Pith. sign in

REVIEW 2 major objections 1 minor 73 references

Generalized-CVO: Fast and Correspondence-Free Local Point Cloud Registration with Second Order Riemannian Optimization

T0 review · 2 major / 1 minor · reviewed 2026-06-27 · grok-4.3

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

Pith's one-line read Point clouds encoded as continuous functions via point-wise anisotropic kernels permit second-order Riemannian optimization that cuts LiDAR drift more than 55 percent while running up to ten times faster than first-order RKHS solvers.

desk verdict This paper introduces anisotropic RKHS kernels plus approximate second-order Riemannian optimization for correspondence-free point cloud registration and reports over 55% drift reduction plus 10x speedup on LiDAR tracking. read the letter →

arxiv 2606.10019 v1 pith:PZDA2DJS submitted 2026-06-08 cs.CV cs.AIcs.RO

Ray Zhang , Marcus Greiff , Thomas Lew , John Subosits This is my paper · ORCID
classification cs.CVcs.AIcs.RO
keywords pointcloudregistrationRKHSembeddingsRiemannianoptimizationLiDARtrackingcorrespondence-freeanisotropickernelssecond-ordersurfacegeometryencoding
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

The paper introduces a correspondence-free registration technique that models each point cloud as a continuous function in a reproducing kernel Hilbert space, with kernels shaped anisotropically to reflect local surface geometry. This encoding tightens alignment along surface normals while loosening it along tangential directions, which the authors argue yields more stable solutions in feature-sparse scenes. The resulting optimization problem is solved by a second-order on-manifold method that uses approximate Riemannian Hessians, delivering the reported speed-up and accuracy gains. Experiments on driving-domain LiDAR sequences, indoor RGB-D streams, and object benchmarks show lower drift and greater robustness than both prior RKHS methods and ICP variants, especially under moderate initial misalignment.

What carries the argument

Second-order on-manifold optimization with approximate Riemannian Hessians applied to the RKHS embedding formed by point-wise anisotropic kernels.

What would settle it

A dataset of feature-sparse LiDAR pairs on which the new method produces higher translational or rotational drift than the first-order RKHS baseline under identical initialization and kernel parameters.

Watch

Extended reading notes

Core claim

Representing point clouds as continuous functions with point-wise anisotropic kernels that encode local geometry allows the registration objective to emphasize normal alignment; solving this objective with second-order Riemannian optimization on the manifold produces both higher accuracy and up to a tenfold reduction in computation time relative to first-order solvers.

Load-bearing premise

The point-wise anisotropic kernels sufficiently capture local surface geometry so that relaxing tangential alignment while tightening normal alignment improves registration without introducing new failure modes.

Editorial extensions

If this is right

  • Frame-to-frame LiDAR and RGB-D tracking accuracy improves across diverse indoor and outdoor datasets.
  • Translational and rotational drift each drop more than 55 percent on driving-domain LiDAR tracking in challenging feature-sparse environments.
  • Registration runs up to ten times faster than the first-order solvers used in earlier correspondence-free RKHS methods.
  • Object registration becomes more robust than ICP-based methods, with additional gains when refining global initializations under moderate misalignment.

Reading between the lines

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

  • If the normal-versus-tangential emphasis generalizes, the same kernel construction could be applied to other surface-based alignment problems such as medical-image registration where tangential features are unreliable.
  • The computational saving might allow real-time systems to maintain higher registration rates without increasing drift accumulation in long trajectories.
  • Because the method avoids explicit correspondences, it could serve as a drop-in refinement stage after global methods that already supply moderate initial alignments.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The paper proposes Generalized-CVO, a correspondence-free local point cloud registration method that represents point clouds as continuous functions via point-wise anisotropic kernels in an RKHS to encode local surface geometry. This formulation prioritizes alignment along normals while relaxing tangential directions. Registration is solved via second-order Riemannian optimization on the manifold using approximate Hessians, yielding claimed speedups of up to 10x over first-order RKHS solvers. Experiments report >55% reduction in both translational and rotational drift on LiDAR frame-to-frame tracking in feature-sparse driving environments, plus improved robustness versus ICP on object registration benchmarks and gains when refining global initializations.

Significance. If the empirical results hold under scrutiny, the work offers a practical advance for real-time registration in sparse environments by combining geometric kernel priors with efficient second-order manifold optimization. The approach extends prior RKHS correspondence-free methods and could impact robotics and vision pipelines requiring fast, robust local alignment without explicit correspondences.

major comments (2)
  1. [Abstract] Abstract: the central claim of >55% reduction in translational and rotational drift (and up to 10x speedup) is presented without error bars, number of trials/sequences, dataset identifiers, or ablation results; this is load-bearing for the headline empirical contribution and prevents assessment of post-hoc selection or variance.
  2. [Experimental Results] Experimental section (assumed §5 or equivalent): the manuscript must supply concrete dataset details, baseline implementations, and separate ablations isolating the anisotropic kernel contribution versus the second-order solver to substantiate that the reported gains are not artifacts of the chosen evaluation protocol.
minor comments (1)
  1. [Abstract] Abstract: the description of the continuous-function representation would benefit from a one-sentence clarification of how the point-wise anisotropic kernels are constructed from local surface estimates.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We agree that the headline empirical claims require stronger supporting details and will revise the manuscript to address both major comments.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim of >55% reduction in translational and rotational drift (and up to 10x speedup) is presented without error bars, number of trials/sequences, dataset identifiers, or ablation results; this is load-bearing for the headline empirical contribution and prevents assessment of post-hoc selection or variance.

    Authors: We acknowledge the abstract presents the >55% drift reduction and 10x speedup without supporting statistics. In the revision we will add explicit references to the number of sequences/trials, dataset identifiers, and error bars (or direct the reader to the experimental section for these quantities). Full per-sequence results with variance will be highlighted in the main text to allow assessment of robustness. revision: yes

  2. Referee: [Experimental Results] Experimental section (assumed §5 or equivalent): the manuscript must supply concrete dataset details, baseline implementations, and separate ablations isolating the anisotropic kernel contribution versus the second-order solver to substantiate that the reported gains are not artifacts of the chosen evaluation protocol.

    Authors: We agree that concrete details and targeted ablations are needed. The revised experimental section will list exact dataset identifiers and sequences, describe baseline implementations (including parameter settings for ICP and prior RKHS methods), and add separate ablations that isolate the anisotropic kernel from the second-order Riemannian solver. These changes will clarify the source of the observed speedups and accuracy gains. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity detected

full rationale

The provided abstract and context describe a registration method that represents point clouds via continuous functions with anisotropic kernels in RKHS, then applies second-order Riemannian optimization. No equations are shown that reduce the claimed drift reduction or speedup to fitted parameters or self-citations by construction. The formulation is presented as an extension of prior RKHS work with an independent optimization contribution, and the empirical gains are reported as measured outcomes on external datasets rather than tautological outputs of the inputs. No self-definitional, fitted-prediction, or load-bearing self-citation patterns appear in the text.

Assumptions & free parameters 0 free parameters · 1 assumptions · 0 invented entities

Abstract-only review supplies insufficient detail to enumerate concrete free parameters or invented entities; the approach relies on standard manifold optimization and RKHS theory whose parameters are not specified here.

assumptions (1)
  • standard math Rigid transformations form a Riemannian manifold on which second-order optimization is well-defined
    Invoked by the proposed on-manifold solver

how reviews work

0 comments
Cite this review

Pith. "Pith review of Generalized-CVO: Fast and Correspondence-Free Local Point Cloud Registration with Second Order Riemannian Optimization." pith.science (2026). https://pith.science/paper/PZDA2DJS

@misc{pith2026260610019,
  author       = {Pith},
  title        = {Pith review of: Generalized-CVO: Fast and Correspondence-Free Local Point Cloud Registration with Second Order Riemannian Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PZDA2DJS}},
  note         = {Machine review of arXiv:2606.10019}
}
abstract

We propose a fast and correspondence-free local point cloud registration method that leverages geometric surface structure and reproducing kernel Hilbert space (RKHS) embeddings. The method represents point clouds as continuous functions with point-wise anisotropic kernels that encode local geometry. This formulation improves alignment along surface normals while relaxing alignment along tangential directions. To solve the resulting registration problem, we propose a second-order on-manifold optimization scheme with approximate Riemannian Hessians, achieving a speedup of up to 10x over the first-order solvers used in prior correspondence-free RKHS-based methods. We demonstrate improved frame-to-frame LiDAR and RGB-D tracking accuracy across diverse indoor and outdoor datasets. On a LiDAR tracking registration task in the driving domain, we achieve a reduction of $>55\%$ in both translational and rotational drift in challenging feature-sparse environments. On object registration benchmarks, we show improved robustness over ICP-based methods and further gains when refining global initialization, particularly under moderate misalignment.

Figures

Figures reproduced from arXiv: 2606.10019 by the authors.

Figure 1
Figure 1. The proposed correspondence-free point cloud regis [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Raw point cloud Z (left), and resampled point cloud (right) in red if Σ(zi;Z¯) has two small eigenvalues (edge) and blue if Σ(zi;Z¯) has one small eigenvalue (surface). sparse and noisy: the kernel attenuates components aligned with high-variance normal directions, which can be detri￾mental to registration performance. To regularize this degeneracy, we impose upper bounds on the dominant eigenmodes of Σ(x, z), const… view at source ↗
Figure 3
Figure 3. G-CVO versus baselines on KITTI sequence [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Ground truth and frame-to-frame tracking with G-CVO [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 6
Figure 6. Figure 6: Normalized Frobenius error with the 1st/2nd-order opti [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 8
Figure 8. Figure 8: Convergence time (IO/downsampling excluded) vs accu￾racy. Accuracy in KITTI metrics at N = {4, 8, 16} · 103 points [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: The RKHS inner product with translational and rotational perturbations. Each line corresponds to a different input pair. [PITH_FULL_IMAGE:figures/full_fig_p014_9.png]
Figure 10
Figure 10. Figure 10: Illustration of the Dirt Track dataset (top), Race Track dataset (middle), and Skid Pad dataset (bottom), with lidar point clouds and images at two points shown at two points along the track [PITH_FULL_IMAGE:figures/full_fig_p014_10.png]
Figure 11
Figure 11. Figure 11: The stacked point clouds using the frame-to-frame tracking poses from KITTI LiDAR sequence [PITH_FULL_IMAGE:figures/full_fig_p015_11.png]
Figure 12
Figure 12. Figure 12: The stacked point clouds using the frame-to-frame tracking poses from ETH3D RGB-D sequence [PITH_FULL_IMAGE:figures/full_fig_p016_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

73 extracted references · 1 canonical work pages

  1. [1]

    Princeton Univer- sity Press, 2008

    P-A Absil, Robert Mahony, and Rodolphe Sepulchre.Opti- mization algorithms on matrix manifolds. Princeton Univer- sity Press, 2008

    2008

  2. [2]

    Convergence of itera- tively re-weighted least squares to robust M-estimators

    Khurrum Aftab and Richard Hartley. Convergence of itera- tively re-weighted least squares to robust M-estimators. In 2015 IEEE Winter Conference on Applications of Computer Vision, pages 480–487, 2015

    2015

  3. [3]

    Balo: A novel point-to-plane balanced lidar odometry

    Andrea Azzini et al. Balo: A novel point-to-plane balanced lidar odometry. InProceedings of the IEEE Intelligent Vehi- cles Symposium (IV), 2025

    2025

  4. [4]

    Efficient surfel-based SLAM using 3D laser range data in urban environments

    Jens Behley and Cyrill Stachniss. Efficient surfel-based SLAM using 3D laser range data in urban environments. In Robotics: science and systems, page 59, 2018

    2018

  5. [5]

    Method for registration of 3D shapes

    Paul J Besl and Neil D McKay. Method for registration of 3D shapes. InSensor fusion IV: control paradigms and data structures, pages 586–606. Spie, 1992

    1992

  6. [6]

    Efficient and robust registration on the 3D special Euclidean group

    Uttaran Bhattacharya and Venu Madhav Govindu. Efficient and robust registration on the 3D special Euclidean group. In Int. Conf. on Computer Vision, pages 5885–5894, 2019

    2019

  7. [7]

    Springer, 2006

    Christopher M Bishop and Nasser M Nasrabadi.Pattern recognition and machine learning. Springer, 2006

    2006

  8. [8]

    Cambridge University Press, 2023

    Nicolas Boumal.An introduction to optimization on smooth manifolds. Cambridge University Press, 2023

    2023

Show all 73 references
  1. [9]

    On-manifold probabilistic iterative closest point: Application to underwa- ter karst exploration.Int

    Yohan Breux, Andr ´e Mas, and Lionel Lapierre. On-manifold probabilistic iterative closest point: Application to underwa- ter karst exploration.Int. Journal of Robotics Research, 41 (9-10):875–902, 2022

    2022

  2. [10]

    Past, present, and future of simultaneous localiza- tion and mapping: Towards the robust-perception age.IEEE Transactions on robotics, 32(6):1309–1332, 2016

    Cesar Cadena, Luca Carlone, Henry Carrillo, Yasir Latif, Davide Scaramuzza, Jose Neira, Ian Reid, John J Leonard, et al. Past, present, and future of simultaneous localiza- tion and mapping: Towards the robust-perception age.IEEE Transactions on robotics, 32(6):1309–1332, 2016

    2016

  3. [11]

    IKD-tree: An incre- mental KD-tree for robotic applications.arXiv preprint arXiv:2102.10808, 2021

    Yixi Cai, Wei Xu, and Fu Zhang. IKD-tree: An incre- mental KD-tree for robotic applications.arXiv preprint arXiv:2102.10808, 2021

    2021

  4. [12]

    An ICP variant using a point-to-line metric

    Andrea Censi. An ICP variant using a point-to-line metric. InInt. Conf. on robotics and automation, pages 19–25. Ieee, 2008

    2008

  5. [13]

    Object modelling by regis- tration of multiple range images.Image and vision comput- ing, 10(3):145–155, 1992

    Yang Chen and G ´erard Medioni. Object modelling by regis- tration of multiple range images.Image and vision comput- ing, 10(3):145–155, 1992

    1992

  6. [14]

    Sc2-pcr: A second order spatial compatibility for efficient and robust point cloud registration

    Zhi Chen, Kun Sun, Fan Yang, and Wenbing Tao. Sc2-pcr: A second order spatial compatibility for efficient and robust point cloud registration. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 13221–13231, 2022

    2022

  7. [15]

    Fully convolutional geometric features

    Christopher Choy, Jaesik Park, and Vladlen Koltun. Fully convolutional geometric features. InProceedings of the IEEE/CVF international conference on computer vision, pages 8958–8966, 2019

    2019

  8. [16]

    Non- parametric continuous sensor registration.Journal of Ma- chine Learning Research, 22(271):1–50, 2021

    William Clark, Maani Ghaffari, and Anthony Bloch. Non- parametric continuous sensor registration.Journal of Ma- chine Learning Research, 22(271):1–50, 2021

    2021

  9. [17]

    Scan registration using segmented region growing NDT.Int

    Arun Das and Steven L Waslander. Scan registration using segmented region growing NDT.Int. journal of Robotics Research, 33(13):1645–1663, 2014

    2014

  10. [18]

    IMLS-SLAM: scan-to-model matching based on 3D data

    Jean-Emmanuel Deschaud. IMLS-SLAM: scan-to-model matching based on 3D data. InInt. conf. on robotics and aut., pages 2480–2485. IEEE, 2018

    2018

  11. [19]

    One model to drift them all: Physics-informed conditional diffusion model for driving at the limits

    Franck Djeumou, Thomas Lew, Nan Ding, Michael Thomp- son, Makoto Suminaka, Marcus Greiff, and John Subosits. One model to drift them all: Physics-informed conditional diffusion model for driving at the limits. InConf. on Robot Learning, 2024

    2024

  12. [20]

    HGMR: Hi- erarchical Gaussian mixtures for adaptive 3D registration

    Benjamin Eckart, Kihwan Kim, and Jan Kautz. HGMR: Hi- erarchical Gaussian mixtures for adaptive 3D registration. In European conf. on computer vision, pages 705–721, 2018

    2018

  13. [21]

    Springer, 2020

    JEAN QUAINTANCE Gallier and Jocelyn Quaintance.Dif- ferential geometry and lie groups. Springer, 2020

    2020

  14. [22]

    Are we ready for autonomous driving? the KITTI vision benchmark suite

    Andreas Geiger, Philip Lenz, and Raquel Urtasun. Are we ready for autonomous driving? the KITTI vision benchmark suite. InConf. on Computer Vision and Pattern Recognition, 2012

    2012

  15. [23]

    Continuous direct sparse visual odometry from RGB-D images.Robotics: Science and Sys- tems, 2019

    Maani Ghaffari, William Clark, Anthony Bloch, Ryan M Eu- stice, and Jessy W Grizzle. Continuous direct sparse visual odometry from RGB-D images.Robotics: Science and Sys- tems, 2019

    2019

  16. [24]

    Approximate Gauss-Newton methods for nonlinear least squares problems.SIAM Journal on Optimization, 18(1): 106–132, 2007

    Serge Gratton, Amos S Lawless, and Nancy K Nichols. Approximate Gauss-Newton methods for nonlinear least squares problems.SIAM Journal on Optimization, 18(1): 106–132, 2007

    2007

  17. [25]

    Approximate KD tree search for efficient ICP

    Michael Greenspan and Mike Yurick. Approximate KD tree search for efficient ICP. InInt. conf. on 3D Digital Imaging and Modeling, pages 442–448. IEEE, 2003

    2003

  18. [26]

    Hierarchical optimization on manifolds for online 2d and 3d mapping

    Giorgio Grisetti, Rainer K ¨ummerle, Cyrill Stachniss, Udo Frese, and Christoph Hertzberg. Hierarchical optimization on manifolds for online 2d and 3d mapping. InInterna- tional Conference on Robotics and Automation, pages 273– 278, 2010

    2010

  19. [27]

    evo: Python package for the evaluation of odometry and SLAM.https : / / github

    Michael Grupp. evo: Python package for the evaluation of odometry and SLAM.https : / / github . com / MichaelGrupp/evo, 2017

    2017

  20. [28]

    Predator: Registration of 3d point clouds with low overlap

    Shengyu Huang, Zan Gojcic, Mikhail Usvyatsov, Andreas Wieser, and Konrad Schindler. Predator: Registration of 3d point clouds with low overlap. InProceedings of the IEEE/CVF Conference on computer vision and pattern recognition, pages 4267–4276, 2021

    2021

  21. [29]

    Robust estimation of a location parameter

    Peter J Huber. Robust estimation of a location parameter. In Breakthroughs in statistics: Methodology and distribution, pages 492–518. Springer, 1992

    1992

  22. [30]

    Learn- ing forward dynamics model and informed trajectory sam- pler for safe quadruped navigation

    Yunho Kim, Chanyoung Kim, and Jemin Hwangbo. Learn- ing forward dynamics model and informed trajectory sam- pler for safe quadruped navigation. InRobotics: Science and Systems (RSS), 2022. Preprint – to appear in CVPR 2026 June 2026

    2022

  23. [31]

    Globally consistent 3D LiDAR mapping with GPU- accelerated GICP matching cost factors.Robotics and Aut

    Kenji Koide, Masashi Yokozuka, Shuji Oishi, and Atsuhiko Banno. Globally consistent 3D LiDAR mapping with GPU- accelerated GICP matching cost factors.Robotics and Aut. Letters, 6(4):8591–8598, 2021

    2021

  24. [32]

    V oxelized GICP for fast and accurate 3D point cloud registration

    Kenji Koide, Masashi Yokozuka, Shuji Oishi, and Atsuhiko Banno. V oxelized GICP for fast and accurate 3D point cloud registration. InInt. conf. on robotics and automation, pages 11054–11059. IEEE, 2021

    2021

  25. [33]

    LiDAR odometry survey: recent advancements and remaining challenges.Intelligent Service Robotics, 17(2): 95–118, 2024

    Dongjae Lee, Minwoo Jung, Wooseong Yang, and Ayoung Kim. LiDAR odometry survey: recent advancements and remaining challenges.Intelligent Service Robotics, 17(2): 95–118, 2024

    2024

  26. [34]

    Risk- averse model predictive control for racing in adverse con- ditions.Int

    Thomas Lew, Marcus Greiff, Franck Djeumou, Makoto Suminaka, Michael Thompson, and John Subosits. Risk- averse model predictive control for racing in adverse con- ditions.Int. Conf. on Robotics and Automation, 2025

    2025

  27. [35]

    Unsupervised learning of edges

    Yin Li, Manohar Paluri, James M Rehg, and Piotr Doll ´ar. Unsupervised learning of edges. InConf. on computer vision and pattern recognition, pages 1619–1627, 2016

    2016

  28. [36]

    Se3et: Se(3)-equivariant transformer for low-overlap point cloud registration.IEEE Robotics and Automation Letters, 2024

    Chien Erh Lin, Minghan Zhu, and Maani Ghaffari. Se3et: Se(3)-equivariant transformer for low-overlap point cloud registration.IEEE Robotics and Automation Letters, 2024

    2024

  29. [37]

    PhD thesis, ¨Orebro university, 2009

    Martin Magnusson.The three-dimensional normal- distributions transform: an efficient representation for reg- istration, surface analysis, and loop detection. PhD thesis, ¨Orebro university, 2009

    2009

  30. [38]

    The pose estimation of mobile robot based on improved point cloud registration.Int

    Yanzi Miao, Yang Liu, Hongbin Ma, and Huijie Jin. The pose estimation of mobile robot based on improved point cloud registration.Int. Journal of Advanced Robotic Sys- tems, 13(2):52, 2016

    2016

  31. [39]

    R-LOAM: Improving LiDAR odometry and mapping with point-to-mesh features of a known 3D reference object

    Martin Oelsch, Mojtaba Karimi, and Eckehard Steinbach. R-LOAM: Improving LiDAR odometry and mapping with point-to-mesh features of a known 3D reference object. Robotics and Automation Letters, 6(2):2068–2075, 2021

  32. [40]

    RT3000 Product Page, 2024

    OxTS. RT3000 Product Page, 2024. Last accessed 05/20/2024

    2024

  33. [41]

    Dgcnn: A convolutional neural network over large-scale labeled graphs.Neural Networks, 108:533– 543, 2018

    Anh Viet Phan, Minh Le Nguyen, Yen Lam Hoang Nguyen, and Lam Thu Bui. Dgcnn: A convolutional neural network over large-scale labeled graphs.Neural Networks, 108:533– 543, 2018

    2018

  34. [42]

    Geotrans- former: Fast and robust point cloud registration with geo- metric transformer.Trans

    Zheng Qin, Hao Yu, Changjian Wang, Yulan Guo, Yuxing Peng, Slobodan Ilic, Dewen Hu, and Kai Xu. Geotrans- former: Fast and robust point cloud registration with geo- metric transformer.Trans. on Pattern Analysis and Machine Intelligence, 45(8):9806–9821, 2023

    2023

  35. [43]

    3D is here: Point cloud library (PCL)

    Radu Bogdan Rusu and Steve Cousins. 3D is here: Point cloud library (PCL). InInt. Conf. on robotics and automa- tion, pages 1–4. IEEE, 2011

    2011

  36. [44]

    Fast point feature histograms (fpfh) for 3d registration

    Radu Bogdan Rusu, Nico Blodow, and Michael Beetz. Fast point feature histograms (fpfh) for 3d registration. In2009 IEEE international conference on robotics and automation, pages 3212–3217. IEEE, 2009

    2009

  37. [45]

    Bad SLAM: Bundle adjusted direct RGB-D SLAM

    Thomas Schops, Torsten Sattler, and Marc Pollefeys. Bad SLAM: Bundle adjusted direct RGB-D SLAM. InProceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 134–144, 2019

    2019

  38. [46]

    Generalized-ICP

    Aleksandr Segal, Dirk Haehnel, and Sebastian Thrun. Generalized-ICP. InRobotics: science and systems, page

  39. [47]

    Lego-loam: Lightweight and ground-optimized LiDAR odometry and mapping on variable terrain

    Tixiao Shan and Brendan Englot. Lego-loam: Lightweight and ground-optimized LiDAR odometry and mapping on variable terrain. InInt. conf. on intelligent robots and sys- tems, pages 4758–4765. IEEE, 2018

    2018

  40. [48]

    The calculus of M- estimation.The American Statistician, 56(1):29–38, 2002

    Leonard A Stefanski and Dennis D Boos. The calculus of M- estimation.The American Statistician, 56(1):29–38, 2002

    2002

  41. [49]

    Scalability in percep- tion for autonomous driving: Waymo open dataset

    Pei Sun, Henrik Kretzschmar, Xerxes Dotiwalla, Aurelien Chouard, Vijaysai Patnaik, Paul Tsui, James Guo, Yin Zhou, Yuning Chai, Benjamin Caine, et al. Scalability in percep- tion for autonomous driving: Waymo open dataset. InConf. on computer vision and pattern recognition (CV...

    2020

  42. [50]

    A correlation-based ap- proach to robust point set registration

    Yanghai Tsin and Takeo Kanade. A correlation-based ap- proach to robust point set registration. InEuropean conf. on computer vision, pages 558–569. Springer Berlin Heidel- berg, 2004

    2004

  43. [51]

    Poisson surface reconstruction for LiDAR odometry and mapping

    Ignacio Vizzo, Xieyuanli Chen, Nived Chebrolu, Jens Behley, and Cyrill Stachniss. Poisson surface reconstruction for LiDAR odometry and mapping. InInt. conf. on robotics and automation (ICRA), pages 5624–5630. IEEE, 2021

    2021

  44. [52]

    Kiss-ICP: In defense of point-to-point ICP–simple, accurate, and robust registration if done the right way.Robotics and Automation Letters, 8(2):1029–1036, 2023

    Ignacio Vizzo, Tiziano Guadagnino, Benedikt Mersch, Louis Wiesmann, Jens Behley, and Cyrill Stachniss. Kiss-ICP: In defense of point-to-point ICP–simple, accurate, and robust registration if done the right way.Robotics and Automation Letters, 8(2):1029–1036, 2023

    2023

  45. [53]

    F- loam: Fast LiDAR odometry and mapping

    Han Wang, Chen Wang, Chun-Lin Chen, and Lihua Xie. F- loam: Fast LiDAR odometry and mapping. InInf. Conf. on Intelligent Robots and Systems, pages 4390–4396. IEEE, 2021

    2021

  46. [54]

    Gaussian processes for machine learning

    Christopher KI Williams and Carl Edward Rasmussen. Gaussian processes for machine learning. MIT press Cam- bridge, MA, 2006

    2006

  47. [55]

    3d shapenets: A deep representation for volumetric shapes

    Zhirong Wu, Shuran Song, Aditya Khosla, Fisher Yu, Lin- guang Zhang, Xiaoou Tang, and Jianxiong Xiao. 3d shapenets: A deep representation for volumetric shapes. In2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 1912–1920, 2015

    1912

  48. [56]

    FAST-LIO: A fast, robust LiDAR-inertial odometry package by tightly-coupled iter- ated kalman filter, 2021

    Wei Xu and Fu Zhang. FAST-LIO: A fast, robust LiDAR-inertial odometry package by tightly-coupled iter- ated kalman filter, 2021

    2021

  49. [57]

    FAST-LIO2: Fast direct LiDAR-inertial odometry.Transac- tions on Robotics, 38(4):2053–2073, 2022

    Wei Xu, Yixi Cai, Dongjiao He, Jiarong Lin, and Fu Zhang. FAST-LIO2: Fast direct LiDAR-inertial odometry.Transac- tions on Robotics, 38(4):2053–2073, 2022

  50. [58]

    Teaser: Fast and certifiable point cloud registration.Trans

    Heng Yang, Jingnan Shi, and Luca Carlone. Teaser: Fast and certifiable point cloud registration.Trans. on Rob., 37 (2):314–333, 2020

    2020

  51. [59]

    Go-ICP: A globally optimal solution to 3D ICP point-set registration.Trans

    Jiaolong Yang, Hongdong Li, Dylan Campbell, and Yunde Jia. Go-ICP: A globally optimal solution to 3D ICP point-set registration.Trans. on pattern analysis and machine intelli- gence, 38(11):2241–2254, 2015

    2015

  52. [60]

    Mac: Maxi- mal cliques for 3d registration.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    Jiaqi Yang, Xiyu Zhang, Peng Wang, Yulan Guo, Kun Sun, Qiao Wu, Shikun Zhang, and Yanning Zhang. Mac: Maxi- mal cliques for 3d registration.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024. Preprint – to appear in CVPR 2026 June 2026

    2024

  53. [61]

    Cofinet: Reliable coarse-to-fine correspondences for robust pointcloud registration.Advances in Neural Infor- mation Processing Systems, 34:23872–23884, 2021

    Hao Yu, Fu Li, Mahdi Saleh, Benjamin Busam, and Slobo- dan Ilic. Cofinet: Reliable coarse-to-fine correspondences for robust pointcloud registration.Advances in Neural Infor- mation Processing Systems, 34:23872–23884, 2021

    2021

  54. [62]

    3dmatch: Learning local geometric descriptors from RGB-D recon- structions

    Andy Zeng, Shuran Song, Matthias Nießner, Matthew Fisher, Jianxiong Xiao, and Thomas Funkhouser. 3dmatch: Learning local geometric descriptors from RGB-D recon- structions. InProceedings of the IEEE conference on com- puter vision and pattern recognition, pages 1802–1811, 2017

    2017

  55. [63]

    LOAM: LiDAR odometry and mapping in real-time

    Ji Zhang, Sanjiv Singh, et al. LOAM: LiDAR odometry and mapping in real-time. InRobotics: Science and systems, pages 1–9. Berkeley, CA, 2014

    2014

  56. [64]

    A new framework for registration of se- mantic point clouds from stereo and RGB-D cameras

    Ray Zhang, Tzu-Yuan Lin, Chien Erh Lin, Steven A Parki- son, William Clark, Jessy W Grizzle, Ryan M Eustice, and Maani Ghaffari. A new framework for registration of se- mantic point clouds from stereo and RGB-D cameras. In Int. Conf. on Robotics and Automation, pages 12214–122...

    2021

  57. [65]

    Correspondence-free SE(3) point cloud reg- istration in RKHS via unsupervised equivariant learning

    Ray Zhang, Zheming Zhou, Min Sun, Omid Ghasemal- izadeh, Cheng-Hao Kuo, Ryan M Eustice, Maani Ghaffari, and Arnie Sen. Correspondence-free SE(3) point cloud reg- istration in RKHS via unsupervised equivariant learning. In European conf. on computer vision, pages 68–86. Springer, 2024

    2024

  58. [66]

    RKHS-BA: A robust correspondence-free multi-view bun- dle adjustment framework for semantic point clouds.Trans

    Ray Zhang, Jingwei Song, Xiang Gao, Junzhe Wu, Tiany Liu, Jinyuan Zhang, Ryan Eustice, and Maani Ghaffari. RKHS-BA: A robust correspondence-free multi-view bun- dle adjustment framework for semantic point clouds.Trans. on Pattern Analysis and Machine Intelligence, 2025

    2025

  59. [67]

    LiDAR- based place recognition for autonomous driving: A survey

    Yongjun Zhang, Pengcheng Shi, and Jiayuan Li. LiDAR- based place recognition for autonomous driving: A survey. ACM Computing Surveys, 57(4):1–36, 2024

    2024

  60. [68]

    Progressive correspondence regenerator for robust 3d regis- tration

    Guiyu Zhao, Sheng Ao, Ye Zhang, Kai Xu, and Yulan Guo. Progressive correspondence regenerator for robust 3d regis- tration. InProceedings of the Computer Vision and Pattern Recognition Conference, pages 1210–1219, 2025

    2025

  61. [69]

    SubT-MRS Dataset: Pushing SLAM towards all-weather en- vironments

    Shibo Zhao, Yuanjun Gao, Tianhao Wu, Damanpreet Singh, Rushan Jiang, Haoxiang Sun, Mansi Sarawata, Yuheng Qiu, Warren Whittaker, Ian Higgins, Yi Du, Shaoshu Su, Can Xu, John Keller, Jay Karhade, Lucas Nogueira, Sourojit Saha, Ji Zhang, Wenshan Wang, Chen Wang, and Sebastian Sc...

    2024

  62. [70]

    Traj-LO: In defense of LiDAR- only odometry using an effective continuous-time trajectory

    Xin Zheng and Jianke Zhu. Traj-LO: In defense of LiDAR- only odometry using an effective continuous-time trajectory. Robotics and Automation Letters, 9(2):1961–1968, 2024

    1961

  63. [71]

    Fast global registration

    Qian-Yi Zhou, Jaesik Park, and Vladlen Koltun. Fast global registration. InEuropean conference on computer vision, pages 766–782. Springer, 2016

    2016

  64. [72]

    Correspondence-free point cloud registration with so(3)- equivariant implicit shape representations

    Minghan Zhu, Maani Ghaffari, and Huei Peng. Correspondence-free point cloud registration with so(3)- equivariant implicit shape representations. InProceedings of the 5th Conference on Robot Learning, pages 1412–1422. PMLR, 2022

    2022

  65. [73]

    ω∧gω +ω ∧gω +v ∧gv +g ∧ ωω+g ∧ v v v∧gω + 2ω∧gv +g ∧ ωv # . (31) Usinga ∧b=−b ∧a, we have v∧gv +g ∧ v v= 0,ω ∧gω +g ∧ ωω= 0,v ∧gω +g ∧ ωv= 0. (32) Hence, Γ∨(ξ∧,gradf(T)) =

    Minghan Zhu, Maani Ghaffari, William A Clark, and Huei Peng. E2PN: Efficient SE(3)-equivariant point network. In Conf. on Computer Vision and Pattern Recognition, pages 1223–1232, 2023. Preprint – to appear in CVPR 2026 June 2026 A. Proof of Lemma 1 Letγ(t) =Texp(tξ ∧). We wri...

    2023

Pith tools

Reviewed June 27, 2026 · model on record in the stance chip above.