pith. sign in

arxiv: 2510.21975 · v2 · submitted 2025-10-24 · 🧮 math.OC

Convex Bound of Nonlinear Dynamical Errors for Covariance Steering

Daniel C. Qi , Kenshiro Oguri This is my paper

Pith reviewed 2026-05-18 04:02 UTC · model grok-4.3

classification 🧮 math.OC
keywords convex optimizationcovariance steeringnonlinear dynamicslinearization remainderTaylor expansionhalo orbitstationkeepingGaussian distribution
0
0 comments X

The pith

A convex upper bound on the Taylor remainder lets linear covariance controllers preserve Gaussian state distributions in nonlinear regimes.

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

The paper develops a method to minimize the nonlinear dynamical errors that arise when linear controllers are applied to nonlinear systems such as those in cislunar space. It constructs an upper bound on the remainder term left after linearization by using higher-order terms from the Taylor series and converts that bound into a convex function. The convex function is then minimized as a cost when choosing controller gains. In halo-orbit stationkeeping simulations the resulting gains keep the closed-loop state distribution closer to Gaussian than gains chosen without the bound.

Core claim

The formulation upper-bounds the remainder term from the linearization process using higher-order terms in a Taylor series expansion, resolves the bound into a convex function, and employs that function as a cost in the gain optimization so that the linear covariance controller experiences smaller nonlinear errors and maintains Gaussianity more effectively in nonlinear simulations.

What carries the argument

The convex upper bound on the nonlinear remainder term obtained from higher-order Taylor terms, serving as the cost function for gain optimization.

If this is right

  • Controller gains can be found efficiently by convex optimization while still accounting for nonlinear effects.
  • The closed-loop state distribution stays closer to the assumed Gaussian shape in nonlinear environments.
  • Linear covariance steering becomes usable over wider regions without requiring full nonlinear propagation.
  • The same bounding technique can be inserted into other linear-controller designs that rely on covariance propagation.

Where Pith is reading between the lines

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

  • The bound could be recomputed online if the reference trajectory changes, allowing the optimizer to trade off gain adjustment against reference adjustment.
  • In systems where the operating region is larger than the local validity of the Taylor expansion, the method may naturally push the trajectory toward flatter regions of the dynamics.
  • The same convex-cost idea might be applied to other linearization-based estimators or planners that currently ignore higher-order terms.

Load-bearing premise

The convex upper bound on the Taylor remainder stays tight enough across the operating region that minimizing it actually reduces the true nonlinear error rather than merely trading one source of mismatch for another.

What would settle it

A halo-orbit stationkeeping simulation in which the gains obtained by minimizing the proposed convex bound produce equal or larger deviation from Gaussianity than gains obtained from the standard linearization without the bound.

read the original abstract

Applying linear controllers to nonlinear systems requires the dynamical linearization about a reference. In highly nonlinear environments such as cislunar space, the region of validity for these linearizations varies widely and can negatively affect controller performance if not carefully formulated. This paper presents a formulation that minimizes the nonlinear errors experienced by linear covariance controllers. The formulation involves upper-bounding the remainder term from the linearization process using higher-order terms in a Taylor series expansion, and resolving it into a convex function. This can serve as a cost function for controller gain optimization, and its convex nature allows for efficient solutions through convex optimization. This formulation is then demonstrated and compared with the current methods within a halo orbit stationkeeping scenario. The results show that the formulation proposed in this paper maintains the Gaussianity of the distribution in nonlinear simulations more effectively, thereby allowing the linear covariance controller to perform more as intended in nonlinear environments.

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

2 major / 2 minor

Summary. The paper proposes constructing a convex upper bound on the nonlinear remainder term arising from Taylor linearization of the dynamics, then using this bound as a cost in a convex optimization problem to select gains for covariance-steering controllers. The resulting controller is claimed to maintain closer adherence to the linear-Gaussian assumption in closed-loop nonlinear simulations than standard linearization-based methods, with the claim supported by a halo-orbit stationkeeping numerical example.

Significance. If the bound proves sufficiently tight along closed-loop trajectories, the approach would supply a computationally tractable surrogate for nonlinear error that preserves the convexity and Gaussianity assumptions underlying covariance steering, offering a practical bridge between linear control theory and highly nonlinear regimes such as cislunar stationkeeping.

major comments (2)
  1. [§3.2] §3.2, Eq. (12)–(15): the convex relaxation of the Taylor remainder (via norm bounds or SDP lifting) is presented without a subsequent tightness analysis or a posteriori verification that the minimized bound correlates with actual pointwise or distributional deviation from the linear model along the optimized trajectories.
  2. [§5] §5, Table 2 and Fig. 7: the reported improvement in Gaussianity (e.g., reduced Kullback-Leibler divergence or covariance mismatch) is shown relative to baseline linear controllers, yet no column or subplot quantifies the realized gap between the proposed convex upper bound and the true nonlinear remainder evaluated on the same closed-loop data; this leaves open whether the performance gain is attributable to the bound or to incidental retuning effects.
minor comments (2)
  1. [§2–§3] The notation for the higher-order remainder tensor R(x) is introduced in §2 but its precise contraction with the state deviation vector is not restated when the bound is formed in §3, forcing the reader to cross-reference.
  2. [Fig. 6] Figure 6 (distribution snapshots) would be clearer if the plotted ellipses were accompanied by the numerical value of the optimized bound at the corresponding time step.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their insightful comments, which help improve the clarity and rigor of our work on convex bounds for nonlinear dynamical errors in covariance steering. We address each major comment point by point below.

read point-by-point responses

  1. Referee: [§3.2] §3.2, Eq. (12)–(15): the convex relaxation of the Taylor remainder (via norm bounds or SDP lifting) is presented without a subsequent tightness analysis or a posteriori verification that the minimized bound correlates with actual pointwise or distributional deviation from the linear model along the optimized trajectories.

    Authors: We agree that a tightness analysis and a posteriori verification would strengthen the presentation. The manuscript focuses on deriving the convex bound and demonstrating its use in optimization, but does not explicitly verify the correlation along trajectories. In the revision, we will include an analysis of the bound's tightness using higher-order Taylor terms and add numerical results from the halo-orbit simulations showing the gap between the bound and the true remainder. This will be added as a new paragraph in §3.2. revision: yes

  2. Referee: [§5] §5, Table 2 and Fig. 7: the reported improvement in Gaussianity (e.g., reduced Kullback-Leibler divergence or covariance mismatch) is shown relative to baseline linear controllers, yet no column or subplot quantifies the realized gap between the proposed convex upper bound and the true nonlinear remainder evaluated on the same closed-loop data; this leaves open whether the performance gain is attributable to the bound or to incidental retuning effects.

    Authors: The referee raises a valid point regarding attribution of the observed improvements. While the results show better adherence to Gaussianity compared to baselines, we did not directly quantify the bound-true remainder gap in the reported tables and figures. To resolve this, we will revise §5 to include additional metrics in Table 2 and a new subplot in Fig. 7 that evaluates the actual nonlinear remainder on the closed-loop data for both the proposed and baseline controllers. This will provide evidence that the performance gains stem from the minimization of the convex bound. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation uses standard Taylor remainder bounding and convex relaxation.

full rationale

The paper constructs an upper bound on the nonlinear remainder term via higher-order Taylor series terms, relaxes the bound to convexity, and employs the result as a cost in gain optimization for covariance steering. This chain rests on classical analytic expansions and standard convex optimization techniques rather than parameter fitting to validation data, self-referential definitions, or load-bearing self-citations. The halo-orbit stationkeeping demonstration compares closed-loop performance but does not reduce the claimed bound to a tautology with the simulation outputs or prior author results. The derivation therefore remains self-contained against external mathematical benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach rests on the standard Taylor theorem with remainder, the assumption that the remainder can be majorized by a convex function of the gains, and the modeling choice that preserving Gaussianity in simulation is the right proxy for controller quality. No new physical entities or free parameters are introduced in the abstract.

axioms (2)
  • standard math Taylor expansion with Lagrange remainder exists and is differentiable enough for the bound construction
    Invoked when the paper states it upper-bounds the remainder term from the linearization process using higher-order terms.
  • domain assumption The resulting upper bound is convex in the decision variables (controller gains)
    Required for the claim that the formulation can be solved efficiently through convex optimization.

pith-pipeline@v0.9.0 · 5677 in / 1417 out tokens · 26552 ms · 2026-05-18T04:02:45.094717+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

  1. [1]

    Optimal Linear Covariance Steering with Minimum Nonlinear Dynamical Errors,

    Qi, D. C., and Oguri, K., “Optimal Linear Covariance Steering with Minimum Nonlinear Dynamical Errors,”AAS/AIAA Astrodynamics Specialist Conference, 2025, pp. 1–20. 19

    work page 2025

  2. [2]

    Identifying Limits of Linear Control Design Validity in Nonlinear Systems: A Continuation-based Approach,

    Nguyen, D. H., Lowenberg, M. H., and Neild, S. A., “Identifying Limits of Linear Control Design Validity in Nonlinear Systems: A Continuation-based Approach,”Nonlinear Dynamics, Vol. 104, 2020, p. 901–921

    work page 2020

  3. [3]

    Research on Gain Scheduling,

    Rugh, W. J., and Shamma, J. S., “Research on Gain Scheduling,”Automatica, Vol. 36, No. 10, 2000, pp. 1401–1425

    work page 2000

  4. [4]

    Covariance Control Theory,

    Hotz, A., and Skelton, R. E., “Covariance Control Theory,”International Journal of Control, Vol. 46, No. 1, 1986, pp. 13–32

    work page 1986

  5. [5]

    Optimal Covariance Control for Stochastic Systems Under Chance Constraints,

    Okamoto, K., Goldshtein, M., and Tsiotras, P., “Optimal Covariance Control for Stochastic Systems Under Chance Constraints,” IEEE Control Systems Letters, Vol. 2, No. 2, 2018, pp. 266–271

    work page 2018

  6. [6]

    Chance-ConstrainedCovarianceControlforLow-thrustMinimum-FuelTrajectory Optimization,

    Ridderhof,J.,Pilipovsky,J.,andTsiotras,P.,“Chance-ConstrainedCovarianceControlforLow-thrustMinimum-FuelTrajectory Optimization,”AIAA/AAS Astrodynamics Specialists Conference, 2021, pp. 1–20

    work page 2021

  7. [7]

    Robust Cislunar Low-Thrust Trajectory Optimization under Uncertainties via Sequential Covariance Steering,

    Kumagai, N., and Oguri, K., “Robust Cislunar Low-Thrust Trajectory Optimization under Uncertainties via Sequential Covariance Steering,”Journal of Guidance, Control, and Dynamics, 2025, pp. 1–19

    work page 2025

  8. [8]

    Trajectory Optimization Under Uncertainty with Nonlinear Programming and Forward–Backward Shooting,

    Varghese, J., and Oguri, K., “Trajectory Optimization Under Uncertainty with Nonlinear Programming and Forward–Backward Shooting,”Journal of Guidance, Control, and Dynamics, 2025, pp. 1–19

    work page 2025

  9. [9]

    Convex Approach to Covariance Control with Application to Stochastic Low-Thrust Trajectory Optimization,

    Benedikter, B., Zavoli, A., Wang, Z., Pizzurro, S., and Cavallini, E., “Convex Approach to Covariance Control with Application to Stochastic Low-Thrust Trajectory Optimization,”Journal of Guidance, Control, and Dynamics, Vol. 45, No. 11, 2022, pp. 2061–2075

    work page 2022

  10. [10]

    Chance-constrained output-feedback control without history feedback: Application to NRHO stationkeeping,

    Aleti, D., Oguri, K., and Kumagai, N., “Chance-constrained output-feedback control without history feedback: Application to NRHO stationkeeping,”AIAA/AAS Astrodynamics Specialists Conference, 2023, pp. 1–20

    work page 2023

  11. [11]

    Chance-Constrained Control for Safe Spacecraft Autonomy: Convex Programming Approach,

    Oguri, K., “Chance-Constrained Control for Safe Spacecraft Autonomy: Convex Programming Approach,”IEEE American Control Conference, Vol. 132, No. 28, 2024

    work page 2024

  12. [12]

    Linear Covariance Techniques for Orbital Rendezvous Analysis and Autonomous Onboard Mission Planning,

    Geller, D. K., “Linear Covariance Techniques for Orbital Rendezvous Analysis and Autonomous Onboard Mission Planning,” Journal of Guidance, Control, and Dynamics, Vol. 29, No. 6, 2006, pp. 1404–1414

    work page 2006

  13. [13]

    Chance-constrainedSensing-optimalPathPlanningforSafeAngles-onlyAutonomousNavigation,

    Ra,M.A.P.,andOguri,K.,“Chance-constrainedSensing-optimalPathPlanningforSafeAngles-onlyAutonomousNavigation,” AAS/AIAA Astrodynamics Specialist Conference, 2024, pp. 1–20

    work page 2024

  14. [14]

    Investigation on Moon-based Sensor Placement for Cislunar Orbit Determination with Exclusion Zones,

    Jarrett-Izzi, E. M., Oguri, K., Carpenter, M., and Danis, J., “Investigation on Moon-based Sensor Placement for Cislunar Orbit Determination with Exclusion Zones,”AAS Guidance, Navigation, and Control Conference, 2024, pp. 1–20

    work page 2024

  15. [15]

    Adaptive Gaussian Mixture Filtering for Multi-sensor Maneuvering Cislunar Space Object Tracking,

    Iannamorelli, J. L., and LeGrand, K. A., “Adaptive Gaussian Mixture Filtering for Multi-sensor Maneuvering Cislunar Space Object Tracking,”The Journal of the Astronautical Sciences, Vol. 72, No. 2, 2025, pp. 1–35

    work page 2025

  16. [16]

    Accurate Uncertainty Characterization of Impulsive Thrust Maneuvers in the Restricted Three Body Problem,

    Sharan, S., Eapen, R., Singla, P., and Melton, R., “Accurate Uncertainty Characterization of Impulsive Thrust Maneuvers in the Restricted Three Body Problem,”The Journal of the Astronautical Sciences, Vol. 70, No. 35, 2023, pp. 1–35. 20

    work page 2023

  17. [17]

    Confidence Regions in Non-Linear Estimation,

    Beale, E. M. L., “Confidence Regions in Non-Linear Estimation,”Journal of the Royal Statistical Society, Vol. 22, No. 1, 1960, pp. 41–76

    work page 1960

  18. [18]

    How Nonlinear Is It? A Tutorial on Nonlinearity of Orbit and Attitude Dynamics,

    Junkins, J. L., and Singla, P., “How Nonlinear Is It? A Tutorial on Nonlinearity of Orbit and Attitude Dynamics,”The Journal of the Astronautical Sciences, Vol. 52, No. 1, 2004, pp. 7–60

    work page 2004

  19. [19]

    SemianalyticalMeasuresofNonlinearityBasedonTensorEigenpairs,

    Jenson, E.L., andScheeres, D.J., “SemianalyticalMeasuresofNonlinearityBasedonTensorEigenpairs,”JournalofGuidance, Control, and Dynamics, Vol. 46, No. 4, 2023, pp. 638–653

    work page 2023

  20. [20]

    MATMPC - A MATLAB Based Toolbox for Real-time Nonlinear Model Predictive Control,

    Chen, Y., Bruschetta, M., Picotti, E., and Beghi, A., “MATMPC - A MATLAB Based Toolbox for Real-time Nonlinear Model Predictive Control,”18th European Control Conference, 2019, pp. 3365–3370

    work page 2019

  21. [21]

    High-Order State and Parameter Transition Tensor Calculations,

    Turner, J. D., Majji, M., and Junkins, J. L., “High-Order State and Parameter Transition Tensor Calculations,”AIAA/AAS Astrodynamics Specialist Conference, 2008, pp. 1–24

    work page 2008

  22. [22]

    OntheComputationandAccuracyofTrajectoryStateTransitionMatrices,

    Pellegrini,E.,andRussell,R.,“OntheComputationandAccuracyofTrajectoryStateTransitionMatrices,”JournalofGuidance, Control, and Dynamics, Vol. 39, No. 11, 2016, pp. 2485–2499

    work page 2016

  23. [23]

    Applications of Induced Tensor Norms to Guidance Navigation and Control,

    Kulik, J., Ruth, M., Orton-Urbina, C., and Savransky, D., “Applications of Induced Tensor Norms to Guidance Navigation and Control,”Journal of Guidance, Control, and Dynamics, Vol. 48, No. 10, 2025, pp. 1–19

    work page 2025

  24. [24]

    An Adaptive Shifted Power Method for Computing Generalized Tensor Eigenpairs,

    Kolda, T. G., and Mayo, J. R., “An Adaptive Shifted Power Method for Computing Generalized Tensor Eigenpairs,”Journal on Matrix Analysis and Applications, Vol. 35, No. 4, 2014, pp. 1–20

    work page 2014

  25. [25]

    Boyd, S., and Vandenberghe, L.,Convex Optimization, 7th ed., Cambridge University Press, 2009

    work page 2009

  26. [26]

    Dynamics Leveraged in Long-term Stationkeeping Strategies For Multi-Body Orbits,

    Williams, D. A. P., Howell, K. C., and Davis, D. C., “Dynamics Leveraged in Long-term Stationkeeping Strategies For Multi-Body Orbits,”AAS/AIAA Astrodynamics Specialist Conference, 2024, pp. 1–20

    work page 2024

  27. [27]

    Near Rectilinear Halo Orbits and Nearby Higher-Period Dynamical Structures: Orbital Stability and Resonance Properties,

    Zimovan-Spreen, E. M., Howell, K. C., and Davis, D. C., “Near Rectilinear Halo Orbits and Nearby Higher-Period Dynamical Structures: Orbital Stability and Resonance Properties,”Celestial Mechanics and Dynamical Astronomy, Vol. 132, No. 28, 2020, pp. 1–25

    work page 2020

  28. [28]

    Multi-Objective Optimization of Covariance and Energy for Asteroid Transfers,

    Jenson, E. L., and Scheeres, D. J., “Multi-Objective Optimization of Covariance and Energy for Asteroid Transfers,”Journal of Guidance, Control, and Dynamics, Vol. 44, No. 7, 2021, pp. 1253–1265

    work page 2021

  29. [29]

    Seber, G. A. F.,A Matrix Handbook for Statisticians, John Wiley & Sons, 2008

    work page 2008

  30. [30]

    Optimal Statistical Moment Steering for Controlling Non-Gaussian Distributions,

    Qi, D. C., Oguri, K., Singla, P., and Akella, M. R., “Optimal Statistical Moment Steering for Controlling Non-Gaussian Distributions,”AAS/AIAA Astrodynamics Specialist Conference, 2025, pp. 1–20. 21

    work page 2025