pith. sign in

arxiv: 2605.08086 · v1 · submitted 2025-12-17 · 💻 cs.GR

Representations of 3D Rotations: Mathematical Foundations and Comparative Analysis

Aizierjiang Aiersilan , Haochen Liu , James Hahn This is my paper

Pith reviewed 2026-05-16 22:04 UTC · model grok-4.3

classification 💻 cs.GR
keywords 3D rotationsquaternionsSO(3)rotation representationsEuler anglescomputer graphicspose estimation
0
0 comments X p. Extension

The pith

Quaternions provide the most compact and computationally efficient representation for 3D rotations among common alternatives.

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

The paper surveys representations of rotations in SO(3) including Euler angles, axis-angle vectors, quaternions, rotation matrices, exponential maps, and newer continuous and probabilistic approaches. It evaluates each on mathematical formulation, continuity, gimbal lock susceptibility, computational cost, storage needs, interpolation behavior, and composition rules. The comparison integrates algebraic details with uses in animation, pose estimation, navigation, and neural networks. Quaternions stand out for compactness and speed while continuous representations like 6D vectors and matrix Fisher distributions improve smoothness and uncertainty handling. The work points to hybrid methods and larger-scale testing as next steps.

Core claim

The paper establishes through side-by-side analysis that quaternions dominate due to their compactness and computational efficiency, while alternatives like 6D continuous representations and matrix Fisher distributions provide enhanced continuity and uncertainty modeling.

What carries the argument

Comparative evaluation of SO(3) representations on continuity, efficiency, storage, interpolation, and composition.

If this is right

  • Quaternions reduce storage and computation time in real-time graphics pipelines.
  • 6D continuous representations avoid discontinuities that can disrupt gradient-based learning.
  • Matrix Fisher distributions enable explicit uncertainty estimates in inertial navigation.
  • Rotation matrices remain useful when direct access to the full linear transformation is required.
  • Exponential maps offer a compact vector form but require care near singularities.

Where Pith is reading between the lines

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

  • Hybrid representations that switch between quaternion speed and continuous smoothness could reduce failures in long motion sequences.
  • Applications with sensor noise may benefit more from probabilistic models even when efficiency drops slightly.
  • Standardizing a small set of representations in libraries could simplify interoperability between graphics and learning systems.

Load-bearing premise

That comparing listed properties such as continuity and efficiency through qualitative review is sufficient to identify the best representation without large-scale quantitative benchmarks across applications.

What would settle it

A controlled benchmark measuring runtime, error rates, and failure frequency across animation, robotics, and neural pose tasks where a non-quaternion method shows clear overall superiority.

Figures

Figures reproduced from arXiv: 2605.08086 by Aizierjiang Aiersilan, Haochen Liu, James Hahn.

Figure 1
Figure 1. Figure 1: Comparison of storage requirements, composition time, and [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Correlation map showing the applicability of different rotation [PITH_FULL_IMAGE:figures/full_fig_p016_2.png] view at source ↗
read the original abstract

Rotation representations are foundational in fields such as computer graphics, robotics, and machine learning, where precise and efficient modeling of 3D orientations is critical. This paper comprehensively investigates diverse representations of the special orthogonal group $SO(3)$, such as Euler angles, axis-angle vectors, quaternions, rotation matrices, exponential maps, and emerging continuous and probabilistic methods, evaluating their mathematical formulations, continuity, susceptibility to gimbal lock, computational efficiency, storage requirements, interpolation properties, and composition operations, while integrating detailed algebraic insights with practical applications in fields like animation, pose estimation, inertial navigation, 3D shape registration, and neural networks. Empirical evidence highlights quaternions' dominance due to their compactness and computational efficiency, while alternatives like 6D continuous representations and matrix Fisher distributions provide enhanced continuity and uncertainty modeling. Future research could explore hybrid methods and thorough large-scale evaluations to help build a solid foundation for improving rotation representation techniques.

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

1 major / 1 minor

Summary. The manuscript reviews mathematical representations of 3D rotations from the special orthogonal group SO(3), covering Euler angles, axis-angle vectors, quaternions, rotation matrices, exponential maps, and emerging continuous and probabilistic methods. It evaluates these on criteria including continuity, gimbal lock, computational efficiency, storage, interpolation, and composition, integrating algebraic insights with applications in animation, pose estimation, navigation, registration, and neural networks. The paper concludes that empirical evidence supports quaternions as dominant for compactness and efficiency, while 6D continuous representations and matrix Fisher distributions offer advantages in continuity and uncertainty modeling.

Significance. If the comparative analysis holds, this work provides a valuable reference for researchers and practitioners in computer graphics, robotics, and machine learning by organizing the trade-offs among rotation representations. It highlights strengths of quaternions based on established literature and points to future directions like hybrid methods. However, the lack of new empirical data limits its novelty to synthesis rather than advancing the state of the art with quantitative evidence.

major comments (1)
  1. Abstract: The assertion that 'Empirical evidence highlights quaternions' dominance due to their compactness and computational efficiency' is not backed by original experiments or benchmarks in the paper; the discussion appears to rely on qualitative synthesis of prior citations without new runtime measurements or large-scale tests across applications such as neural pose estimators.
minor comments (1)
  1. The future research directions paragraph could benefit from more concrete suggestions for the proposed large-scale evaluations, such as specific metrics or application domains to test.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comment point by point below.

read point-by-point responses

  1. Referee: Abstract: The assertion that 'Empirical evidence highlights quaternions' dominance due to their compactness and computational efficiency' is not backed by original experiments or benchmarks in the paper; the discussion appears to rely on qualitative synthesis of prior citations without new runtime measurements or large-scale tests across applications such as neural pose estimators.

    Authors: We agree that the manuscript is a comparative review and does not present new runtime benchmarks or large-scale experiments. The phrase 'empirical evidence' refers to the synthesis of results reported in the cited literature, which includes multiple prior studies demonstrating the compactness and efficiency advantages of quaternions. To avoid any ambiguity, we will revise the abstract to clarify that the conclusion is drawn from a review of existing empirical studies rather than new experiments conducted here. A possible rephrasing is: 'A synthesis of the literature highlights quaternions' dominance due to their compactness and computational efficiency'. This revision will be made in the updated manuscript. revision: yes

Circularity Check

0 steps flagged

No circularity: review paper synthesizes external literature without self-referential derivations or fitted predictions

full rationale

This is a comparative survey of established SO(3) representations (Euler angles, quaternions, matrices, etc.). All algebraic properties, continuity arguments, and efficiency rankings are drawn from standard group theory and prior published results; the abstract explicitly attributes the 'quaternions' dominance' claim to 'empirical evidence' in the cited literature rather than any new derivation or internal fit performed in the paper. No equations are defined in terms of their own outputs, no parameters are fitted to a subset and then relabeled as predictions, and no load-bearing uniqueness theorem is imported from the authors' own prior work. The manuscript therefore contains no reduction of claims to self-constructed inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper rests on standard properties of the special orthogonal group SO(3) and algebraic operations on common representations without introducing fitted parameters or new entities.

axioms (1)
  • standard math SO(3) is the group of 3x3 orthogonal matrices with determinant 1 representing 3D rotations
    Invoked throughout the abstract and title as the mathematical foundation for all compared representations.

pith-pipeline@v0.9.0 · 5460 in / 1088 out tokens · 32255 ms · 2026-05-16T22:04:53.901090+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

44 extracted references · 44 canonical work pages · 2 internal anchors

  1. [1]

    Formulae generales pro translatione quacunque cor- porum rigidorum,

    L. Euler, “Formulae generales pro translatione quacunque cor- porum rigidorum,” Novi commentarii academiae scientiarum petropolitanae, pp. 189–207, 1776

    work page

  2. [2]

    On a new species of imaginary quantities, connected with the theory of quaternions,

    W. R. Hamilton, “On a new species of imaginary quantities, connected with the theory of quaternions,” Proceedings of the Royal Irish Academy (1836-1869), vol. 2, pp. 424–434, 1840

    work page

  3. [3]

    Animating rotation with quaternion curves,

    K. Shoemake, “Animating rotation with quaternion curves,” in Proceedings of the 12th annual conference on Computer graphics and interactive techniques, 1985, pp. 245–254

    work page 1985

  4. [4]

    6d rotation representation for unconstrained head pose estimation,

    T. Hempel, A. A. Abdelrahman, and A. Al-Hamadi, “6d rotation representation for unconstrained head pose estimation,” in 2022 IEEE International Conference on Image Processing (ICIP). IEEE, 2022, pp. 2496–2500

    work page 2022

  5. [5]

    A vector-based repre- sentation to enhance head pose estimation,

    Z. Cao, Z. Chu, D. Liu, and Y . Chen, “A vector-based repre- sentation to enhance head pose estimation,” in Proceedings of the IEEE/CVF Winter Conference on applications of computer vision, 2021, pp. 1188–1197

    work page 2021

  6. [6]

    Let’s remove quaternions from every 3d en- gine,

    M. ten Bosch, “Let’s remove quaternions from every 3d en- gine,” URL: https://marctenbosch. com/quaternions, 2020

    work page 2020

  7. [7]

    We don’t really need quaternions in geometric modeling, computer graphics and animation: Here is why,

    F. Cheng, T. L. Johnson, A. Kazadi, E. G. Toney, J. I. Watson, and A. J. Lin, “We don’t really need quaternions in geometric modeling, computer graphics and animation: Here is why,” Computer-Aided Design and Applications, vol. 20, no. 6, pp. 1061–1073, 2023, publisher Copyright: © 2023 CAD Solutions, LLC

    work page 2023

  8. [8]

    Goldman, Rethinking quaternions

    R. Goldman, Rethinking quaternions. Springer Nature, 2022

    work page 2022

  9. [9]

    V oight, Quaternion algebras

    J. V oight, Quaternion algebras. Springer Nature, 2021

    work page 2021

  10. [10]

    Visualizing quaternion rotation,

    J. C. Hart, G. K. Francis, and L. H. Kauffman, “Visualizing quaternion rotation,” ACM Transactions on Graphics (TOG), vol. 13, no. 3, pp. 256–276, 1994

    work page 1994

  11. [11]

    O. Rodrigues, “Des lois g ´eom´etriques qui r ´egissent les d´eplacements d’un syst`eme solide dans l’espace, et de la varia- tion des coordonn´ees provenant de ces d´eplacements consid´er´es ind´ependamment des causes qui peuvent les produire,” Journal de math´ematiques pures et appliqu´ees, vol. 5, pp. 380–440, 1840

    work page

  12. [12]

    On the continuity of rotation representations in neural networks,

    Y . Zhou, C. Barnes, J. Lu, J. Yang, and H. Li, “On the continuity of rotation representations in neural networks,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2019, pp. 5745–5753

    work page 2019

  13. [13]

    Probabilistic orien- tation estimation with matrix fisher distributions,

    D. Mohlin, J. Sullivan, and G. Bianchi, “Probabilistic orien- tation estimation with matrix fisher distributions,” Advances in Neural Information Processing Systems, vol. 33, pp. 4884– 4893, 2020

    work page 2020

  14. [14]

    Properties and applications of fisher distribution on the rotation group,

    T. Sei, H. Shibata, A. Takemura, K. Ohara, and N. Takayama, “Properties and applications of fisher distribution on the rotation group,” Journal of Multivariate Analysis, vol. 116, pp. 440–455, 2013

    work page 2013

  15. [15]

    Bayesian attitude estimation with the matrix fisher dis- tribution on so (3),

    T. Lee, “Bayesian attitude estimation with the matrix fisher dis- tribution on so (3),” IEEE Transactions on Automatic Control, vol. 63, no. 10, pp. 3377–3392, 2018

    work page 2018

  16. [16]

    An antipodally symmetric distribution on the sphere,

    C. Bingham, “An antipodally symmetric distribution on the sphere,” The Annals of Statistics, pp. 1201–1225, 1974

    work page 1974

  17. [17]

    Bingham distribution-based linear filter for online pose estimation,

    R. A. Srivatsan, M. Xu, N. Zevallos, and H. Choset, “Bingham distribution-based linear filter for online pose estimation,” in Robotics: Science and Systems. Robotics Science and Systems Foundation, 2017

    work page 2017

  18. [18]

    Representing attitude: Euler angles, unit quaternions, and rotation vectors,

    J. Diebel et al., “Representing attitude: Euler angles, unit quaternions, and rotation vectors,” Matrix, vol. 58, no. 15-16, pp. 1–35, 2006

    work page 2006

  19. [19]

    Rotation representations and their con- versions,

    S. Kim and M. Kim, “Rotation representations and their con- versions,” Ieee Access, vol. 11, pp. 6682–6699, 2023

    work page 2023

  20. [20]

    Adaptive sliding mode attitude tracking control for rigid spacecraft considering the unwinding problem,

    B. Huo, M. Du, and Z. Yan, “Adaptive sliding mode attitude tracking control for rigid spacecraft considering the unwinding problem,” Mathematics, vol. 11, no. 20, p. 4372, 2023

    work page 2023

  21. [21]

    QuaterNet: A Quaternion-based Recurrent Model for Human Motion

    D. Pavllo, D. Grangier, and M. Auli, “Quaternet: A quaternion- based recurrent model for human motion,” arXiv preprint arXiv:1805.06485, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  22. [22]

    Why and how to avoid the flipped quaternion multiplication,

    H. Sommer, I. Gilitschenski, M. Bloesch, S. Weiss, R. Siegwart, and J. Nieto, “Why and how to avoid the flipped quaternion multiplication,” Aerospace, vol. 5, no. 3, p. 72, 2018

    work page 2018

  23. [23]

    On quaternion- based attitude control and the unwinding phenomenon,

    C. G. Mayhew, R. G. Sanfelice, and A. R. Teel, “On quaternion- based attitude control and the unwinding phenomenon,” in Proceedings of the 2011 american control conference. IEEE, 2011, pp. 299–304

    work page 2011

  24. [24]

    Nova methodus motum corporum rigidorum degerminandi,

    L. Euler, “Nova methodus motum corporum rigidorum degerminandi,” Novi commentarii academiae scientiarum Petropolitanae, pp. 208–238, 1776

    work page

  25. [25]

    Lie and F

    S. Lie and F. Engel, Theorie der transformationsgruppen. Teub- ner, 1893, vol. 3

    work page

  26. [26]

    Maximum likelihood esti- mation of models for residual covariance in spatial regression,

    K. V . Mardia and R. J. Marshall, “Maximum likelihood esti- mation of models for residual covariance in spatial regression,” Biometrika, vol. 71, no. 1, pp. 135–146, 1984

    work page 1984

  27. [27]

    Optimizing software in c++,

    A. Fog, “Optimizing software in c++,” URL: http://www. agner. org/optimize/optimizing cpp. pdf, 2006

    work page 2006

  28. [28]

    Optimizing matrix mul- tiplication for a short-vector simd architecture–cell processor,

    J. Kurzak, W. Alvaro, and J. Dongarra, “Optimizing matrix mul- tiplication for a short-vector simd architecture–cell processor,” Parallel Computing, vol. 35, no. 3, pp. 138–150, 2009

    work page 2009

  29. [29]

    R. M. Murray, Z. Li, and S. S. Sastry, A mathematical introduction to robotic manipulation. CRC press, 2017

    work page 2017

  30. [30]

    Optimal rotation alignment of 3d objects using a gpu-based similarity function,

    M. Martinek and R. Grosso, “Optimal rotation alignment of 3d objects using a gpu-based similarity function,” Computers & Graphics, vol. 33, no. 3, pp. 291–298, 2009

    work page 2009

  31. [31]

    On the Efficacy and High-Performance Implementation of Quaternion Matrix Multiplication

    D. Williams-Young and X. Li, “On the efficacy and high- performance implementation of quaternion matrix multiplica- tion,” arXiv preprint arXiv:1903.05575, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1903

  32. [32]

    Practical parameterization of rotations using the exponential map,

    F. S. Grassia, “Practical parameterization of rotations using the exponential map,” Journal of graphics tools, vol. 3, no. 3, pp. 29–48, 1998

    work page 1998

  33. [33]

    From quaternion to matrix and back,

    J. Van Waveren, “From quaternion to matrix and back,” Id Software, Inc, 2005

    work page 2005

  34. [34]

    Yi, A Study of Performance Programming of CPU, GPU accelerated Computers and SIMD Architecture, arXiv:2409.10661 (2024)

    X. Yi, “A study of performance programming of cpu, gpu accelerated computers and simd architecture,” arXiv preprint arXiv:2409.10661, 2024

    work page arXiv 2024

  35. [35]

    Learning with 3d rotations: a hitchhiker’s guide to so (3),

    A. R. Geist, J. Frey, M. Zhobro, A. Levina, and G. Martius, “Learning with 3d rotations: a hitchhiker’s guide to so (3),” in Proceedings of the 41st International Conference on Machine Learning, 2024, pp. 15 331–15 350

    work page 2024

  36. [36]

    Rins-w: Robust inertial navigation system on wheels,

    M. Brossard, A. Barrau, and S. Bonnabel, “Rins-w: Robust inertial navigation system on wheels,” in 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2019, pp. 2068–2075

    work page 2019

  37. [37]

    Unscented orientation estimation based on the bingham dis- tribution,

    I. Gilitschenski, G. Kurz, S. J. Julier, and U. D. Hanebeck, “Unscented orientation estimation based on the bingham dis- tribution,” IEEE Transactions on Automatic Control, vol. 61, no. 1, pp. 172–177, 2015

    work page 2015

  38. [38]

    Closed-form solution of absolute orientation using unit quaternions,

    B. K. Horn, “Closed-form solution of absolute orientation using unit quaternions,” Journal of the optical society of America A, vol. 4, no. 4, pp. 629–642, 1987

    work page 1987

  39. [39]

    Method for registration of 3- d shapes,

    P. J. Besl and N. D. McKay, “Method for registration of 3- d shapes,” in Sensor fusion IV: control paradigms and data structures, vol. 1611. Spie, 1992, pp. 586–606

    work page 1992

  40. [40]

    Learning rotations,

    A. Pepe, J. Lasenby, and P. Chac ´on, “Learning rotations,” Mathematical Methods in the Applied Sciences, vol. 47, no. 3, pp. 1204–1217, 2024

    work page 2024

  41. [41]

    Dual quaternion operations for rigid body motion and their application to the hand–eye calibration,

    X. Wang, H. Sun, C. Liu, and H. Song, “Dual quaternion operations for rigid body motion and their application to the hand–eye calibration,” Mechanism and Machine Theory, vol. 193, p. 105566, 2024

    work page 2024

  42. [42]

    Revisiting rotation averaging: Uncertainties and robust losses,

    G. Zhang, V . Larsson, and D. Barath, “Revisiting rotation averaging: Uncertainties and robust losses,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 17 215–17 224

    work page 2023

  43. [43]

    Geometric deep learning and equivariant neural networks,

    J. E. Gerken, J. Aronsson, O. Carlsson, H. Linander, F. Ohlsson, C. Petersson, and D. Persson, “Geometric deep learning and equivariant neural networks,” Artificial Intelligence Review, vol. 56, no. 12, pp. 14 605–14 662, 2023

    work page 2023

  44. [44]

    E (3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials,

    S. Batzner, A. Musaelian, L. Sun, M. Geiger, J. P. Mailoa, M. Kornbluth, N. Molinari, T. E. Smidt, and B. Kozinsky, “E (3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials,” Nature communications, vol. 13, no. 1, p. 2453, 2022

    work page 2022