Pith. sign in

REVIEW 5 major objections 5 minor 37 references

PG-LIO: Photometric-Geometric fusion for Robust LiDAR-Inertial Odometry

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read PG-LIO claims that adding photometric constraints from LiDAR intensity to a geometric-inertial factor graph keeps odometry accurate in geometrically self-similar tunnels and fields.

desk verdict A credible incremental advance over COIN-LIO with two genuinely useful changes; the tunnel headline is weaker than advertised but the paper deserves a fair review. read the letter →

arxiv 2506.18583 v1 pith:SS22Q2EF submitted 2025-06-23 cs.RO

Nikhil Khedekar , Kostas Alexis This is my paper
classification cs.RO
keywords LiDAR-InertialOdometryphotometric-geometricfusionfactorgraphoptimizationgeometricdegeneracyintensityimagenormalizedcrosscorrelationself-similarenvironmentsreal-timestateestimation
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

PG-LIO claims that a LiDAR-inertial odometry system can stay accurate in geometrically self-similar environments by treating the LiDAR's intensity return as a second sensing modality, not just reflectivity coloring for the geometric map. The paper argues that when walls, tunnels, or fields present little geometric structure for point-to-plane matching, the texture captured in the intensity image provides the missing constraints, and that this can be done in real time on a factor graph using only the LiDAR and IMU already on board. The headline evidence is a 1 km manually piloted aerial trajectory inside a self-similar bicycle tunnel, flown at 7.5 m/s on average, that ends with about 1 m of drift. In geometrically well-structured scenes the method matches the accuracy of state-of-the-art LIO, and in degenerate scenes it improves substantially on the leading intensity-fusing baseline, per the paper's reported numbers.

What carries the argument

The load-bearing piece is the normalized-cross-correlation (NCC) photometric factor, which compares whole patches of filtered intensity rather than individual pixel values, making the cost invariant to unknown affine brightness changes that violate the brightness-constancy assumption for LiDAR intensity images. It is paired with a projection model that computes a per-point azimuth bias from a lookup table, correcting for the offset between the laser head and the rotation axis that varies with measured range. These two elements give the photometric subsystem something reliable to lock onto, while the geometric factor and the IMU preintegration factor hold the system together when texture is sparse. A sliding-window factor graph smoothed incrementally keeps the whole loop real time.

What would settle it

Run PG-LIO in a geometrically self-similar environment that is also photometrically uniform, such as a plain painted tunnel or a snow-covered field, and compare end-to-end drift with a geometry-only LIO baseline; if the errors converge despite the added photometric factor, the degeneracy-robustness claim is confined to textured scenes. A controlled ablation that disables the photometric factor on the paper's own tunnel sequence would show directly how much of the 1 m result comes from intensity information.

Watch

Extended reading notes

Core claim

The central claim is that degeneracy in LiDAR-inertial odometry is a missing-data problem, and that the LiDAR's own intensity channel supplies the missing data when geometry is self-similar. The paper's contribution is a tightly coupled factor-graph estimator with three constraint types: a geometric point-to-plane factor, an IMU preintegration factor, and a photometric factor built from normalized cross correlation over intensity patches. The photometric factor is designed to tolerate the brightness inconsistency of raw LiDAR intensity, and a self-calibrating projection model corrects the range-dependent azimuth bias that otherwise misprojects points into the intensity image. The reported result is accuracy on par with state-of-the-art LIO in well-structured environments and a roughly 1 m end-to-end drift over a 1047 m tunnel flight, where geometry-only and reflectivity-only methods fail or drift far more.

Load-bearing premise

The whole robustness argument depends on the scene having some intensity texture for the photometric factor to track; in a featureless corridor, tunnel, snowfield, or field with no reflectivity variation, the photometric constraint has nothing to lock onto and the system falls back to geometry plus IMU, where degeneracy returns.

Editorial extensions

If this is right

  • In geometrically well-structured settings, PG-LIO reports accuracy comparable to state-of-the-art LIO, so adopting the photometric machinery does not degrade normal operation.
  • In self-similar environments with intensity texture, the method reports substantially lower trajectory error than the leading intensity-fusing LIO baseline on the ENWIDE sequences.
  • The tunnel experiment shows roughly 1 m drift over a 1047 m trajectory at 7.5 m/s average speed, indicating stable high-speed operation in a visually repetitive corridor.
  • Since only the LiDAR and IMU are used and the graph is smoothed over a sliding window, the approach is real-time and adds no sensor weight or power burden.
  • The planned source release would let others reproduce the tunnel result and extend the method to other platforms.

Reading between the lines

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

  • The robustness claim implicitly assumes the scene has some intensity texture; in a photometrically uniform environment such as a plain painted tunnel or snowfield, the photometric factor has nothing to lock onto and degeneracy returns.
  • The NCC-based patch factor could be reused in other LiDAR-intensity odometry pipelines because it relaxes the brightness-consistency assumption that direct photometric methods usually require.
  • The range-dependent azimuth-bias lookup projection could benefit other spinning-LiDAR intensity projection systems independently of the full PG-LIO pipeline.
  • A controlled ablation that disables the photometric factor on the paper's own tunnel sequence would quantify how much of the 1 m result comes from intensity information rather than geometry and IMU alone.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The manuscript presents PG-LIO, a tightly-coupled LiDAR-inertial odometry system that augments a geometric point-to-plane factor graph with a photometric factor computed from LiDAR intensity images. The photometric factor uses normalized cross-correlation (NCC) patch similarity rather than an explicit brightness-constancy error, and the projection model includes a range-dependent azimuth-offset correction. The system is evaluated on the Newer College and ENWIDE datasets and on a newly collected 1047 m bicycle-tunnel flight. The authors claim accuracy comparable to state-of-the-art LIO in well-structured scenes and improved accuracy over COIN-LIO in most geometrically degenerate scenarios, with a headline result of roughly 1 m drift on the tunnel trajectory.

Significance. If the results hold, PG-LIO is a useful practical advance: it mitigates geometric degeneracy by reusing information the LiDAR already provides, without adding cameras or other sensors. The two technical novelties are clearly described: a range-corrected azimuth projection model (Section III-F.2) and a differentiable NCC factor (Section III-F.4) with explicit Jacobians. The ENWIDE comparison shows large and consistent gains over purely geometric baselines and modest gains over COIN-LIO in 8 of 10 sequences. The promised open-source release is a concrete benefit to the community. However, the headline tunnel figure is supported only by an endpoint-return measurement and no ablation isolates the photometric factor from the projection correction, so the significance of the central claim is currently conditional on additional evidence.

major comments (5)
  1. [IV-C] The abstract and Section IV-C claim only about 1 m drift over the 1047 m tunnel trajectory, but the only quantitative support is that the vehicle returns to the starting location and the accumulated point clouds of PG-LIO and COIN-LIO are compared qualitatively. No surveyed ground-truth trajectory (total station, surveyed markers, or post-processed RTK) and no full-trajectory error metric are reported. For a round-trip trajectory, large errors on the outward and return legs can cancel at the endpoint, so the 1 m figure does not by itself establish robustness along the flight. Please report a ground-truth-based ATE/RE, or at least per-leg errors and a drift-versus-time plot.
  2. [III-F / IV-B] The two contributions relative to COIN-LIO are the corrected projection model (Eq. (16)) and the NCC-based photometric factor (Eqs. (20)-(25)). The evaluation tables compare the full system only; no ablation separates these changes. Because the projection correction changes which points project to which pixels and can itself improve geometric consistency, the central claim that photometric-geometric fusion, rather than the better projection model, provides degeneracy robustness is not yet established. An ablation on the degenerate ENWIDE sequences (geometry-only, geometry plus corrected projection, geometry plus NCC without the projection correction, and full PG-LIO) is needed.
  3. [IV (Tables I and II)] The baseline numbers for all compared methods are taken from [12] rather than recomputed with the same evaluation harness; Section IV states that the values are as reported by [12]. This is acceptable only if the evaluation settings (evo options, trajectory alignment, start and end times) are identical, and the paper does not confirm this. In addition, each reported number is a single run with no variance or repeated-trials information, which matters for the small margins over COIN-LIO in several ENWIDE sequences. Please clarify the protocol and, where feasible, recompute the baselines or provide a reproducibility statement.
  4. [IV-C] The tunnel experiment does not cover a texture-free degenerate environment: Section IV-C explicitly says the walls contain a mix of plain and decorative mural sections that provide textural cues in the intensity image. The photometric factor therefore has informative signal precisely where geometry is uninformative. The robustness claim should be scoped to geometrically self-similar but photometrically textured environments; plain tunnels, snowfields, and uniform fields are not demonstrated, and in those scenes PG-LIO reduces to geometry-plus-IMU and degeneracy can return. Please state this limitation explicitly and, ideally, evaluate a texture-free segment.
  5. [IV-B] The two acknowledged unfavorable cases, TunnelS and RunwayD, are in the target regime of the paper's central claim. In TunnelS the estimator does not converge to the correct gravity and accelerometer bias, and in RunwayD a single agile maneuver breaks the map. The current one-sentence explanations lack quantitative diagnostics (state/error plots, timing of the divergence, sensitivity to the new factors). The table also shows that the comparison is mixed in these cases: better ATE but worse RE in TunnelS, and worse ATE but better RE in RunwayD. A discussion of whether these cases reveal systematic limits of the photometric fusion, rather than isolated incidents, would strengthen the robustness claim.
minor comments (5)
  1. [Eq. (12)] The projection model uses negative focal lengths (fx = -w/2π, fy = -h/θ_fov); the sign convention and image-axis orientation should be explained for reproducibility.
  2. [Eq. (25)] The function Γ(·) is used before it is defined; please define it as the per-point homogeneous transformation (including T_IL, T_WI_e, and the per-point deskew transform T_LeL_t) and state the dimensions of each term.
  3. [Tables I and II] The × notation for failed methods is not defined in the captions, and in some rows an RE value is still printed next to a × ATE; please define what × means and how the RE was computed for failed runs.
  4. [Abstract / repository URL] The github URL in the abstract points to a repository named mimosa, which is not obviously PG-LIO; please align the repository name or add a description so the code release is discoverable.
  5. [Section III-E] The map-update thresholds (2 m, 30 degrees), the subsampling limits (20 points, 0.1 m), and the plane-correspondence thresholds d1 and d2 are stated without any sensitivity analysis or a sentence on how they were chosen; a brief note would aid reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: PG-LIO's photometric-geometric fusion claim is empirically benchmarked against external datasets and does not reduce to its inputs by construction.

full rationale

PG-LIO's central claim—that fusing photometric and geometric LiDAR cues with IMU preintegration improves robustness in geometrically degenerate environments—is an empirical claim tested on Newer College, ENWIDE, and a new tunnel dataset against external baselines. The photometric factor is derived from NCC (Eqs. 17–24) and a projection model (Eqs. 12–16) with stated sensor geometry; neither definition encodes the accuracy result. Baseline numbers in Tables I and II are taken from [12] or measured externally (Leica total station ground truth; evo evaluation). No fitted parameter is renamed as a prediction, and no self-citation is load-bearing: references to the authors' prior work [3], [4], [10], [37] concern platform and system context, not the derivation of the photometric or geometric factors. The tunnel result is an endpoint-return measurement and may be weak evidence for mid-trajectory robustness, and the lack of an ablation separating the projection-model correction from the NCC factor limits attribution of the improvement, but these are evidentiary or correctness concerns, not circularity: the robustness claim does not reduce to its inputs by construction. No circular step can be quoted from the paper.

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

The central claim rests on standard LIO machinery plus a sensor-specific projection model and an environmental assumption that intensity texture exists where geometry degenerates. The free parameters are hand-set thresholds common in SLAM tuning; none are fit to the benchmark trajectories in a way that would manufacture the result, but several are unreported. No new physical entities are introduced.

free parameters (6)
  • Plane correspondence thresholds d1, d2 = not reported
    Used in Section III-Ec to accept or reject geometric point-to-plane correspondences. The omitted values affect which constraints enter the factor graph.
  • Map update translation threshold = 2 m
    Section III-Ed: the map is updated when the current pose is more than 2 m from any pose used to update the map. Chosen by hand.
  • Map update rotation threshold = 30 degrees roll/pitch
    Section III-Ed: the map is also updated on significant roll or pitch variation. Chosen by hand.
  • Subsampling voxel constraints = 20 points maximum, 0.1 m minimum spacing
    Section III-Eb: point density reduction thresholds chosen to balance real-time operation and preservation of edge structure.
  • Static initialization duration = 0.5 s, corresponding to 50 IMU samples at 100 Hz
    Section III-C: the system assumes a static start for bias and gravity direction initialization. The TunnelS failure is traced to gravity and accelerometer bias non-convergence.
  • Patch selection and NCC outlier thresholds = not reported
    Section III-F5 and III-F6: high-gradient patch scoring and the minimum ENCC outlier rejection threshold are not given numeric values, though they determine which photometric constraints survive.
assumptions (6)
  • domain assumption IMU measurements are constant angular velocity and linear acceleration over each sample period.
    Used in Eq. 6 for state propagation, deskewing, and preintegration, inherited from [25]. If violated during aggressive maneuvers, the propagated state and deskew degrade.
  • domain assumption LiDAR and IMU are synchronized with a fixed, apriori calibrated extrinsic transform T_IL.
    Section III-A: all fusion assumes this transform; no online calibration is provided.
  • domain assumption The system is initially static for Ns=0.5 s, so W=I0 and the initial gravity direction equals the mean accelerometer direction.
    Section III-C: initialization procedure. The TunnelS failure is attributed to gravity and accelerometer bias non-convergence, showing this premise is load-bearing.
  • domain assumption The Ouster OS0-128 projection geometry has a laser origin offset n and per-laser azimuth offsets theta_a in the range [-11, +11] degrees, as given by the manufacturer documentation.
    Section III-F2, Eqs. 13-16: the lookup-table projection correction assumes this sensor geometry and known theta_a per laser. No per-sensor estimation procedure is described.
  • domain assumption Degenerate directions can be identified from the projected localizability of the geometric Jacobian and supplemented by intensity patches.
    Section III-Ec and III-F: patch selection uses the eigenvectors of the translation block of J^T J and assumes intensity constrains those degenerate directions.
  • domain assumption Normalized cross-correlation is an appropriate similarity measure under unknown affine brightness changes.
    Section III-F3 and III-F4: this motivates replacing brightness constancy with the NCC factor. If intensity noise exceeds the affine model, the photometric factor loses information.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PG-LIO: Photometric-Geometric fusion for Robust LiDAR-Inertial Odometry." pith.science (2026). https://pith.science/paper/SS22Q2EF

@misc{pith2026250618583,
  author       = {Pith},
  title        = {Pith review of: PG-LIO: Photometric-Geometric fusion for Robust LiDAR-Inertial Odometry},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SS22Q2EF}},
  note         = {Machine review of arXiv:2506.18583}
}
read the original abstract

LiDAR-Inertial Odometry (LIO) is widely used for accurate state estimation and mapping which is an essential requirement for autonomous robots. Conventional LIO methods typically rely on formulating constraints from the geometric structure sampled by the LiDAR. Hence, in the lack of geometric structure, these tend to become ill-conditioned (degenerate) and fail. Robustness of LIO to such conditions is a necessity for its broader deployment. To address this, we propose PG-LIO, a real-time LIO method that fuses photometric and geometric information sampled by the LiDAR along with inertial constraints from an Inertial Measurement Unit (IMU). This multi-modal information is integrated into a factor graph optimized over a sliding window for real-time operation. We evaluate PG-LIO on multiple datasets that include both geometrically well-conditioned as well as self-similar scenarios. Our method achieves accuracy on par with state-of-the-art LIO in geometrically well-structured settings while significantly improving accuracy in degenerate cases including against methods that also fuse intensity. Notably, we demonstrate only 1 m drift over a 1 km manually piloted aerial trajectory through a geometrically self-similar tunnel at an average speed of 7.5m/s (max speed 10.8 m/s). For the benefit of the community, we shall also release our source code https://github.com/ntnu-arl/mimosa

Figures

Figures reproduced from arXiv: 2506.18583 by the authors.

Figure 1
Figure 1. Architecture of the maintained factor graph. Factors are represented [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 3
Figure 3. Variation of projected pixel location u with the sampled range r for the projection model in Equation 12. bias values to correct the projected locations to be exactly at the PBID locations. Different from CPBRA this accounts for the sampled range. The final projection function is then given by  u v  = Πn(pLt ) = Π(pLt ) +  bu 0  (16) where bu is the bias offset obtained from the generated lookup table. In case o… view at source ↗
Figure 4
Figure 4. Top-Down Orthographic view of the accumulated registered pointcloud for the [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: Qualitative comparison demonstrating our improved accuracy over COIN-LIO [12] on the [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Comparison against COIN-LIO [12] in a 500 m long geometrically self-similar section of a bicycle tunnel collected from an aerial robot manually [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

37 extracted references · 30 canonical work pages

  1. [12]

    Coin-lio: Complementary intensity-augmented lidar inertial odometry,

    P. Pfreundschuh, H. Oleynikova, C. Cadena, R. Siegwart, and O. An- dersson, “Coin-lio: Complementary intensity-augmented lidar inertial odometry,” in 2024 IEEE International Conference on Robotics and Automation (ICRA) , 2024, pp. 1730–1737

    work page 2024

  2. [1]

    Present and future of slam in extreme environments: The darpa subt challenge,

    K. Ebadi, L. Bernreiter, H. Biggie, G. Catt, Y . Chang, A. Chatterjee, C. E. Denniston, S.-P. Desch ˆenes, K. Harlow, S. Khattak, L. Nogueira, M. Palieri, P. Petr´aˇcek, M. Petrl´ık, A. Reinke, V . Kr´atk´y, S. Zhao, A.-a. Agha-mohammadi, K. Alexis, C. Heckman, K. Khosoussi, N. Kottege, B. Morrell, M. Hutter, F. Pauling, F. Pomerleau, M. Saska, S. Scherer...

    work page 2024

  3. [2]

    A method for registration of 3-D shapes,

    P. Besl and N. D. McKay, “A method for registration of 3-D shapes,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 14, no. 2, pp. 239–256, Feb. 1992

    work page 1992

  4. [3]

    MIMOSA: A Multi-Modal SLAM Framework for Resilient Autonomy against Sensor Degradation,

    N. Khedekar, M. Kulkarni, and K. Alexis, “MIMOSA: A Multi-Modal SLAM Framework for Resilient Autonomy against Sensor Degradation,” in 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). Kyoto, Japan: IEEE, Oct. 2022, pp. 7153–7159

    work page 2022

  5. [4]

    Comple- mentary Multi–Modal Sensor Fusion for Resilient Robot Pose Estima- tion in Subterranean Environments,

    S. Khattak, H. Nguyen, F. Mascarich, T. Dang, and K. Alexis, “Comple- mentary Multi–Modal Sensor Fusion for Resilient Robot Pose Estima- tion in Subterranean Environments,” in 2020 International Conference on Unmanned Aircraft Systems (ICUAS) . Athens, Greece: IEEE, Sep. 2020, pp. 1024–1029

    work page 2020

  6. [5]

    Super odometry: Imu-centric lidar-visual-inertial estimator for challenging en- vironments,

    S. Zhao, H. Zhang, P. Wang, L. Nogueira, and S. Scherer, “Super odometry: Imu-centric lidar-visual-inertial estimator for challenging en- vironments,” in 2021 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , 2021, pp. 8729–8736

    work page 2021

  7. [6]

    LVI-SAM: Tightly-coupled Lidar-Visual-Inertial Odometry via Smoothing and Mapping,

    T. Shan, B. Englot, C. Ratti, and D. Rus, “LVI-SAM: Tightly-coupled Lidar-Visual-Inertial Odometry via Smoothing and Mapping,” in 2021 IEEE International Conference on Robotics and Automation (ICRA) . Xi’an, China: IEEE, May 2021, pp. 5692–5698

    work page 2021

  8. [7]

    LIVER: A Tightly Coupled LiDAR-Inertial-Visual State Estimator With High Robustness for Underground Environments,

    T. Wen, Y . Fang, B. Lu, X. Zhang, and C. Tang, “LIVER: A Tightly Coupled LiDAR-Inertial-Visual State Estimator With High Robustness for Underground Environments,” IEEE Robotics and Automation Letters , vol. 9, no. 3, pp. 2399–2406, Mar. 2024

    work page 2024

Show all 37 references
  1. [8]

    Fast-livo2: Fast, direct lidar-inertial-visual odometry,

    C. Zheng, W. Xu, Z. Zou, T. Hua, C. Yuan, D. He, B. Zhou, Z. Liu, J. Lin, F. Zhu, Y . Ren, R. Wang, F. Meng, and F. Zhang, “Fast-livo2: Fast, direct lidar-inertial-visual odometry,” 2024. [Online]. Available: https://arxiv.org/abs/2408.14035

    2024 arXiv

  2. [9]

    Informed, constrained, aligned: A field analysis on degeneracy-aware point cloud registration in the wild,

    T. Tuna, J. Nubert, P. Pfreundschuh, C. Cadena, S. Khattak, and M. Hutter, “Informed, constrained, aligned: A field analysis on degeneracy-aware point cloud registration in the wild,” 2025. [Online]. Available: https://arxiv.org/abs/2408.11809

    2025 arXiv

  3. [10]

    Degradation resilient lidar- radar-inertial odometry,

    M. Nissov, N. Khedekar, and K. Alexis, “Degradation resilient lidar- radar-inertial odometry,” in 2024 IEEE International Conference on Robotics and Automation (ICRA) , 2024, pp. 8587–8594

    2024

  4. [11]

    Garlio: Gravity enhanced radar-lidar-inertial odometry,

    C. Noh, W. Yang, M. Jung, S. Jung, and A. Kim, “Garlio: Gravity enhanced radar-lidar-inertial odometry,” in Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) , Atlanta, May. 2025

    2025

  5. [13]

    Lidar odometry survey: recent advancements and remaining challenges,

    D. Lee, M. Jung, W. Yang, and A. Kim, “Lidar odometry survey: recent advancements and remaining challenges,” Intelligent Service Robotics , vol. 17, no. 2, pp. 95–118, 2024

    2024

  6. [14]

    Fast Sparse LiDAR Odometry Using Self-Supervised Feature Selection on Intensity Images,

    T. Guadagnino, X. Chen, M. Sodano, J. Behley, G. Grisetti, and C. Stachniss, “Fast Sparse LiDAR Odometry Using Self-Supervised Feature Selection on Intensity Images,” IEEE Robotics and Automation Letters, vol. 7, no. 3, pp. 7597–7604, Jul. 2022

    2022

  7. [15]

    Md-slam: Multi-cue direct slam,

    L. Di Giammarino, L. Brizi, T. Guadagnino, C. Stachniss, and G. Grisetti, “Md-slam: Multi-cue direct slam,” in 2022 IEEE/RSJ In- ternational Conference on Intelligent Robots and Systems (IROS) , 2022, pp. 11 047–11 054

    2022

  8. [16]

    Real-Time Simultaneous Localization and Mapping with LiDAR Intensity,

    W. Du and G. Beltrame, “Real-Time Simultaneous Localization and Mapping with LiDAR Intensity,” in 2023 IEEE International Conference on Robotics and Automation (ICRA) . London, United Kingdom: IEEE, May 2023, pp. 4164–4170

    2023

  9. [17]

    An Intensity-Augmented LiDAR- Inertial SLAM for Solid-State LiDARs in Degenerated Environments,

    H. Li, B. Tian, H. Shen, and J. Lu, “An Intensity-Augmented LiDAR- Inertial SLAM for Solid-State LiDARs in Degenerated Environments,” IEEE Transactions on Instrumentation and Measurement , vol. 71, pp. 1–10, 2022

    2022

  10. [18]

    Loam: Lidar odometry and mapping in real- time,

    J. Zhang and S. Singh, “Loam: Lidar odometry and mapping in real- time,” in Proceedings of Robotics: Science and Systems , Berkeley, USA, July 2014

    2014

  11. [19]

    Fast-lio2: Fast direct lidar- inertial odometry,

    W. Xu, Y . Cai, D. He, J. Lin, and F. Zhang, “Fast-lio2: Fast direct lidar- inertial odometry,” IEEE Transactions on Robotics , vol. 38, no. 4, pp. 2053–2073, 2022

    2022

  12. [20]

    RI-LIO: Reflectivity Image Assisted Tightly-Coupled LiDAR-Inertial Odometry,

    Y . Zhang, Y . Tian, W. Wang, G. Yang, Z. Li, F. Jing, and M. Tan, “RI-LIO: Reflectivity Image Assisted Tightly-Coupled LiDAR-Inertial Odometry,” IEEE Robotics and Automation Letters , vol. 8, no. 3, pp. 1802–1809, Mar. 2023

    2023

  13. [21]

    iSAM2: Incremental smoothing and mapping using the Bayes tree,

    M. Kaess, H. Johannsson, R. Roberts, V . Ila, J. J. Leonard, and F. Dellaert, “iSAM2: Incremental smoothing and mapping using the Bayes tree,” The International Journal of Robotics Research , vol. 31, no. 2, pp. 216–235, Feb. 2012

    2012

  14. [22]

    Visual-Inertial-Aided Navigation for High- Dynamic Motion in Built Environments Without Initial Conditions,

    T. Lupton and S. Sukkarieh, “Visual-Inertial-Aided Navigation for High- Dynamic Motion in Built Environments Without Initial Conditions,” IEEE Transactions on Robotics , vol. 28, no. 1, pp. 61–76, Feb. 2012

    2012

  15. [23]

    Spherical harmonic modelling to ultra-high degree of bouguer and isostatic anomalies,

    G. Balmino, N. Vales, S. Bonvalot, and A. Briais, “Spherical harmonic modelling to ultra-high degree of bouguer and isostatic anomalies,” Journal of Geodesy , vol. 86, no. 7, pp. 499–520, 2012

    2012

  16. [24]

    Faster-LIO: Lightweight Tightly Coupled Lidar-Inertial Odometry Using Parallel Sparse Incremental V oxels,

    C. Bai, T. Xiao, Y . Chen, H. Wang, F. Zhang, and X. Gao, “Faster-LIO: Lightweight Tightly Coupled Lidar-Inertial Odometry Using Parallel Sparse Incremental V oxels,” IEEE Robotics and Automation Letters , vol. 7, no. 2, pp. 4861–4868, Apr. 2022

    2022

  17. [25]

    Imu preinte- gration on manifold for efficient visual-inertial maximum-a-posteriori estimation,

    C. Forster, L. Carlone, F. Dellaert, and D. Scaramuzza, “Imu preinte- gration on manifold for efficient visual-inertial maximum-a-posteriori estimation,” in Proceedings of Robotics: Science and Systems , Rome, Italy, July 2015

    2015

  18. [26]

    KISS-ICP: In Defense of Point-to-Point ICP – Simple, Accurate, and Robust Registration If Done the Right Way,

    I. Vizzo, T. Guadagnino, B. Mersch, L. Wiesmann, J. Behley, and C. Stachniss, “KISS-ICP: In Defense of Point-to-Point ICP – Simple, Accurate, and Robust Registration If Done the Right Way,” IEEE Robotics and Automation Letters , vol. 8, no. 2, pp. 1029–1036, Feb. 2023

    2023

  19. [27]

    Robust Estimation of a Location Parameter,

    P. J. Huber, “Robust Estimation of a Location Parameter,” The Annals of Mathematical Statistics , vol. 35, no. 1, pp. 73 – 101, 1964. [Online]. Available: https://doi.org/10.1214/aoms/1177703732

    1964

  20. [28]

    Ouster Sensor Documentation,

    “Ouster Sensor Documentation,” https://static.ouster.dev/sensor-docs/ image route1/image route2/sensor data/sensor-data.html, [Accessed 11-04-2025]

    2025

  21. [29]

    Towards intensity-augmented SLAM with LiDAR and ToF sensors,

    R. A. Hewitt and J. A. Marshall, “Towards intensity-augmented SLAM with LiDAR and ToF sensors,” in 2015 IEEE/RSJ International Confer- ence on Intelligent Robots and Systems (IROS) . Hamburg, Germany: IEEE, Sep. 2015, pp. 1956–1961

    2015

  22. [30]

    Reflectivity is all you need!: Advancing lidar semantic segmentation,

    K. Viswanath, P. Jiang, and S. Saripalli, “Reflectivity is all you need!: Advancing lidar semantic segmentation,” 2024. [Online]. Available: https://arxiv.org/abs/2403.13188

    2024 arXiv

  23. [31]

    Large scale photometric bundle adjustment,

    O. J. Woodford and E. Rosten, “Large scale photometric bundle adjustment,” in Proceedings of the 31st British Machine Vision Conference (BMVC) . BMV A Press, 2020. [Online]. Available: https://www.bmvc2020-conference.com/assets/papers/0822.pdf

    2020

  24. [32]

    Szeliski, Computer Vision: Algorithms and Applications , ser

    R. Szeliski, Computer Vision: Algorithms and Applications , ser. Texts in Computer Science. Cham: Springer International Publishing, 2022

    2022

  25. [33]

    Least squares normalized cross correlation,

    O. J. Woodford, “Least squares normalized cross correlation,” 2022. [Online]. Available: https://arxiv.org/abs/1810.04320

    2022 arXiv

  26. [34]

    LIO- SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping,

    T. Shan, B. Englot, D. Meyers, W. Wang, C. Ratti, and D. Rus, “LIO- SAM: Tightly-coupled Lidar Inertial Odometry via Smoothing and Mapping,” in 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . Las Vegas, NV , USA: IEEE, Oct. 2020, pp. 5135–5142

    2020

  27. [35]

    evo: Python package for the evaluation of odometry and slam

    M. Grupp, “evo: Python package for the evaluation of odometry and slam.” https://github.com/MichaelGrupp/evo, 2017

    2017

  28. [36]

    Multi-Camera LiDAR Inertial Extension to the Newer College Dataset,

    L. Zhang, M. Camurri, and M. Fallon, “Multi-Camera LiDAR Inertial Extension to the Newer College Dataset,” arXiv:2112.08854 [cs] , Dec. 2021

    2021 arXiv

  29. [37]

    RMF-Owl: A Collision-Tolerant Flying Robot for Autonomous Subterranean Exploration,

    P. D. Petris, H. Nguyen, M. Dharmadhikari, M. Kulkarni, N. Khedekar, F. Mascarich, and K. Alexis, “RMF-Owl: A Collision-Tolerant Flying Robot for Autonomous Subterranean Exploration,” in 2022 International Conference on Unmanned Aircraft Systems (ICUAS) . Dubrovnik, Croatia: I...

    2022

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.