pith. sign in

arxiv: 2605.17146 · v1 · pith:GJG2I4POnew · submitted 2026-05-16 · 💻 cs.CE · cs.LG· cs.SY· eess.SY

Weighted Flow Matching and Physics-Informed Nonlinear Filtering for Parameter Estimation in Digital Twins

Yasar Yanik , Himadri Basu , Ricardo G. Sanfelice , Daniele Venturi This is my paper

Pith reviewed 2026-05-20 14:07 UTC · model grok-4.3

classification 💻 cs.CE cs.LGcs.SYeess.SY
keywords digital twinsparameter estimationweighted flow matchingunscented Kalman filternonlinear filteringspacecraft dynamicsphysics-informed modeling
0
0 comments X

The pith

A framework pairing weighted flow matching with unscented Kalman filtering delivers stable parameter estimates for digital twins under noisy conditions.

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

The paper sets out to show that digital-twin synchronization can be improved by letting a generative model learn probability transport while a physics-based filter enforces consistency with known dynamics. Digital twins must track changing parameters in real time, yet low observability, weak signals, and measurement noise often defeat standard estimators. The method reweights training samples on the fly so the generative component focuses on parameter values most relevant to the current state, then feeds those distributions into an unscented Kalman filter that respects the underlying equations. Demonstrated on a spacecraft model, the combined system produces steadier estimates of moment of inertia than either extended or ensemble Kalman filters alone.

Core claim

The authors claim that dynamic reweighting inside weighted flow matching guides the generative model toward parameter regimes most informative of the evolving state, and that tightly coupling this component with a physics-informed unscented Kalman filter yields a unified framework whose state and parameter estimates remain physically consistent even when sensing is uncertain and noisy.

What carries the argument

Weighted flow matching that applies dynamic reweighting of training samples to steer the generative model toward the most informative parameter regimes, then couples the resulting distributions to an unscented Kalman filter for physically consistent estimation.

If this is right

  • Stable online estimation of moment of inertia becomes possible for spacecraft even with noisy or biased sensors.
  • The same coupling of weighted generative modeling and unscented filtering can be applied to other nonlinear systems that require joint state and parameter tracking.
  • Real-time digital-twin synchronization improves without sacrificing physical consistency.
  • Performance gains over extended and ensemble Kalman filters appear in low-observability regimes.

Where Pith is reading between the lines

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

  • The reweighting idea could be tested on hardware-in-the-loop spacecraft simulators to check whether simulation gains survive sensor bias and latency.
  • Similar weighting schemes might improve other generative models used inside Kalman-type filters for chemical-process or biomedical digital twins.
  • If the method scales, it could reduce the need for high-fidelity excitation signals during system identification.
  • The framework suggests a broader pattern: let data-driven transport handle uncertainty while physics-based filters enforce conservation laws.

Load-bearing premise

Dynamic reweighting of training samples will reliably steer the generative model toward the parameter values that best describe the current physical state.

What would settle it

In the spacecraft digital-twin simulation, removing the dynamic reweighting step and finding that moment-of-inertia error remains as large as or larger than the error produced by a plain unscented Kalman filter would falsify the performance claim.

Figures

Figures reproduced from arXiv: 2605.17146 by Daniele Venturi, Himadri Basu, Ricardo G. Sanfelice, Yasar Yanik.

Figure 1
Figure 1. Figure 1: FIGURE 1 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: FIGURE 2 [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: FIGURE 3 [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIGURE 4 [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
read the original abstract

Digital twins (DTs) rely on continuous synchronization between physical systems and their virtual counterparts through online parameter estimation under uncertainty. In many practical settings, however, this task is challenged by low observability, weak excitation, nonlinear dynamics, and noisy or biased measurements. In this work, we develop a new mathematical framework that integrates Weighted Flow Matching (WFM) generative modeling with physics-informed nonlinear filtering to enhance parameter estimation in DTs. WFM relies on dynamic reweighting of training samples, which guides the generative model toward parameter regimes most informative of the evolving system state. This generative component is tightly coupled with a physics-informed filtering architecture based on the Unscented Kalman Filter (UKF), yielding a unified DT framework that combines data-driven probability transport with physically consistent state and parameter estimation. The effectiveness of the new integrated framework is demonstrated within a spacecraft DT architecture, where stable moment of inertia estimation is achieved under uncertain and noisy sensing, with significant performance improvements over established approaches such as Extended Kalman Filtering (EKF) and Ensemble Kalman Filtering (EnKF). These results highlight the potential of weighted generative modeling as a core mechanism for real-time DT synchronization in operational and mission-critical systems.

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 / 2 minor

Summary. The paper claims to develop a new mathematical framework integrating Weighted Flow Matching (WFM) generative modeling with physics-informed nonlinear filtering based on the Unscented Kalman Filter (UKF) for enhanced parameter estimation in digital twins. Dynamic reweighting of training samples in WFM guides the generative model toward informative parameter regimes of the evolving system state; this component is tightly coupled with the UKF to produce a unified DT framework combining data-driven probability transport with physically consistent state and parameter estimation. Effectiveness is demonstrated in a spacecraft DT architecture achieving stable moment of inertia estimation under uncertain and noisy sensing, with significant performance improvements over EKF and EnKF.

Significance. If the integration and performance claims hold with rigorous validation, the work could advance digital-twin synchronization for nonlinear systems under low observability by using generative reweighting to inform filtering. The spacecraft demonstration provides a concrete test case for mission-critical applications, and the emphasis on dynamic reweighting offers a potentially useful extension of flow-matching ideas to online estimation problems.

major comments (1)
  1. [§3 (framework integration)] §3 (framework integration) and the description of the 'tightly coupled' architecture: the central claim requires that WFM reweighting produces parameter samples that, when fed into the physics-informed UKF, yield stable inertia estimates superior to EKF/EnKF. However, no explicit update rule or interface is derived showing how generative samples modify filter covariances, augment the state vector, replace process noise, or supply a prior. Without this mapping it is impossible to verify that the unscented transform assumptions remain valid or that physical consistency is enforced by construction rather than observed empirically.
minor comments (2)
  1. [Abstract and results section] The abstract states 'significant performance improvements' without reporting specific quantitative metrics, error bars, or the exact simulation conditions; these details should be added to the results section for reproducibility.
  2. [Methods (WFM component)] Notation for the dynamic reweighting factors in WFM is introduced at a high level but lacks an explicit equation distinguishing it from standard flow-matching objectives; a dedicated equation would improve clarity.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive and detailed review of our manuscript. The comment on the framework integration in §3 is well taken, and we address it directly below while committing to a revision that strengthens the presentation of the coupling without altering the core claims.

read point-by-point responses

  1. Referee: [§3 (framework integration)] §3 (framework integration) and the description of the 'tightly coupled' architecture: the central claim requires that WFM reweighting produces parameter samples that, when fed into the physics-informed UKF, yield stable inertia estimates superior to EKF/EnKF. However, no explicit update rule or interface is derived showing how generative samples modify filter covariances, augment the state vector, replace process noise, or supply a prior. Without this mapping it is impossible to verify that the unscented transform assumptions remain valid or that physical consistency is enforced by construction rather than observed empirically.

    Authors: We agree that an explicit derivation of the interface is necessary for rigorous verification. The current manuscript describes the coupling at the architectural level but does not provide the step-by-step update rules. In the revised version we will add a dedicated subsection in §3 that derives the mapping as follows: WFM reweighted samples augment the UKF state vector with the parameter (moment of inertia) and its uncertainty; the reweighting factor, computed from the current UKF posterior, directly scales the process-noise covariance matrix to reflect parameter variability; the generative distribution supplies the prior mean and covariance for the prediction step, which is then propagated through the unscented transform. Because the WFM objective already incorporates the physics-informed residual, the resulting prior is consistent with the nonlinear dynamics by construction. We will also include an algorithmic box and a short proof sketch confirming that the sigma-point selection remains valid under the augmented state. These additions will allow readers to verify the assumptions without relying solely on empirical results. revision: yes

Circularity Check

0 steps flagged

No circularity: new integration of WFM and UKF presented as independent construction

full rationale

The manuscript introduces Weighted Flow Matching with dynamic reweighting as a generative component that is then coupled to a physics-informed UKF for digital-twin parameter estimation. The abstract and high-level description position this coupling as a novel unified framework whose performance gains are demonstrated empirically against EKF and EnKF baselines. No load-bearing equation or claim is shown to reduce by construction to a fitted parameter, a self-citation chain, or an ansatz imported from the authors' prior work; the central claims therefore remain self-contained and externally falsifiable.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the central claim rests on the effectiveness of dynamic sample reweighting and the assumption that the generative-filter coupling preserves physical consistency; no explicit free parameters, axioms, or invented entities are quantified.

free parameters (1)
  • dynamic reweighting factors
    Used in WFM to guide the model toward informative parameter regimes for the evolving state.
axioms (1)
  • domain assumption System dynamics are nonlinear with low observability, weak excitation, and noisy or biased measurements.
    Invoked to motivate the need for the new framework in digital twin synchronization.

pith-pipeline@v0.9.0 · 5761 in / 1272 out tokens · 77449 ms · 2026-05-20T14:07:52.240767+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

19 extracted references · 19 canonical work pages

  1. [1]

    Y anik, S

    Y . Y anik, S. Ekwaro-Osire, J. P . Dias, E. H. Porto, D. S. Alves, T. H. Machado, G. Bregion Daniel, H. F. de Castro, and K. L. Cavalca, ‘‘V er- ification and validation of rotating machinery using digital twin,’’ASCE- ASME Journal of Risk and Uncertainty in Engineering Systems, Part B: Mechanical Engineering, vol. 10, no. 1, p. 011104, 2024

    work page 2024

  2. [2]

    Y anik, ‘‘Applying digital twin in a rotating mechanical system and a photovoltaic system,’’ Ph.D

    Y . Y anik, ‘‘Applying digital twin in a rotating mechanical system and a photovoltaic system,’’ Ph.D. dissertation, Texas Tech University, 2024

    work page 2024

  3. [3]

    C. Yüce, O. Gecgel, O. Doğan, S. Dabetwar, Y . Y anik, O. C. Kalay, E. Karpat, and F. Karpat, ‘‘Prognostics and health management of wind energy infrastructure systems,’’ASCE-ASME Journal of Risk and Uncer- tainty in Engineering Systems, Part B: Mechanical Engineering, vol. 8, no. 2, p. 020801, 2022. 12 VOLUME 1 , 2026 Yaniket al.: Weighted Flow Matching ...

    work page 2022

  4. [4]

    Y anik, ‘‘Quantification of parametric uncertainties effects in structural failure criteria,’’ Ph.D

    Y . Y anik, ‘‘Quantification of parametric uncertainties effects in structural failure criteria,’’ Ph.D. dissertation, Universidade Estadual Paulista (Un- esp), 2019

    work page 2019

  5. [5]

    Y anik, S

    Y . Y anik, S. da Silva, and A. Cunha Jr, ‘‘Uncertainty quantification in the comparison of structural criterions of failure,’’ inProceedings of the X Congresso Nacional de Engenharia Mecânica (CONEM 2018), 2018

    work page 2018

  6. [6]

    Henao-Garcia, M

    S. Henao-Garcia, M. Kapteyn, K. E. Willcox, M. Tezzele, M. Castroviejo- Fernandez, T. Kim, M. Ambrosino, I. Kolmanovsky, H. Basu, P . Jir- wankaret al., ‘‘Digital-twin-enabled multi-spacecraft on-orbit operations,’’ inAIAA SCITECH 2025 F orum, 2025, p. 1432

    work page 2025

  7. [7]

    Fernandes and V

    G. Fernandes and V . Maldonado, ‘‘The u.s. air force next-generation air-refueling system: A resurgence of the blended wing body?’’Aerospace, vol. 11, no. 6, p. 494, Jun. 2024. [Online]. Available: http://dx.doi.org/10. 3390/aerospace11060494

    work page 2024

  8. [8]

    Lipman, R

    Y . Lipman, R. T. Q. Chen, H. Ben-Hamu, M. Nickel, and M. Le, ‘‘Flow matching for generative modeling,’’ inProc. International Conference on Learning Representations (ICLR), 2023

    work page 2023

  9. [9]

    J. Ho, A. Jain, and P . Abbeel, ‘‘Denoising diffusion probabilistic models,’’ inAdvances in Neural Information Processing Systems, vol. 33, 2020, pp. 6840–6851

    work page 2020

  10. [10]

    Y . Song, J. Sohl-Dickstein, D. P . Kingma, A. Kumar, S. Ermon, and B. Poole, ‘‘Score-based generative modeling through stochastic differential equations,’’ inProc. International Conference on Learning Representa- tions (ICLR), 2021

    work page 2021

  11. [11]

    Y . Liu, Y . Chen, D. Xiu, and G. Zhang, ‘‘A training-free conditional diffusion model for learning stochastic dynamical systems,’’SIAM Journal on Scientific Computing, vol. 47, no. 5, pp. C1144–C1171, 2025

    work page 2025

  12. [12]

    M. Ren, W. Zeng, B. Y ang, and R. Urtasun, ‘‘Learning to reweight ex- amples for robust deep learning,’’ inProceedings of the International Conference on Machine Learning (ICML). PMLR, 2018, pp. 4334–4343

    work page 2018

  13. [13]

    Flores-Abad, O

    A. Flores-Abad, O. Ma, K. Pham, and S. Ulrich, ‘‘A review of space robotics technologies for on-orbit servicing,’’Progress in Aerospace Sci- ences, vol. 68, pp. 1–26, 2014

    work page 2014

  14. [14]

    Q. Meng, J. Liang, and O. Ma, ‘‘Identification of all the inertial parameters of a non-cooperative object in orbit,’’Aerospace Science and Technology, vol. 91, pp. 571–582, 2019

    work page 2019

  15. [15]

    M. D. Lichter and S. Dubowsky, ‘‘Estimation of state, shape, and in- ertial parameters of space objects from sequences of range images,’’ in Proceedings of Intelligent Robots and Computer Vision XXI: Algorithms, Techniques, and Active Vision, vol. 5267. SPIE, 2003, pp. 194–205

    work page 2003

  16. [16]

    Y anik, ‘‘BoostedUKF,’’ https://github.com/HybridSystemsLab/ BoostedUKF, 2026

    Y . Y anik, ‘‘BoostedUKF,’’ https://github.com/HybridSystemsLab/ BoostedUKF, 2026

    work page 2026

  17. [17]

    G. Dufflis Fernandes, ‘‘Design of a stealth, flying wing bomber aircraft using vortex panel methods,’’ Doctoral dissertation, Texas Tech University, Lubbock, Texas, December 2024, accessed: 2025-01-15. [Online]. Available: https://hdl.handle.net/2346/100823

    work page 2024

  18. [18]

    H. Basu, M. C. Fernandez, R. G. Sanfelice, and I. Kolmanovsky, ‘‘Hybrid model predictive control approach for spacecraft proximity maneuvering and docking accounting for collisions,’’ in2025 American Control Confer- ence (ACC). IEEE, 2025, pp. 4954–4959

    work page 2025

  19. [19]

    E. A. Wan and R. V an Der Merwe, ‘‘The unscented kalman filter for nonlinear estimation,’’ inProceedings of the IEEE 2000 Adaptive Systems for Signal Processing, Communications, and Control Symposium. IEEE, 2000, pp. 153–158. YASAR YANIKreceived the B.S. degree in me- chanical engineering from Istanbul Technical Uni- versity, Istanbul, Turkey, in 2016, th...

    work page 2000