pith. sign in

arxiv: 2605.04480 · v1 · submitted 2026-05-06 · 🧮 math.NA · cs.NA

Geometric Milstein Scheme for Stochastic Differential Equations on SO(n) and SE(n)

Xi Wang , Victor Solo This is my paper

Pith reviewed 2026-05-08 16:46 UTC · model grok-4.3

classification 🧮 math.NA cs.NA
keywords stochastic differential equationsLie groupsMilstein schemetangent space parameterizationSO(n)SE(n)strong convergencegeometric integration
0
0 comments X

The pith

A tangent-space corrected Milstein scheme converges strongly at order 1 for SDEs on SO(n) and SE(n) while preserving the group structure.

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

The paper develops a numerical integrator for stochastic differential equations that evolve on the Lie groups of rotations and rigid motions. It extends an existing tangent-space parameterization technique from ordinary differential equations to the stochastic setting by adding new correction terms inside a Milstein step. This extension overcomes the obstacle that Magnus expansions create for higher-order stochastic integrators. A reader would care because many applications require simulating noisy rotations or motions without the solution drifting off the manifold. If the claim holds, the method supplies a practical, provably order-1 accurate scheme that works whether or not the driving noise commutes.

Core claim

The tangent-space parameterization corrected Milstein method (TaSP-CM) keeps the numerical trajectory exactly on SO(n) or SE(n) at every step and delivers strong convergence of order 1 for both commutative and non-commutative noise.

What carries the argument

Tangent-space parameterization of the SDE together with explicit stochastic correction terms inside the Milstein increment.

If this is right

  • The numerical solution remains exactly on the Lie group for any time step without projection or retraction.
  • The same scheme attains order-1 strong convergence whether the noise vector fields commute or not.
  • The construction bypasses the need to expand the exponential map to higher orders via the Magnus series.
  • Numerical experiments on rotation and rigid-body problems confirm both the theoretical rate and practical stability.

Where Pith is reading between the lines

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

  • The same correction strategy may permit construction of strong order 1.5 schemes on the same groups.
  • The approach could be carried over to other matrix Lie groups once an analogous tangent-space chart is available.
  • Applications that require long-time preservation of invariants, such as stochastic rigid-body dynamics, would inherit the exact group membership property.

Load-bearing premise

The added stochastic corrections simultaneously restore geometric consistency and recover the full order-1 strong accuracy that the classical Milstein scheme possesses in Euclidean space.

What would settle it

A direct computation of the strong error for a sequence of halved step sizes on a simple non-commutative test SDE on SO(3) that fails to decrease proportionally to the step size, or any observed departure of the numerical matrix from exact orthogonality or determinant one.

Figures

Figures reproduced from arXiv: 2605.04480 by Victor Solo, Xi Wang.

Figure 1
Figure 1. Figure 1: Numerical performance of TaSP–CM and competing me view at source ↗
Figure 2
Figure 2. Figure 2: Numerical performance of TaSP–CM and competing me view at source ↗
Figure 3
Figure 3. Figure 3: Numerical performance of TaSP–CM and competing me view at source ↗
read the original abstract

In the paper, we propose a higher-order geometry-preserving numerical method for stochastic differential equations (SDEs) evolving on the Lie groups SO(n) and SE(n). Most existing Lie group integrators rely on Magnus expansion of the exponential map, which makes the construction of higher-order stochastic schemes difficult. To overcome this limitation, we develop a tangent-space parameterization corrected Milstein method (TaSP-CM), extending the tangent space parameterization (TaSP) framework from Lie-group ODEs to the stochastic setting. Although TaSP is a well-established method for Lie ODEs, the extension to SDEs is non-trivial and requires new stochastic corrections that ensure both geometric consistency and higher-order accuracy. We prove that the proposed scheme achieves strong convergence of order 1 under both commutative and non-commutative noise. Numerical experiments illustrate the theoretical results and demonstrate the efficiency and robustness of the proposed method.

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 / 3 minor

Summary. The paper proposes the tangent-space parameterization corrected Milstein (TaSP-CM) scheme for SDEs on the Lie groups SO(n) and SE(n). It extends the deterministic TaSP framework by deriving new stochastic correction terms to achieve geometric consistency. The central claim is a proof that the scheme attains strong order-1 convergence for both commutative and non-commutative noise, accompanied by numerical experiments demonstrating efficiency and robustness.

Significance. If the convergence analysis holds, the result would represent a meaningful advance in geometric integrators for stochastic problems on Lie groups. By avoiding Magnus expansions and providing explicit stochastic corrections that preserve both the manifold structure and Milstein order, the method fills a gap between existing low-order geometric schemes and the need for higher-order accuracy in applications such as rigid-body dynamics and robotics under noise. The explicit treatment of non-commutative noise and the numerical validation are particular strengths.

minor comments (3)
  1. The abstract states that new stochastic corrections are derived, but the precise form of these corrections (e.g., the additional terms beyond the deterministic TaSP drift) should be displayed explicitly in the main text near the scheme definition to aid readability.
  2. In the convergence theorem, the dependence of the order-1 constant on the Lie-group dimension n and on the noise commutativity should be stated more clearly; currently the bound appears uniform but the proof sketch leaves the n-dependence implicit.
  3. The numerical experiments section would benefit from an additional table or plot comparing CPU time versus error against a standard Euler–Maruyama method on the same manifold to quantify the efficiency gain claimed in the abstract.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive summary of our work on the TaSP-CM scheme, for recognizing its potential significance in geometric stochastic integrators, and for recommending minor revision. No specific major comments were raised in the report.

Circularity Check

0 steps flagged

No significant circularity detected in derivation chain

full rationale

The paper extends an established TaSP framework for deterministic Lie-group ODEs to the stochastic setting by introducing new stochastic correction terms in the Milstein scheme to maintain geometric consistency on SO(n) and SE(n). The central result is a proof of strong order-1 convergence for both commutative and non-commutative noise, which rests on standard stochastic Taylor expansions and Lie-group properties rather than redefining any input quantities or relying on self-citations for the core convergence claim. No self-definitional loops, fitted inputs renamed as predictions, or load-bearing self-citations appear in the described structure; the derivation introduces independent corrections and is therefore self-contained against external benchmarks in stochastic analysis.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard Lie-group differential geometry and the non-trivial extension of the existing TaSP framework to the stochastic case.

axioms (2)
  • standard math Lie group structure and exponential map properties for SO(n) and SE(n)
    Invoked implicitly as the setting for the SDEs and the requirement of geometry preservation.
  • domain assumption Tangent space parameterization is well-defined and extendable from ODEs to SDEs
    Stated as the basis for the new corrections in the abstract.

pith-pipeline@v0.9.0 · 5444 in / 1363 out tokens · 53413 ms · 2026-05-08T16:46:18.345441+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

33 extracted references · 33 canonical work pages

  1. [1]

    Brockett

    Roger W. Brockett. Lie algebras and Lie groups in control theory. In Geometric Methods in System Theory, pages 43–82, Dordrecht, 1973. Springer Netherlands

    work page 1973

  2. [2]

    Theory and algorithms for diffusion processes on riemannian manifolds.arXiv preprint arXiv:2204.13665, 2022

    Xiang Cheng, Jingzhao Zhang, and Suvrit Sra. Theory and a lgorithms for diffusion processes on Riemannian manifolds. arXiv preprint arXiv:2204.13665 , 2022

    work page arXiv 2022

  3. [3]

    Batista, and Kipton Barros

    David Dahlbom, Hao Zhang, Cole Miles, Xiaojian Bai, Cris tian D. Batista, and Kipton Barros. Ge- ometric integration of classical spin dynamics via a mean-fi eld Schr¨ odinger equation. Phys. Rev. B , 106:054423, Aug 2022

    work page 2022

  4. [4]

    Riemannian score-based generative modelling

    Valentin De Bortoli, Emile Mathieu, Michael Hutchinson , James Thornton, Yee Whye Teh, and Ar- naud Doucet. Riemannian score-based generative modelling . In NeurIPS, volume 35, pages 2406–2422, 2022

    work page 2022

  5. [5]

    Geometric integration for particle accele rators

    ´Etienne Forest. Geometric integration for particle accele rators. J. Phys. A: Math. Gen. , 39(19):5321, apr 2006

    work page 2006

  6. [6]

    Differential geometry and Lie groups , vol- ume 12

    JEAN QUAINTANCE Gallier and Jocelyn Quaintance. Differential geometry and Lie groups , vol- ume 12. Springer, Cham, 2020

    work page 2020

  7. [7]

    T. H. Gronwall. Note on the derivatives with respect to a p arameter of the solutions of a system of differential equations. Annals of Mathematics , 20(4):292–296, 1919

    work page 1919

  8. [8]

    Geometric numerical integration

    Ernst Hairer, Marlis Hochbruck, Arieh Iserles, and Chri stian Lubich. Geometric numerical integration. Oberwolfach Reports, 3(1):805–882, 2006

    work page 2006

  9. [9]

    An Introduction to the Numerical Simulation of Stochastic Differential Equations

    Desmond Higham and Peter Kloeden. An Introduction to the Numerical Simulation of Stochastic Differential Equations . SIAM, Philadelphia, PA, US, 2021

    work page 2021

  10. [10]

    Stochastic analysis on manifolds

    Elton P Hsu. Stochastic analysis on manifolds . Number 38. American Mathematical Soc., Providence, RI, US, 2002

    work page 2002

  11. [11]

    An analysis of appro ximation algorithms for iterated stochastic integrals and a julia and matlab simulation toolbox

    Felix Kastner and Andreas R¨ oßler. An analysis of appro ximation algorithms for iterated stochastic integrals and a julia and matlab simulation toolbox. Numer. Alg. , 93(1):27–66, 2023

    work page 2023

  12. [12]

    Kloeden and Eckhard Platen

    Peter E. Kloeden and Eckhard Platen. Numerical Solution of Stochastic Differential Equations , vol- ume 23 of Stochastic Modelling and Applied Probability . Springer, Berlin, Heidelberg, 1992

    work page 1992

  13. [13]

    Junghyun Kwon, Minseok Choi, F. C. Park, and Changmook C hun. Particle filtering on the Euclidean group: framework and applications. Robotica, 25(6):725–737, November 2007

    work page 2007

  14. [14]

    Brownian motion, martingales, and stochastic calculus

    Jean-Fran¸ cois Le Gall. Brownian motion, martingales, and stochastic calculus . Springer, Cham, 2016

    work page 2016

  15. [15]

    J. Lee. Introduction to Riemannian manifolds . Springer, Cham, 2018

    work page 2018

  16. [16]

    Simon J. A. Malham and Anke Wiese. Stochastic Lie group i ntegrators. SIAM J. Sci. Comput. , 30(2):597–617, 2008

    work page 2008

  17. [17]

    Numerical methods fo r stochastic differential equations in matrix Lie groups made simple

    Goran Marjanovic and Victor Solo. Numerical methods fo r stochastic differential equations in matrix Lie groups made simple. IEEE Trans. Autom. Control , 63(12):4035–4050, 2018. 35

    work page 2018

  18. [18]

    Continuous Markov processes and stochastic equations

    Gisirˆ o Maruyama. Continuous Markov processes and stochastic equations. Rend. Circ. Mat. Palermo. , 4:48–90, 1955

    work page 1955

  19. [19]

    Approximate integration of stochastic di fferential equations

    GN Milstein. Approximate integration of stochastic di fferential equations. Theory Probab. Appl. , 19(3):557–562, 1975

    work page 1975

  20. [20]

    A Lie algebraic approach to numerical i ntegration of stochastic differential equations

    Tetsuya Misawa. A Lie algebraic approach to numerical i ntegration of stochastic differential equations. SIAM J. Sci. Comput. , 23(3):866–890, 2001

    work page 2001

  21. [21]

    Higher strong order methods for linear Itˆ o SDEs on matrix Lie groups

    Michelle Muniz, Matthias Ehrhardt, Michael G¨ unther, and Renate Winkler. Higher strong order methods for linear Itˆ o SDEs on matrix Lie groups. BIT Numer. Math. , 62(4):1095–1119, 2022

    work page 2022

  22. [22]

    Strong stochastic Runge- Kutta–Munthe-Kaas methods for nonlinear Itˆ o SDEs on manifolds

    Michelle Muniz, Matthias Ehrhardt, Michael G¨ unther, and Renate Winkler. Strong stochastic Runge- Kutta–Munthe-Kaas methods for nonlinear Itˆ o SDEs on manifolds. Appl. Numer. Math , 193:196–203, 2023

    work page 2023

  23. [23]

    Runge-kutta methods on Lie groups

    Hans Munthe-Kaas. Runge-kutta methods on Lie groups. BIT Numer. Math. , 38:92–111, 1998

    work page 1998

  24. [24]

    High order Runge-Kutta methods on ma nifolds

    Hans Munthe-Kaas. High order Runge-Kutta methods on ma nifolds. Appl. Numer. Math. , 29(1):115– 127, 1999

    work page 1999

  25. [25]

    Riemannian neural SDE: Learning stochastic representations on manifolds

    Sung Woo Park, Hyomin Kim, Kyungjae Lee, and Junseok Kwo n. Riemannian neural SDE: Learning stochastic representations on manifolds. In NeurIPS, volume 35, pages 1434–1444, 2022

    work page 2022

  26. [26]

    Piggott and V

    M. Piggott and V. Solo. Geometric Euler–Maruyama schem es for stochastic differential equations in SO(n) and SE(n). SIAM J. Numer. Anal. , 54(4):2490–2516, 2016

    work page 2016

  27. [27]

    Rabier and Werner C

    Patrick J. Rabier and Werner C. Rheinboldt. On the numer ical solution of the Euler–Lagrange equa- tions. SIAM J. Numer. Anal. , 32(1):318–329, 1995

    work page 1995

  28. [28]

    Stochastic simulation

    Brian D Ripley. Stochastic simulation . John Wiley & Sons, New York, NY, 2009

    work page 2009

  29. [29]

    Strong and weak approximation method s for stochastic differential equations—some recent developments

    Andreas R¨ oßler. Strong and weak approximation method s for stochastic differential equations—some recent developments. In Recent Developments in Applied Probability and Statistics : Dedicated to the Memory of J¨ urgen Lehn, pages 127–153. Springer, Heidelberg, 2010

    work page 2010

  30. [30]

    Soatto, R

    S. Soatto, R. Frezza, and P. Perona. Motion estimation v ia dynamic vision. IEEE Trans. Autom. Control,, 41(3):393–413, 1996

    work page 1996

  31. [31]

    Stratonovich, Itˆ o and numerical analysi s on Lie groups

    Victor Solo. Stratonovich, Itˆ o and numerical analysi s on Lie groups. In 2024 IEEE 63rd CDC , pages 4328–4333, 2024

    work page 2024

  32. [32]

    Tangent space parametrization for stochastic differential equations on SO(n)

    Xi Wang and Victor Solo. Tangent space parametrization for stochastic differential equations on SO(n). IEEE Control Syst. Lett. , 9:1369–1374, 2025

    work page 2025

  33. [33]

    Gl obal optimization with orthogonality constraints via stochastic diffusion on manifold

    Honglin Yuan, Xiaoyi Gu, Rongjie Lai, and Zaiwen Wen. Gl obal optimization with orthogonality constraints via stochastic diffusion on manifold. J. Sci. Comput. , 80(2):1139–1170, 2019. 36

    work page 2019