San Francisco, and urban Los Angeles respectively, outperforming WLS (12.02 m), r-WLS (10.26 m), and Kalman filtering (8.97 m) while using roughly half the parameters of the earlier neural-network approach [17].

Load-bearing premise

The claimed accuracy rests on the assumption that the training, validation, and test split separates entire driving routes, so no road driven during training appears in the test set; if the split instead mixes individual time samples across routes, the network could memorize familiar road segments and the reported gains would shrink on routes it has never seen.

Editorial extensions

If this is right

  • If the reported numbers hold under trace-level generalization, a smartphone-grade receiver can cut horizontal error by roughly a quarter relative to r-WLS and by about 13 percent relative to KF, with no ground infrastructure.
  • The same trained model can be applied when satellites enter or leave view, because sum-pooling makes the output independent of satellite ordering and count.
  • The corrected position is a simple sum of the r-WLS fix and the network output, so the framework can be dropped into existing least-squares GNSS pipelines.
  • The claimed computational savings (88,033 parameters, about 340 KB) make on-device inference plausible on the resource-constrained hardware typical of IoT nodes.

Reading between the lines

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

  • Editorial inference: the headline 7.77 m average hides the dense-urban scenario-I score of 11.98 m, so the real-world benefit in deep urban canyons may be smaller than the aggregate suggests.
  • Editorial inference: the decisive generalization test is cross-city transfer (train on San Francisco, test on Los Angeles, or the reverse); the paper's within-city scenario splits do not fully rule out route-level memorization.
  • Editorial inference: the permutation-invariant encoder could be extended with learned attention weights or an uncertainty head, producing per-fix confidence estimates and possibly better corrections than plain sum-pooling; these are testable variants the paper does not explore.
  • Editorial inference: because the paper states that the current implementation relies on GPS signals alone, combining the learned correction with other constellations or with inertial and 5G measurements is a natural next step whose benefit is not quantified here.
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 / 10 minor

Summary. The paper proposes PC-DeepNet, a permutation-invariant deep neural network that maps per-satellite GNSS features (pseudorange residual, LOS vector, GDOP, C/N0, and elevation angle) to a 3D position correction applied on top of an r-WLS initial solution. The encoder-aggregation-decoder architecture uses sum pooling so that the output is invariant to the number and order of visible satellites. The authors evaluate on two public Google Android raw-GNSS datasets with three scenarios (urban and suburban San Francisco, urban Los Angeles) and report an aggregate horizontal score of 7.77 m versus 12.02 m (WLS), 10.26 m (r-WLS), and 8.97 m (KF), as well as 88,033 model parameters versus 151,107 for the learning-based baseline of Kanhere et al. The central claim is that the learned r-WLS residual correction generalizes to held-out urban and suburban traces.

Significance. If the results survive a properly documented trace-disjoint evaluation, this is a useful, modest contribution: it demonstrates that a set-based deep network with NLOS/multipath-indicator features (C/N0, elevation, GDOP, pseudorange residuals, LOS vector) can consistently beat WLS, r-WLS, and KF across three scenarios from two public datasets, at a smaller model size than a previous learning-based method. The supervised residual-learning setup (predicting ground-truth-minus-r-WLS corrections from features available at inference) is methodologically sound and not circular. Strengths include a clearly defined score (Eq. 20), an architecture-depth ablation (Table VI), a parameter ledger (Table V), and evaluation on publicly available datasets. Its significance is limited by the absence of code, by the lack of trace-level split documentation, and by the absence of error bars for the headline scores; the margin over the strongest baseline (KF) is about 1.2 m (13%) on the aggregate score, which could be sensitive to split choices and test-trace composition.

major comments (3)
  1. [Section V (and Section IV)] The central claim of the paper is that the learned correction generalizes to held-out traces, but a trace-level train/test split is never documented. Section V states only that "a training-validation-test split of 75%, 10%, and 15% of data" was used, which is a statement about data fractions rather than about which driving traces belong to which split, and the sentence "For all the cases, the training and test traces are different" is an assertion unsupported by any trace-ID table or by a description of how the 75/10/15 rule is applied to trajectories. This matters because adjacent epochs within one trace share satellite geometry, multipath environment, and receiver bias; an epoch-level random split would let the network memorize route-specific error patterns and could largely erase the reported 7.77 m score on truly unseen routes. The manuscript also contradicts itself: Section IV says dataset-I has 29 training traces, while Section V says 55 trajectories in dataset-I are used in scenario-I, and a six-trace scenario-III would leave fewer than one trace for validation under a 10% rule. Finally, the reported aggregate score cannot be reconciled with Table I without per-scenario test-trace counts: under the natural reading that the 75/10/15 rule is applied separately to the stated 55/55/6 trajectories, the weighted aggregate of the Table I scores is about 9.1 m, not the reported 7.77 m. Please provide a trace-level split table (trace IDs and counts per split per scenario), state the exact split rule, and re-run the evaluation with a strictly trace-disjoint split.
  2. [Section V, Fig. 13, Table IV, Eq. (20)] The comparison with the learning-based baseline is not verifiable as reported. The paper reports Kanhere et al.'s scenario-I values (6.4±5.2, 5.9±5.0, 6.2±4.9 m in Table IV) but never states whether these numbers were obtained by re-running their model (or a re-implementation) on the same trace-disjoint test traces used for PC-DeepNet, or whether they are quoted from [17]; since the score in Eq. (20) is a trace-averaged percentile, values computed on different test-trace sets are not comparable. In addition, the headline comparison in Fig. 13 (7.77 m vs 8.97 m for KF) is presented without error bars, confidence intervals, or the number of test traces per scenario; Table III provides intervals only for NED MAEs, not for the score. Please state the protocol used for the Kanhere comparison, report the distribution of per-trace scores, and provide an error estimate (for example, a bootstrap confidence interval over test traces) for the headline numbers.
  3. [Section V, Table IV] Table IV, the direct comparison with the learning-based baseline, contains corrupted entries: the KF down value is printed as "39.8.33±9.62" and the PC-DeepNet down value as "5.26.2±4.9", which cannot be interpreted as printed. The surrounding paragraph also misassigns scenario labels: PC-DeepNet's 2D MAE of 6.95 m is attributed to scenario-II within a passage describing scenario-I, an r-WLS MAE is attributed to scenario-I in the middle of the scenario-II discussion, and scenario-III is called a "sub-urban" area after being introduced as urban. Because the per-scenario MAE values appear nowhere else and do not match Table I (which reports percentiles), these errors make the per-scenario accuracy claims impossible to verify. Please correct the table, add scenario labels to each reported MAE, and reconcile the MAE reporting with the score reporting.
minor comments (10)
  1. [Table V] The decoder row labeled "Dropout-15" contradicts the 2% dropout stated in Section III-B and in the caption of Fig. 3; the label should read "Dropout" with the stated rate.
  2. [Table V] The Output row reports 33 parameters; a linear layer mapping 32 features to 3 outputs would have 96 weights plus 3 biases, i.e., 99 parameters. Please verify the parameter ledger and the reported total of 88,033.
  3. [Section V] In the MAE paragraph, the final PC-DeepNet sentence ("ranging from 0.16 m to 22.80 m with an MAE of 3.27 m") has no scenario label, and KF's MAE of 3.97 m in scenario-III repeats the r-WLS value of 3.97 m; please check these numbers.
  4. [Section III-A and III-B] The seven features have incompatible units and scales (pseudorange residuals in meters, LOS unit-vector components in [-1,1], elevation in degrees, C/N0 in dB-Hz, GDOP dimensionless), but no standardization or normalization is described, even though training uses a fixed Adam learning rate of 1e-3; please document the preprocessing including any clipping or per-feature scaling.
  5. [Section V (complexity analysis)] The "lower computational complexity" claim is supported only by parameter count and stored model size; reporting inference FLOPs or latency on a mobile-class processor would substantiate the claim, since the comparison architecture differs in operations beyond parameter count.
  6. [Fig. 13 and Eq. (20)] The pooled "50th percentile of error" and "95th percentile of error" shown in Fig. 13 are not defined; Eq. (20) defines only the trace-averaged score. Please state how the pooled percentiles are computed from the per-trace distributions.
  7. [Sections I and II] The critique that Kanhere et al. "does not train the model with proper position corrections" overstates the matter: adding uniform noise to the true position is a legitimate way to generate initial-guess training targets. Please rephrase or support this criticism with an experiment.
  8. [General (reproducibility)] No code, random seed, preprocessing scripts, or software versions are provided; given the sensitivity of percentile-based scores to split composition, releasing the exact split files and training code is essential for independent verification.
  9. [Throughout] Minor typographical issues include "BeioDou" in the conclusion, the malformed summation index in Eq. (15), and the phrase "α=0.1 (set value empirically)" without a sensitivity analysis for the Leaky ReLU slope and the 2% dropout rate.
  10. [Section V (KF baseline)] The KF baseline configuration is not described: state dimension, process and measurement noise parameters, and whether it consumes the same pseudoranges as WLS/r-WLS. Given the very large down-direction errors reported for KF in Table IV (about 40 m), this configuration should be specified.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation found: PC-DeepNet is supervised residual learning evaluated against external baselines; only a trivial non-load-bearing self-citation is present.

full rationale

The paper's core pipeline is not circular. The network is trained to map per-satellite features (pseudorange residual, LOS vector, GDOP, C/N0, elevation) to the correction δx = x_true − x_initial, where x_initial is the r-WLS solution, and the reported final position is x̂ = x_initial + δx̂ (Eq. 17). This is standard supervised residual learning: the correction is predicted from features on held-out traces and is not copied from the label, and the headline score of 7.77 m is computed from predicted positions using Eqs. (18)–(20). The comparison targets (WLS, r-WLS, KF, and Kanhere et al.) are external methods, not outputs of the model. The only self-citation is reference [1] (Kabir, Lee, Shin), used in the introductory sentence that GPS, Galileo, BeiDou, and GLONASS provide global services; this is a standard factual claim and is not load-bearing for the architecture, the training objective, or the reported accuracy. A separate evaluation-validity concern remains, namely whether the 75/10/15 split is truly trace-disjoint given the discrepancy between 29 training traces in Section IV and 55 trajectories in Section V; that is a possible generalization/leakage risk, not a circularity of the derivation. Therefore no circular step is identified; the score of 2 reflects only the presence of one minor, non-load-bearing self-citation.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central claim rests on standard GNSS measurement modeling, a universal-approximation justification for sum pooling, and several domain assumptions about ground truth and split hygiene. The only hand-fitted quantities are model hyperparameters; no new physical entities are introduced.

free parameters (4)
  • Leaky ReLU negative slope alpha = 0.1
    Set empirically per Section III-B; affects which negative activations propagate and is tuned on the validation set.
  • Dropout rate = 2%
    Applied after the third encoder layer and in the decoder; chosen to control overfitting, with no sensitivity analysis reported.
  • Encoder/decoder depth and widths = 4 layers of sizes 32, 64, 128, 256 (encoder) and 256, 128, 64, 32 (decoder)
    Selected by comparing model-1 through model-4 in Table VI; the central accuracy claim depends on this architecture choice.
  • Training epochs and batch size = 100 epochs, batch size 8
    Model trained for up to 100 epochs with batch size 8; no early stopping or schedule is described.
assumptions (6)
  • domain assumption The pseudorange after corrections follows rho_m = ||x - x_m|| + b_r + eps_tilde_m (Eq. 5), with residual errors treated as learnable signal.
    Standard GNSS measurement model; the paper relies on it to define residuals and LOS vectors in Section III-A.
  • domain assumption First-order Taylor linearization around the r-WLS initial guess is accurate enough for computing residuals (Eqs. 6-9).
    The method computes pseudorange residuals from this linearization; if the initial guess is far off, residuals are less informative.
  • standard math A sum-pooling permutation-invariant network with this encoder capacity can approximate the position-correction function (universal approximation for set functions, Wagstaff et al. [36]).
    The architecture is justified by the cited universal approximation result; no proof is given in this paper.
  • domain assumption The seven features are sufficient to predict the position correction of r-WLS, and the mapping learned on training traces transfers to held-out traces.
    This is the core learning assumption; if features do not capture the dominant error sources, the reported gains would not generalize.
  • domain assumption The NovAtel SPAN ground truth in the Android datasets is accurate enough to serve as training labels.
    Target corrections are computed as ground truth minus r-WLS position; errors in ground truth would bias all labels.
  • domain assumption The training, validation, and test split separates whole trajectories so that temporally correlated samples do not leak.
    The paper states training and test traces are different but does not document trace-level assignment; leakage would inflate the reported accuracy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PC-DeepNet: A GNSS Positioning Error Minimization Framework Using Permutation-Invariant Deep Neural Network." pith.science (2026). https://pith.science/paper/HG5462FE

@misc{pith2026250413990,
  author       = {Pith},
  title        = {Pith review of: PC-DeepNet: A GNSS Positioning Error Minimization Framework Using Permutation-Invariant Deep Neural Network},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HG5462FE}},
  note         = {Machine review of arXiv:2504.13990}
}
read the original abstract

Global navigation satellite systems (GNSS) face significant challenges in urban and sub-urban areas due to non-line-of-sight (NLOS) propagation, multipath effects, and low received power levels, resulting in highly non-linear and non-Gaussian measurement error distributions. In light of this, conventional model-based positioning approaches, which rely on Gaussian error approximations, struggle to achieve precise localization under these conditions. To overcome these challenges, we put forth a novel learning-based framework, PC-DeepNet, that employs a permutation-invariant (PI) deep neural network (DNN) to estimate position corrections (PC). This approach is designed to ensure robustness against changes in the number and/or order of visible satellite measurements, a common issue in GNSS systems, while leveraging NLOS and multipath indicators as features to enhance positioning accuracy in challenging urban and sub-urban environments. To validate the performance of the proposed framework, we compare the positioning error with state-of-the-art model-based and learning-based positioning methods using two publicly available datasets. The results confirm that proposed PC-DeepNet achieves superior accuracy than existing model-based and learning-based methods while exhibiting lower computational complexity compared to previous learning-based approaches.

Figures

Figures reproduced from arXiv: 2504.13990 by the authors.

Figure 1
Figure 1. The proposed PC-DeepNet to minimize the positioning error in NLOS and multipath environment. [PITH_FULL_IMAGE:figures/full_fig_p009_1.png] view at source ↗
Figure 2
Figure 2. Proposed framework (PC-DeepNet) for positioning error minimization. [PITH_FULL_IMAGE:figures/full_fig_p010_2.png] view at source ↗
Figure 3
Figure 3. Architecture of PI-DNN model to minimize the positioning error in NLOS and multipath environments. [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Test and train location map in San Francisco. [PITH_FULL_IMAGE:figures/full_fig_p016_4.png]
Figure 5
Figure 5. Figure 5: Test and train location map in Los Angeles. [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: Kendall’s tau correlation among different features including pseudorange residuals, LOS vector, GDOP, elevation angle, [PITH_FULL_IMAGE:figures/full_fig_p019_6.png]
Figure 7
Figure 7. Figure 7: Evaluation map for urban area in San Francisco. [PITH_FULL_IMAGE:figures/full_fig_p020_7.png]
Figure 8
Figure 8. Figure 8: Evaluation map for the sub-urban area in San Francisco. [PITH_FULL_IMAGE:figures/full_fig_p021_8.png]
Figure 9
Figure 9. Figure 9: Evaluation map for urban area in Los Angeles. [PITH_FULL_IMAGE:figures/full_fig_p021_9.png]
Figure 10
Figure 10. Figure 10: Time series comparison of error between different methods in scenario-I. [PITH_FULL_IMAGE:figures/full_fig_p022_10.png]
Figure 11
Figure 11. Figure 11: Time series comparison of error between different methods in scenario-II. [PITH_FULL_IMAGE:figures/full_fig_p022_11.png]
Figure 12
Figure 12. Figure 12: Time series comparison of error between different methods in scenario-III. [PITH_FULL_IMAGE:figures/full_fig_p023_12.png]
Figure 13
Figure 13. Figure 13: Comparison of horizontal positioning error among state-of-the-art methods with the proposed framework (PC-DeepNet). [PITH_FULL_IMAGE:figures/full_fig_p024_13.png]
Figure 14
Figure 14. Figure 14: Performance comparison of positioning MAE among state-of-the-art methods with the proposed framework (PC [PITH_FULL_IMAGE:figures/full_fig_p025_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 39 canonical work pages

  1. [17]

    Improving GNSS positioning using neural-network-based corrections,

    A. V . Kanhere, S. Gupta, A. Shetty, and G. Gao, “Improving GNSS positioning using neural-network-based corrections,” Navig. J. Inst. Navig. , vol. 69, no. 4, 2022

    work page 2022

  2. [1]

    Performance evaluation of GNSS positioning with geometric dilution of precision,

    M. H. Kabir, S. Lee, and W. Shin, “Performance evaluation of GNSS positioning with geometric dilution of precision,” in Proc. 13th Intl. Conf. on Information and Communication Technology Convergence (ICTC) , 2022, pp. 910–912

    work page 2022

  3. [2]

    Towards millimeter-level accuracy in GNSS-based space geodesy: a review of error budget for GNSS precise point positioning,

    X. Li, J. Barriot, Y . Lou, W. Zhang, P. Li, and C. Shi, “Towards millimeter-level accuracy in GNSS-based space geodesy: a review of error budget for GNSS precise point positioning,” Surveys in Geophysics, vol. 44, no. 6, pp. 1691–1780, 2023

    work page 2023

  4. [3]

    Ionospheric time-delay algorithm for single-frequency GPS users,

    J. A. Klobuchar, “Ionospheric time-delay algorithm for single-frequency GPS users,” IEEE Trans. aerosp. electron. syst. , no. 3, pp. 325–331, 2007

    work page 2007

  5. [4]

    Atmospheric correction for the troposphere and stratosphere in radio ranging satellites,

    J. Saastamoinen, “Atmospheric correction for the troposphere and stratosphere in radio ranging satellites,” The use of artificial satellites for geodesy , vol. 15, pp. 247–251, 1972. 30

    work page 1972

  6. [5]

    Research on improving accuracy of GPS positioning based on particle filter,

    E. Wang, W. Zhao, and M. Cai, “Research on improving accuracy of GPS positioning based on particle filter,” in Proc. IEEE 8th Conf. on Industrial Electronics and Applications (ICIEA) , 2013, pp. 1167–1171

    work page 2013

  7. [6]

    GNSS NLOS signal classification based on machine learning and pseudorange residual check,

    T. Ozeki and N. Kubo, “GNSS NLOS signal classification based on machine learning and pseudorange residual check,” Frontiers in Robotics and AI , vol. 9, pp. 1–12, May. 2022

    work page 2022

  8. [7]

    The multipath estimating delay lock loop: approaching theoretical accuracy limits,

    R. D. Van Nee, J. Siereveld, P. C. Fenton, and B. R. Townsend, “The multipath estimating delay lock loop: approaching theoretical accuracy limits,” in Proc. of IEEE Position, Location and Navigation Symposium-PLANS’94, 1994, pp. 246–251

    work page 1994

Show all 42 references
  1. [8]

    A practical approach to the reduction of pseudorange multipath errors in a L1 GPS receiver,

    B. Townsend and P. Fenton, “A practical approach to the reduction of pseudorange multipath errors in a L1 GPS receiver,” in Proc. of the 7th Intl. Technical Meeting of the Satellite Division of the Institute of Navigation, Salt Lake City, UT, USA , 1994, pp. 20–23

    1994

  2. [9]

    NLOS GPS signal detection using a dual-polarisation antenna,

    Z. Jiang and P. D. Groves, “NLOS GPS signal detection using a dual-polarisation antenna,” GPS Solut., vol. 18, pp. 15–26, 2014

    2014

  3. [10]

    Analysis of a choke ring groundplane for multipath control in global positioning system (GPS) applications,

    J. M. Tranquilla, J. Carr, and H. M. Al-Rizzo, “Analysis of a choke ring groundplane for multipath control in global positioning system (GPS) applications,” IEEE Trans. Antennas Propag. , vol. 42, no. 7, pp. 905–911, 1994

    1994

  4. [11]

    A unified framework for GPS code and carrier-phase multipath mitigation using support vector regression,

    Q.-H. Phan, S.-L. Tan, I. McLoughlin, and D.-L. Vu, “A unified framework for GPS code and carrier-phase multipath mitigation using support vector regression,” Advances in Artificial Neural Systems , vol. 2013, 2013

    2013

  5. [12]

    A robust GNSS LOS/NLOS signal classifier,

    R. Yozevitch, B. B. Moshe, and A. Weissman, “A robust GNSS LOS/NLOS signal classifier,” Navig. J. Inst. Navig., vol. 63, no. 4, pp. 429–442, 2016

    2016

  6. [13]

    A machine learning approach to detect non-line-of-sight GNSS signals in Nav2Nav,

    M. Socharoentum, H. A. Karimi, and Y . Deng, “A machine learning approach to detect non-line-of-sight GNSS signals in Nav2Nav,” in 23rd ITS World Congress , 2016, pp. 10–14

    2016

  7. [14]

    Convolutional neural network based multipath detection method for static and kinematic GPS high precision positioning,

    Y . Quan, L. Lau, G. W. Roberts, X. Meng, and C. Zhang, “Convolutional neural network based multipath detection method for static and kinematic GPS high precision positioning,” Remote Sensing, vol. 10, no. 12, p. 2052, 2018

    2018

  8. [15]

    GPS signal reception classification using adaptive neuro-fuzzy inference system,

    R. Sun, L.-T. Hsu, D. Xue, G. Zhang, and W. Y . Ochieng, “GPS signal reception classification using adaptive neuro-fuzzy inference system,” The Journal of Navigation , vol. 72, no. 3, pp. 685–701, 2019

    2019

  9. [16]

    Improving GPS code phase positioning accuracy in urban environments using machine learning,

    R. Sun, G. Wang, Q. Cheng, L. Fu, K.-W. Chiang, L.-T. Hsu, and W. Y . Ochieng, “Improving GPS code phase positioning accuracy in urban environments using machine learning,” IEEE Internet Things J. , vol. 8, no. 8, pp. 7065–7078, 2020

    2020

  10. [18]

    scipy optimize least squares,

    “scipy optimize least squares,” https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.least squares.html, 2023, [Online], Accessed: 3-April-2025

    2023

  11. [19]

    Android raw GNSS measurement datasets for precise positioning,

    G. M. Fu, M. Khider, and F. van Diggelen, “Android raw GNSS measurement datasets for precise positioning,” in Proc. of the 33rd Intl. technical meeting of the satellite division of the Institute of Navigation (ION GNSS+ 2020) , 2020, pp. 1925–1937

    2020

  12. [20]

    Google smartphone decimeter challenge 2022,

    H. Addison et al. , “Google smartphone decimeter challenge 2022,” https://www.kaggle.com/competitions/ smartphone-decimeter-2022/, 2022, [Online], Accessed: 3-April-2025

    2022

  13. [21]

    Y . J. Morton, F. van Diggelen, J. J. Spilker Jr, B. W. Parkinson, S. Lo, and G. Gao, Position, navigation, and timing technologies in the 21st century: Integrated satellite navigation, sensor systems, and civil applications, volume 1 . John Wiley & Sons, 2021

    2021

  14. [22]

    GNSS-based position estimation with Kalman filters,

    Sever, Mert and Hajiyev, Chingiz, “GNSS-based position estimation with Kalman filters,” Int. J.Sustainable Aviation, vol. 8, no. 3, pp. 181–197, 2022

    2022

  15. [23]

    Multipath analysis using code-minus-carrier for dynamic testing of GNSS receivers,

    N. Blanco-Delgado and M. U. de Haag, “Multipath analysis using code-minus-carrier for dynamic testing of GNSS receivers,” in Proc. Intl. Conf. on Loc. and GNSS (ICL-GNSS) , 2011, pp. 25–30. 31

    2011

  16. [24]

    Urban multipath detection and mitigation with dynamic 3d maps for reliable land vehicle localization,

    M. Obst, S. Bauer, and G. Wanielik, “Urban multipath detection and mitigation with dynamic 3d maps for reliable land vehicle localization,” in Proc. of the 2012 IEEE/ION position, location and navigation symposium , 2012, pp. 685–691

    2012

  17. [25]

    The iterated kalman filter update as a gauss-newton method,

    B. M. Bell and F. W. Cathey, “The iterated kalman filter update as a gauss-newton method,” IEEE Trans. Autom. Control, vol. 38, no. 2, pp. 294–297, 1993

    1993

  18. [26]

    T. D. Barfoot, State estimation for robotics . Cambridge University Press, 2017

    2017

  19. [27]

    A comparison of EKF and SGD applied to a view-based SLAM approach with omnidirectional images,

    D. Valiente, A. Gil, L. Fern ´andez, and ´O. Reinoso, “A comparison of EKF and SGD applied to a view-based SLAM approach with omnidirectional images,” Robot. Auton. Syst. , vol. 62, no. 2, pp. 108–119, 2014

    2014

  20. [28]

    Tightly coupled GNSS/INS integration via factor graph and aided by fish-eye camera,

    W. Wen, X. Bai, Y . C. Kan, and L.-T. Hsu, “Tightly coupled GNSS/INS integration via factor graph and aided by fish-eye camera,” IEEE Trans. Veh. Technol., vol. 68, no. 11, pp. 10 651–10 662, 2019

    2019

  21. [29]

    Optimization-based estimator design for vision-aided inertial navigation,

    Li, Mingyang and Mourikis, Anastasios I, “Optimization-based estimator design for vision-aided inertial navigation,” in Robotics: Science and Systems , vol. 8. Berlin Germany, 2013, pp. 241–248

    2013

  22. [30]

    It is time for Factor Graph Optimization for GNSS/INS Integration: Comparison between FGO and EKF,

    Wen, Weisong and Pfeifer, Tim and Bai, Xiwei and Hsu, Li-Ta, “It is time for Factor Graph Optimization for GNSS/INS Integration: Comparison between FGO and EKF,” arXiv preprint arXiv:2004.10572 , 2020

    2004 arXiv

  23. [31]

    GNSS-5G hybrid positioning based on multi-rate measurements fusion and proactive measurement uncertainty prediction,

    l. Bai, C. Sun, A. G. Dempster, H. Zhao, J. W. Cheong, and W. Feng, “GNSS-5G hybrid positioning based on multi-rate measurements fusion and proactive measurement uncertainty prediction,” IEEE Trans. on Instrum. Meas. , vol. 71, pp. 1–15, 2022

    2022

  24. [32]

    Combining machine learning and dynamic time wrapping for vehicle driving event detection using smartphones,

    R. Sun, Q. Cheng, F. Xie, W. Zhang, T. Lin, and W. Y . Ochieng, “Combining machine learning and dynamic time wrapping for vehicle driving event detection using smartphones,” IEEE Trans. Intell. Transp. Syst., vol. 22, no. 1, pp. 194–207, 2019

    2019

  25. [33]

    Noureldin, T

    A. Noureldin, T. B. Karamat, and J. Georgy, Fundamentals of inertial navigation, satellite-based positioning and their integration. Springer Science & Business Media, 2012

    2012

  26. [34]

    Satellite selection based on WDOP concept and convex geometry,

    N. Blanco-Delgado and F. D. Nunes, “Satellite selection based on WDOP concept and convex geometry,” in Proc. 2010 5th ESA Workshop on Satellite Navigation Technologies and European Workshop on GNSS Signals and Signal Processing (NAVITEC). IEEE, 2010, pp. 1–8

    2010

  27. [35]

    Relation between GDOP and the geometry of the satellite constellation,

    N. Blanco-Delgado, F. D. Nunes, and G. Seco-Granados, “Relation between GDOP and the geometry of the satellite constellation,” in 2011 Int. Conf. on Localization and GNSS (ICL-GNSS) . IEEE, 2011, pp. 175–180

    2011

  28. [36]

    Universal approximation of functions on sets,

    E. Wagstaff, F. B. Fuchs, M. Engelcke, M. A. Osborne, and I. Posner, “Universal approximation of functions on sets,” J. Mach. Learn. Res. , vol. 23, no. 1, pp. 6762–6817, 2022

    2022

  29. [37]

    Stochastic gradient descent tricks,

    Bottou, L ´eon, “Stochastic gradient descent tricks,” in Neural Networks: Tricks of the Trade: Second Edition . Springer, 2012, pp. 421–436

    2012

  30. [38]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  31. [39]

    An overview of gradient descent optimization algorithms,

    Ruder, Sebastian, “An overview of gradient descent optimization algorithms,” arXiv preprint arXiv:1609.04747 , 2016

    2016 arXiv

  32. [40]

    A comparison of the pearson, spearman rank and kendall tau correlation coefficients using quantitative variables,

    F. Essam, H. El, and S. R. H. Ali, “A comparison of the pearson, spearman rank and kendall tau correlation coefficients using quantitative variables,” Asian J. Probab. Stat , pp. 36–48, 2022

    2022

  33. [41]

    Geodetic inverse solution between antipodal points,

    T. Vincenty, “Geodetic inverse solution between antipodal points,” Richard Rapp Geodetic Science Ohio State University , 1975

    1975

  34. [42]

    The haversine in nautical astronomy,

    H. Goodwin, “The haversine in nautical astronomy,” in US Naval Institute Proceedings, vol. 36, no. 3, 1910, pp. 735–746

    1910

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.