Picard Iteration for Parameter Estimation in Nonlinear Dynamic Models of Aircraft and Spacecraft

Aleksandr Talitckii; Matthew Peet

arxiv: 2604.12885 · v1 · submitted 2026-04-14 · 🧮 math.DS

Picard Iteration for Parameter Estimation in Nonlinear Dynamic Models of Aircraft and Spacecraft

Aleksandr Talitckii , Matthew Peet This is my paper

Pith reviewed 2026-05-10 14:14 UTC · model grok-4.3

classification 🧮 math.DS

keywords parameter estimationPicard mappingnonlinear dynamicsODE modelsspacecraft attitudeaircraft dynamicsgradient contractioninertia tensor

0 comments

The pith

A gradient contraction algorithm with the Picard mapping estimates parameters in nonlinear ODE models of spacecraft and aircraft dynamics without state derivatives.

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

The paper proposes using the Picard mapping to impose an integral constraint on solutions of parameterized ODEs for attitude dynamics. This allows parameter estimation from sampled outputs that may be noisy or partial, without computing derivatives. The constraint is enforced through a gradient contraction algorithm to optimize the estimates. It is tested on estimating the inertia tensor for a spacecraft with four control-moment gyros and the 28 control surface coefficients for an F/A-18 aircraft model. Readers would care because in-flight data for such systems is typically limited and imperfect, requiring robust identification methods.

Core claim

The Picard mapping serves as an integral constraint on the solution of the parameterized ODE, which is enforced by a gradient contraction algorithm to obtain optimal parameter estimates for nonlinear dynamic models, demonstrated by recovering the inertia tensor in a 4-CMG spacecraft model and 28 higher-order control surface coefficients in an F/A-18 aircraft model.

What carries the argument

The Picard mapping as an integral constraint on the ODE solution, used to couple sampled outputs and avoid derivative measurements, combined with the gradient contraction algorithm for optimization.

If this is right

The method works on realistic, nonlinear models with complex torque generation mechanisms like CMGs and control surfaces.
It applies to cases where states are not fully measurable and data is noisy or sparse.
Optimal estimates are found without needing to differentiate the data directly.
Successful application to both spacecraft and aircraft suggests broad utility in aerospace parameter identification.

Where Pith is reading between the lines

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

Such techniques could be extended to other nonlinear systems in engineering where derivative data is unavailable.
Real-time implementation might allow adaptive control during flight.
Further analysis could determine convergence rates for highly nonlinear cases.

Load-bearing premise

The Picard mapping remains a reliable constraint and the gradient contraction algorithm converges to accurate global parameter values even with noisy, sparsely sampled data and highly nonlinear ODEs.

What would settle it

Running the algorithm on simulated noisy and sparsely sampled data from the F/A-18 model with known true coefficients and checking if the estimated 28 coefficients match the true values within small error bounds would falsify the claim if they do not.

Figures

Figures reproduced from arXiv: 2604.12885 by Aleksandr Talitckii, Matthew Peet.

read the original abstract

The attitude dynamics of aircraft and spacecraft exhibit significantly nonlinear behaviour. In spacecraft, torque is generated through reaction wheels and control moment gyros. In aircraft, torque is generated using lift on control surfaces. In both cases, complex geometries, unique configurations, and internal/environmental changes imply that models must be identified, verified, and updated using in-flight experimental data. However, this data is often noisy, sparsely sampled, and partial in that modeled states may not be directly measurable. In this paper, we propose a method for estimating key parameters in realistic Ordinary Differential Equation (ODE) models of both spacecraft and aircraft dynamics. This method avoids the need to directly measure state derivatives by coupling sampled outputs using the Picard mapping -- an integral constraint on the solution of the parameterized ODE. This constraint is then enforced, and optimal parameter estimates are found using a gradient contraction algorithm. This algorithm is applied to well-studied models of spacecraft and aircraft motion. First, the algorithm is used to estimate the inertia tensor in a 4 control-moment gyro (CMG) model of spacecraft motion. Second, we estimate the 28 higher-order control surface coefficients in a model of the F/A-18 aircraft.

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.

Desk Editor's Note private letter to a colleague

Picard integral constraint plus gradient contraction for ODE parameter fitting in aerospace models, but no noise or convergence tests shown for the 28-parameter F/A-18 case.

read the letter

The paper's core move is to replace derivative measurements with the Picard integral mapping so that sampled outputs alone enforce the ODE solution constraint, then use a gradient contraction algorithm to recover parameters. They demonstrate this on the inertia tensor of a 4-CMG spacecraft model and on 28 higher-order control-surface coefficients for an F/A-18 aircraft model. That framing is useful because flight data really is noisy and sparse, and derivatives are often unavailable or unreliable. The approach stays within standard models from the literature and avoids explicit differentiation, which is a practical step forward for this subfield. The math itself looks like a direct application of Picard iteration followed by contraction steps; nothing obviously broken in the setup or the way the constraint is written. Citations cover the usual spacecraft and aircraft dynamics references plus basic fixed-point theory, so the background is solid. The main gap is evidence. The abstract and description give the algorithm and the two target models but no error metrics, no Monte-Carlo noise trials, no basin-of-attraction checks, and no head-to-head numbers against shooting or multiple-shooting baselines. For the 28-parameter F/A-18 case especially, it is not clear whether the contraction remains reliable when the vector field is strongly nonlinear and measurements are sparse. If the full paper contains those quantitative checks, the claim strengthens; without them the practical reliability stays unproven. This is aimed at aerospace control engineers who do model identification from flight data. A reader already working on nonlinear system ID might borrow the integral-constraint trick if the numerics hold up. I would send it to peer review so the authors can supply the missing robustness tests and a referee can check the contraction analysis in detail.

Referee Report

2 major / 1 minor

Summary. The paper claims that the Picard mapping provides an integral constraint allowing parameter estimation in nonlinear ODE models of spacecraft and aircraft dynamics without direct state-derivative measurements, and that a gradient contraction algorithm can then be used to recover accurate global estimates. This is demonstrated by estimating the inertia tensor in a 4-CMG spacecraft model and the 28 higher-order control-surface coefficients in an F/A-18 aircraft model.

Significance. If the gradient contraction step can be shown to converge reliably, the approach would supply a derivative-free alternative for system identification in aerospace dynamics when data are noisy and sparse, complementing existing shooting methods.

major comments (2)

[Applications to Spacecraft and Aircraft Models] The central claim that the method produces accurate global parameter estimates for the 28-parameter F/A-18 case rests on the unproven assumption that the Picard mapping remains contractive and the gradient steps reach the global minimizer under realistic sensor noise and sparse sampling; no Monte-Carlo noise study, basin-of-attraction analysis, or comparison against multiple-shooting baselines is supplied to support this.
[Gradient Contraction Algorithm] No convergence proof or contraction-mapping analysis is given for the gradient contraction algorithm when the underlying vector field is highly nonlinear, which is load-bearing for the method's validity on both the 4-CMG and F/A-18 examples.

minor comments (1)

The abstract would benefit from a concise statement of any theoretical guarantees (or lack thereof) and the precise quantitative metrics used to assess success on the two examples.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for their constructive comments. We address the major concerns below, indicating revisions where appropriate. We believe the numerical demonstrations in the manuscript provide supporting evidence for the method's applicability, though we acknowledge limitations in theoretical guarantees.

read point-by-point responses

Referee: [Applications to Spacecraft and Aircraft Models] The central claim that the method produces accurate global parameter estimates for the 28-parameter F/A-18 case rests on the unproven assumption that the Picard mapping remains contractive and the gradient steps reach the global minimizer under realistic sensor noise and sparse sampling; no Monte-Carlo noise study, basin-of-attraction analysis, or comparison against multiple-shooting baselines is supplied to support this.

Authors: We agree that the manuscript would benefit from additional empirical validation. In the revised version, we will add a Monte-Carlo study examining the effect of sensor noise and sparse sampling on the F/A-18 parameter estimates. We will also include a comparison with a multiple-shooting baseline for the 4-CMG case. A full basin-of-attraction analysis for the high-dimensional 28-parameter problem is computationally prohibitive at this stage and will be noted as a direction for future work. revision: partial
Referee: [Gradient Contraction Algorithm] No convergence proof or contraction-mapping analysis is given for the gradient contraction algorithm when the underlying vector field is highly nonlinear, which is load-bearing for the method's validity on both the 4-CMG and F/A-18 examples.

Authors: The gradient contraction algorithm is presented as a practical method that leverages the contractive properties of the Picard mapping in the examples considered. While we do not provide a general convergence proof for arbitrary nonlinear systems, the manuscript demonstrates reliable convergence in the specific aerospace models through numerical experiments. A theoretical analysis for highly nonlinear cases is beyond the current scope but could be explored in subsequent research. revision: no

standing simulated objections not resolved

A general convergence proof for the gradient contraction algorithm under highly nonlinear dynamics.

Circularity Check

0 steps flagged

No significant circularity; method is an independent algorithmic construction

full rationale

The paper introduces the Picard mapping as an integral constraint derived directly from the ODE solution and couples it with a gradient contraction algorithm to enforce the constraint for parameter estimation. This construction is applied to externally defined, standard models (4-CMG spacecraft inertia tensor and 28-coefficient F/A-18 aerodynamics) without reducing any claimed result to a fitted input by definition or to a self-citation chain. No load-bearing step equates a prediction to its own inputs; the algorithm is presented as a self-contained alternative to direct differentiation of noisy data. The absence of Monte-Carlo validation or convergence proofs under noise is a limitation on reliability, not a circularity in the derivation.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard existence/uniqueness results for ODEs and on the unproven convergence of the custom gradient contraction procedure under realistic noise.

free parameters (1)

Contraction step-size or regularization parameter
The gradient contraction algorithm almost certainly requires at least one tunable scalar whose value is not derived from first principles.

axioms (1)

standard math Solutions to the parameterized ODE exist and are unique (Picard-Lindelöf theorem).
Invoked to guarantee that the Picard mapping is well-defined for any candidate parameter vector.

pith-pipeline@v0.9.0 · 5508 in / 1060 out tokens · 57069 ms · 2026-05-10T14:14:30.347582+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

14 extracted references · 14 canonical work pages

[1]

Perspectives on system identification,

L. Ljung, “Perspectives on system identification,”Annual Reviews in Control, vol. 34, no. 1, pp. 1–12, 2010

work page 2010
[2]

Approximate solution of the trust region problem by minimization over two-dimensional subspaces,

R. H. Byrd, R. B. Schnabel, and G. A. Shultz, “Approximate solution of the trust region problem by minimization over two-dimensional subspaces,”Mathematical Programming, vol. 40, no. 1-3, pp. 247– 263, 1988

work page 1988
[3]

The Levenberg-Marquardt algorithm: implementation and theory,

J. J. Mor ´e, “The Levenberg-Marquardt algorithm: implementation and theory,” inProceedings of the Biennial Conference held at Dundee,

work page
[4]

Springer, 2006, pp. 105–116

work page 2006
[5]

An accurate method for real-time aircraft dynamics simulation based on predictor-corrector scheme,

J. Zhao, H. Wang, and H. Zhang, “An accurate method for real-time aircraft dynamics simulation based on predictor-corrector scheme,” Mathematical Problems in Engineering, vol. 2015, no. 1, p. 193179, 2015

work page 2015
[6]

Picard iteration for parameter esti- mation in nonlinear ordinary differential equations,

A. Talitckii and M. M. Peet, “Picard iteration for parameter esti- mation in nonlinear ordinary differential equations,”arXiv preprint arXiv:2412.20216, 2024

work page arXiv 2024
[7]

Feedback control law for variable speed control moment gyros,

H. Schaub, S. R. Vadali, and J. L. Junkins, “Feedback control law for variable speed control moment gyros,”The Journal of the astronautical sciences, vol. 46, no. 3, pp. 307–328, 1998

work page 1998
[8]

Estimation of the longitudinal aerodynamic parameters from flight data for the NASA F/A-18 HARV,

M. Napolitano, A. Paris, B. Seanor, and A. Bowers, “Estimation of the longitudinal aerodynamic parameters from flight data for the NASA F/A-18 HARV,” in21st Atmospheric Flight Mechanics Conference, 1996, p. 3419

work page 1996
[9]

D. G. Luenberger,Optimization by Vector Space Methods. John Wiley & Sons, 1997

work page 1997
[10]

V . I. Arnold,Ordinary Differential Equations. Springer Science & Business Media, 1992

work page 1992
[11]

A converse sum-of-squares Lyapunov result: an existence proof based on the Picard iteration,

M. M. Peet and A. Papachristodoulou, “A converse sum-of-squares Lyapunov result: an existence proof based on the Picard iteration,” in 49th IEEE Conference on Decision and Control (CDC). IEEE, 2010, pp. 5949–5954

work page 2010
[12]

Directional singularity escape and avoidance for single-gimbal control moment gyroscopes,

L. Valk, A. Berry, and H. Vallery, “Directional singularity escape and avoidance for single-gimbal control moment gyroscopes,”Journal of Guidance, Control, and Dynamics, vol. 41, no. 5, pp. 1095–1107, 2018

work page 2018
[13]

A comparative study of real-time aircraft parameter identification schemes applied to NASA F/A-18 HARV flight data,

Y . Song and M. Napolitano, “A comparative study of real-time aircraft parameter identification schemes applied to NASA F/A-18 HARV flight data,”Transactions of the Japan Society for Aeronautical and Space Sciences, vol. 45, no. 149, pp. 180–188, 2002

work page 2002
[14]

Susceptibility of F/A-18 flight controllers to the falling-leaf mode: nonlinear analysis,

A. Chakraborty, P. Seiler, and G. J. Balas, “Susceptibility of F/A-18 flight controllers to the falling-leaf mode: nonlinear analysis,”Journal of Guidance, Control, and Dynamics, vol. 34, no. 1, pp. 73–85, 2011

work page 2011

[1] [1]

Perspectives on system identification,

L. Ljung, “Perspectives on system identification,”Annual Reviews in Control, vol. 34, no. 1, pp. 1–12, 2010

work page 2010

[2] [2]

Approximate solution of the trust region problem by minimization over two-dimensional subspaces,

R. H. Byrd, R. B. Schnabel, and G. A. Shultz, “Approximate solution of the trust region problem by minimization over two-dimensional subspaces,”Mathematical Programming, vol. 40, no. 1-3, pp. 247– 263, 1988

work page 1988

[3] [3]

The Levenberg-Marquardt algorithm: implementation and theory,

J. J. Mor ´e, “The Levenberg-Marquardt algorithm: implementation and theory,” inProceedings of the Biennial Conference held at Dundee,

work page

[4] [4]

Springer, 2006, pp. 105–116

work page 2006

[5] [5]

An accurate method for real-time aircraft dynamics simulation based on predictor-corrector scheme,

J. Zhao, H. Wang, and H. Zhang, “An accurate method for real-time aircraft dynamics simulation based on predictor-corrector scheme,” Mathematical Problems in Engineering, vol. 2015, no. 1, p. 193179, 2015

work page 2015

[6] [6]

Picard iteration for parameter esti- mation in nonlinear ordinary differential equations,

A. Talitckii and M. M. Peet, “Picard iteration for parameter esti- mation in nonlinear ordinary differential equations,”arXiv preprint arXiv:2412.20216, 2024

work page arXiv 2024

[7] [7]

Feedback control law for variable speed control moment gyros,

H. Schaub, S. R. Vadali, and J. L. Junkins, “Feedback control law for variable speed control moment gyros,”The Journal of the astronautical sciences, vol. 46, no. 3, pp. 307–328, 1998

work page 1998

[8] [8]

Estimation of the longitudinal aerodynamic parameters from flight data for the NASA F/A-18 HARV,

M. Napolitano, A. Paris, B. Seanor, and A. Bowers, “Estimation of the longitudinal aerodynamic parameters from flight data for the NASA F/A-18 HARV,” in21st Atmospheric Flight Mechanics Conference, 1996, p. 3419

work page 1996

[9] [9]

D. G. Luenberger,Optimization by Vector Space Methods. John Wiley & Sons, 1997

work page 1997

[10] [10]

V . I. Arnold,Ordinary Differential Equations. Springer Science & Business Media, 1992

work page 1992

[11] [11]

A converse sum-of-squares Lyapunov result: an existence proof based on the Picard iteration,

M. M. Peet and A. Papachristodoulou, “A converse sum-of-squares Lyapunov result: an existence proof based on the Picard iteration,” in 49th IEEE Conference on Decision and Control (CDC). IEEE, 2010, pp. 5949–5954

work page 2010

[12] [12]

Directional singularity escape and avoidance for single-gimbal control moment gyroscopes,

L. Valk, A. Berry, and H. Vallery, “Directional singularity escape and avoidance for single-gimbal control moment gyroscopes,”Journal of Guidance, Control, and Dynamics, vol. 41, no. 5, pp. 1095–1107, 2018

work page 2018

[13] [13]

A comparative study of real-time aircraft parameter identification schemes applied to NASA F/A-18 HARV flight data,

Y . Song and M. Napolitano, “A comparative study of real-time aircraft parameter identification schemes applied to NASA F/A-18 HARV flight data,”Transactions of the Japan Society for Aeronautical and Space Sciences, vol. 45, no. 149, pp. 180–188, 2002

work page 2002

[14] [14]

Susceptibility of F/A-18 flight controllers to the falling-leaf mode: nonlinear analysis,

A. Chakraborty, P. Seiler, and G. J. Balas, “Susceptibility of F/A-18 flight controllers to the falling-leaf mode: nonlinear analysis,”Journal of Guidance, Control, and Dynamics, vol. 34, no. 1, pp. 73–85, 2011

work page 2011