pith. sign in

arxiv: 2501.03972 · v2 · submitted 2025-01-07 · 💻 cs.RO

MAD-BA: 3D LiDAR Bundle Adjustment -- from Uncertainty Modelling to Structure Optimization

Krzysztof \'Cwian , Luca Di Giammarino , Simone Ferrari , Thomas Ciarfuglia , Giorgio Grisetti , Piotr Skrzypczy\'nski This is my paper

Pith reviewed 2026-05-23 05:31 UTC · model grok-4.3

classification 💻 cs.RO
keywords LiDAR bundle adjustmentsurfel mapuncertainty modelingpose optimizationstructure optimization3D mappingrobotics state estimation
0
0 comments X

The pith

Joint LiDAR pose and surfel map optimization with a generalized uncertainty model improves accuracy over most state-of-the-art methods.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper presents a framework that simultaneously optimizes LiDAR sensor poses and the 3D map structure, where the map consists of surfels. It introduces a generalized uncertainty model to reduce the influence of less reliable measurements that arise in different scenarios. A sympathetic reader would care because most existing LiDAR systems optimize poses while treating structure refinement separately or implicitly, so a joint explicit approach could produce more consistent maps and poses in one step. The paper supports this with experiments showing better results than most comparable methods on public datasets and releases the code as open source.

Core claim

The central claim is that representing the 3D map as surfels and performing bundle adjustment to jointly optimize sensor poses and map structure, guided by a generalized LiDAR uncertainty model that accounts for measurement reliability in varying scenarios, results in improved performance compared to most state-of-the-art methods on public datasets.

What carries the argument

Generalized LiDAR uncertainty model combined with surfel-based joint pose and structure optimization inside bundle adjustment.

If this is right

  • The resulting 3D maps achieve higher accuracy and consistency than those from pose-only optimization.
  • Less reliable LiDAR measurements exert less influence on the final optimized result.
  • The surfel representation enables explicit structure refinement during the same optimization pass as pose estimation.
  • The open-source release allows direct application and testing on additional LiDAR datasets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The uncertainty weighting scheme could be adapted to fuse data from additional range sensors beyond LiDAR.
  • Surfel maps may simplify integration with geometric tasks such as surface reconstruction or collision checking compared with implicit representations.
  • The method could be tested on sequences that include moving objects to check whether the static surfel assumption holds.

Load-bearing premise

The generalized uncertainty model accurately captures measurement reliability across scenarios and the surfel representation plus joint optimization produces genuinely better results without dataset-specific biases.

What would settle it

Running the open-source implementation on the same public datasets and finding that performance metrics do not improve when the uncertainty model is removed or when a different map representation is used would falsify the central claim.

Figures

Figures reproduced from arXiv: 2501.03972 by Giorgio Grisetti, Krzysztof \'Cwian, Luca Di Giammarino, Piotr Skrzypczy\'nski, Simone Ferrari, Thomas Ciarfuglia.

Figure 1
Figure 1. Figure 1: Pipeline. Given a trajectory estimated by a LiDAR SLAM system our method jointly optimizes the poses and map. It leverages on our uncertainty model to weight the optimization. The two bottom-right insets qualitatively demonstrate the map refinement achieved by employing our system. association is also employed in our proposed framework. SuMa [2], on the other hand, employs surfel-based map rep￾resentations… view at source ↗
Figure 2
Figure 2. Figure 2: LiDAR uncertainty model. In the top image, a single LiDAR beam is simulated by casting a set of sub￾beams towards a leaf l. At the bottom, the uncertainty of this measurement is modeled as a Gaussian distribution computed from the sampled ranges. simulation prevents the training models from generalizing to real data. To synthesize realistic scans from novel viewpoints, a physically motivated model of the s… view at source ↗
Figure 3
Figure 3. Figure 3: ATE for each iteration of the bundle adjustment algorithm. The plot compares three versions of our system for math-easy sequence and shows that both versions of our BA reduce the trajectory error and accelerate the converge of the algorithm related to the pose-only optimization. For BA with uncertainty, the error didn’t increase in the first iteration because measurements with higher uncertainty are weight… view at source ↗
Figure 4
Figure 4. Figure 4: Chamfer-L1 distance of the cloister sequence for different distance thresholds. The initial map is a surfel map that was created using the initial trajectory. Both variants of our BA notably enhance the quality of the map, however, integrating uncertainty information results in additional im￾provements. convergence, especially compared to pose-only optimization. The map evaluation was performed by comparin… view at source ↗
read the original abstract

The joint optimization of sensor poses and 3D structure is fundamental for state estimation in robotics and related fields. Current LiDAR systems often prioritize pose optimization, with structure refinement either omitted or treated separately using implicit representations. This paper introduces a framework for simultaneous optimization of sensor poses and 3D map, represented as surfels. A generalized LiDAR uncertainty model is proposed to address less reliable measurements in varying scenarios. Experimental results on public datasets demonstrate improved performance over most comparable state-of-the-art methods. The system is provided as open-source software to support further research.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 2 minor

Summary. The paper introduces MAD-BA, a framework for simultaneous optimization of sensor poses and 3D map structure represented as surfels in LiDAR bundle adjustment. It proposes a generalized uncertainty model to handle less reliable measurements across varying scenarios. Experimental results on public datasets are reported to show improved performance over most comparable state-of-the-art methods, with the system released as open-source software.

Significance. If the results hold, this contributes to LiDAR-based state estimation by providing an integrated approach to joint pose-structure optimization with explicit uncertainty modeling, potentially improving accuracy in robotics mapping and SLAM applications. The open-source release is a clear strength that enables reproducibility and community validation.

minor comments (2)
  1. Abstract: The claim of 'improved performance over most comparable state-of-the-art methods' is stated without naming the specific metrics (e.g., ATE, RPE), datasets, or quantitative improvements; adding these details would strengthen the summary for readers.
  2. The manuscript would benefit from an explicit statement in the introduction or experiments section clarifying the exact baselines compared and any data exclusion criteria used in the evaluation.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary, significance assessment, and recommendation of minor revision. No major comments were listed in the report.

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper introduces a new optimization framework for LiDAR bundle adjustment with surfel-based structure and a generalized uncertainty model, validated via experiments on public datasets. No load-bearing derivation step reduces by construction to fitted inputs, self-definitions, or self-citation chains; the central claims rest on the proposed model and empirical comparisons rather than renaming or smuggling prior results. The open-source release and external dataset evaluation further confirm the result is not internally forced.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; the uncertainty model likely rests on unstated modeling assumptions about measurement reliability but these cannot be audited from the given text.

pith-pipeline@v0.9.0 · 5644 in / 1021 out tokens · 30862 ms · 2026-05-23T05:31:06.398896+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

  1. [1]

    Airborne laser scanning: basic relations and formulas

    Emmanuel P Baltsavias. Airborne laser scanning: basic relations and formulas. ISPRS Journal of Photogrammetry and Remote Sensing (JPRS), 54(2-3):199–214, 1999

    work page 1999

  2. [2]

    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 Hadas Kress-Gazit, Siddhartha S. Srinivasa, Tom Howard, and Nikolay Atanasov, editors, Robotics: Science and Systems XIV , Carnegie Mellon University, Pittsburgh, Pennsylvania, USA, June 26-30, 2018 , 2018

    work page 2018

  3. [3]

    VBR: A vision benchmark in Rome

    Leonardo Brizi, Emanuele Giacomini, Luca Di Giammarino, Simone Ferrari, Omar Salem, Lorenzo De Rebotti, and Giorgio Grisetti. VBR: A vision benchmark in Rome. In Proc. of the IEEE Intl. Conf. on Robotics & Automation (ICRA) , pages 15868–15874, 2024

    work page 2024

  4. [4]

    Nowicki, Jan Wietrzykowski, and Piotr Skrzypczy´nski

    Krzysztof ´Cwian, Michał R. Nowicki, Jan Wietrzykowski, and Piotr Skrzypczy´nski. Large-scale LiDAR SLAM with factor graph opti- mization on high-level geometric features. Sensors, 21(10), 2021

    work page 2021

  5. [5]

    NeRF-LOAM: Neural implicit representation for large-scale incremental lidar odometry and mapping

    Junyuan Deng, Qi Wu, Xieyuanli Chen, Songpengcheng Xia, Zhen Sun, Guoqing Liu, Wenxian Yu, and Ling Pei. NeRF-LOAM: Neural implicit representation for large-scale incremental lidar odometry and mapping. In IEEE/CVF International Conference on Computer Vision (ICCV), pages 8184–8193, 2023

    work page 2023

  6. [6]

    Photometric LiDAR and RGB-D bundle adjustment

    Luca Di Giammarino, Emanuele Giacomini, Leonardo Brizi, Omar Salem, and Giorgio Grisetti. Photometric LiDAR and RGB-D bundle adjustment. IEEE Robotics and Automation Letters (RA-L) , 2023

    work page 2023

  7. [7]

    Bundle adjustment rules

    Chris Engels, Henrik Stew ´enius, and David Nist ´er. Bundle adjustment rules. Photogrammetric computer vision , 2(32), 2006

    work page 2006

  8. [8]

    MAD-ICP: It is all about matching data – robust and informed LiDAR odometry

    Simone Ferrari, Luca Di Giammarino, Leonardo Brizi, and Giorgio Grisetti. MAD-ICP: It is all about matching data – robust and informed LiDAR odometry. IEEE Robotics and Automation Letters (RA-L) , 9(11):9175–9182, 2024

    work page 2024

  9. [9]

    Geiger, P

    A. Geiger, P. Lenz, C. Stiller, and R. Urtasun. Vision meets robotics: The KITTI dataset. Intl. Journal of Robotics Research (IJRR) , 32(11):1231–1237, 2013

    work page 2013

  10. [10]

    Di Giammarino, L

    L. Di Giammarino, L. Brizi, T. Guadagnino, C. Stachniss, and G. Grisetti. Md-slam: Multi-cue direct SLAM. In Proc. of the IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems (IROS) , pages 11047–11054. IEEE, 2022

    work page 2022

  11. [11]

    Grisetti, T

    G. Grisetti, T. Guadagnino, I. Aloise, M. Colosi, B. Della Corte, and D. Schlegel. Least squares optimization: From theory to practice. Robotics, 9(3):51, 2020

    work page 2020

  12. [12]

    B. Horn, H. Hilden, and S. Negahdaripour. Closed-form solution of absolute orientation using orthonormal matrices. JOSA A , 5(7):1127– 1135, 1988

    work page 1988

  13. [13]

    Analytical formula to investigate the modulation of sloped targets using LiDAR waveform

    Yihua Hu, Ahui Hou, Qingli Ma, Nanxiang Zhao, Shilong Xu, and Jiajie Fang. Analytical formula to investigate the modulation of sloped targets using LiDAR waveform. Transactions on Geoscience and Remote Sensing , 60:1–12, 2021

    work page 2021

  14. [14]

    Neural LiDAR fields for novel view synthesis

    Shengyu Huang, Zan Gojcic, Zian Wang, Francis Williams, Yoni Kasten, Sanja Fidler, Konrad Schindler, and Or Litany. Neural LiDAR fields for novel view synthesis. In Proc. of the IEEE Intl. Conf. on Computer Vision (ICCV) , pages 18236–18246, 2023

    work page 2023

  15. [15]

    Large-scale LiDAR consistent mapping using hierarchical lidar bundle adjustment

    Xiyuan Liu, Zheng Liu, Fanze Kong, and Fu Zhang. Large-scale LiDAR consistent mapping using hierarchical lidar bundle adjustment. IEEE Robotics and Automation Letters (RA-L) , 8(3):1523–1530, 2023

    work page 2023

  16. [16]

    Liu and F

    Z. Liu and F. Zhang. Balm: Bundle adjustment for LiDAR mapping. IEEE Robotics and Automation Letters (RA-L) , 6(2):3184–3191, 2021

    work page 2021

  17. [17]

    A fast nearest-neighbor algorithm based on a principal axis search tree

    James McNames. A fast nearest-neighbor algorithm based on a principal axis search tree. IEEE Trans. on Pattern Analalysis and Machine Intelligence (TPAMI) , 23(9):964–976, 2001

    work page 2001

  18. [18]

    Ramezani, Y

    M. Ramezani, Y . Wang, M. Camurri, D. Wisth, M. Mattamala, and M. Fallon. The Newer College dataset: Handheld LiDAR, inertial and vision with ground truth. In Proc. of the IEEE/RSJ Intl. Conf. on Intelligent Robots and Systems (IROS) , pages 4353–4360, 2020

    work page 2020

  19. [19]

    Schops, T

    T. Schops, T. Sattler, and M. Pollefeys. Bad SLAM: Bundle adjusted direct RGBD-D SLAM. In Proc. of the IEEE Conf. on Computer Vision and Pattern Recognition (CVPR) , pages 134–144, 2019

    work page 2019

  20. [20]

    Spatial uncertainty management for simultaneous localization and mapping

    Piotr Skrzypczynski. Spatial uncertainty management for simultaneous localization and mapping. In Proceedings 2007 IEEE International Conference on Robotics and Automation , pages 4050–4055, 2007

    work page 2007

  21. [21]

    Stoyanov, M

    T. Stoyanov, M. Magnusson, H. Andreasson, and A. J. Lilienthal. Fast and accurate scan registration through minimization of the distance between compact 3D NDT representations. Intl. Journal of Robotics Research (IJRR), 31(12):1377–1393, 2012

    work page 2012

  22. [22]

    McLauchlan, Richard I

    Bill Triggs, Philip F. McLauchlan, Richard I. Hartley, and Andrew W. Fitzgibbon. Bundle adjustment — a modern synthesis. In Bill Triggs, Andrew Zisserman, and Richard Szeliski, editors, Vision Al- gorithms: Theory and Practice , pages 298–372, Berlin, Heidelberg,

    work page

  23. [23]

    Springer Berlin Heidelberg

    work page

  24. [24]

    Velas, M

    M. Velas, M. Spanel, and A. Herout. Collar line segments for fast odometry estimation from Velodyne point clouds. In Proc. of the IEEE Intl. Conf. on Robotics & Automation (ICRA) , pages 4486–4495, 2016

    work page 2016

  25. [25]

    Vizzo, T

    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 (RA-L) , 8(2):1029–1036, 2023

    work page 2023

  26. [26]

    Airborne laser scanning—an introduc- tion and overview

    Aloysius Wehr and Uwe Lohr. Airborne laser scanning—an introduc- tion and overview. ISPRS Journal of Photogrammetry and Remote Sensing (JPRS) , 54(2-3):68–82, 1999

    work page 1999

  27. [27]

    VDB-GPDF: Online Gaussian process distance field with VDB structure

    Lan Wu, Cedric Le Gentil, and Teresa Vidal-Calleja. VDB-GPDF: Online Gaussian process distance field with VDB structure. IEEE Robotics and Automation Letters (RA-L) , 10(1):374–381, 2025

    work page 2025

  28. [28]

    Correction of linear-array lidar intensity data using an optimal beam shaping approach

    Fan Xu, Yuanqing Wang, Xingyu Yang, Bingqing Zhang, and Fenfang Li. Correction of linear-array lidar intensity data using an optimal beam shaping approach. Optics and Lasers in Engineering , 83:90–98, 2016

    work page 2016

  29. [29]

    Zhang, M

    J. Zhang, M. Kaess, and S. Singh. On degeneracy of optimization- based state estimation problems. In Proc. of the IEEE Intl. Conf. on Robotics & Automation (ICRA) , pages 809–816, 2016

    work page 2016

  30. [30]

    Zhang and S

    J. Zhang and S. Singh. LOAM: Lidar odometry and mapping in real- time. In Proc. of Robotics: Science and Systems (RSS) , 2014

    work page 2014