arxiv: 2604.24451 · v1 · submitted 2026-04-27 · 🌌 astro-ph.IM

Recognition: unknown

Robust Angles-Only Initial Relative Orbit Determination Using Polynomial Optimization

Dong Qiao, Malcolm Macdonald, Roberto Armellin, Xiangyu Li, Xingyu Zhou

Authors on Pith no claims yet

Pith reviewed 2026-05-07 17:28 UTC · model grok-4.3

classification 🌌 astro-ph.IM

keywords angles-only IRODpolynomial optimizationdifferential algebrarelative orbit determinationspacecraft navigationrobust estimationnonlinear dynamicsinitial orbit determination

0 comments

The pith

Polynomial optimization with reduced-order weighting solves robust angles-only initial relative orbit determination for nonlinear dynamics.

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

The paper establishes a method for determining initial relative orbits of spacecraft using only angle measurements under arbitrary nonlinear dynamics. It approximates the relative motion with high-order Taylor polynomials inside a differential algebra framework and minimizes the cross-product residual through a recursive polynomial optimization procedure. A reduced-order weighting strategy projects the residual onto the two-dimensional tangent subspace of the line of sight to remove the intrinsic singularity of conventional three-dimensional weighting. Robustness against poor initialization, strong noise, and bad geometries is added via a zero-solution-avoidance constraint and an adaptive threshold mechanism. Numerical simulations indicate the approach yields roughly three orders of magnitude better accuracy than baselines while lowering the load on later orbit refinement.

Core claim

Approximating relative motion by high-order Taylor polynomials within the differential algebra framework allows the cross-product-residual minimization problem to be solved through recursive polynomial optimization; the reduced-order weighting obtained by projecting the residual onto the line-of-sight tangent subspace, together with zero-solution avoidance and adaptive thresholds, yields a robust estimator that improves IROD accuracy by about three orders of magnitude over baseline methods in numerical tests.

What carries the argument

Recursive polynomial optimization applied to high-order Taylor polynomial approximations of relative motion, combined with reduced-order weighting that projects the residual onto the two-dimensional tangent subspace of the line of sight.

If this is right

IROD accuracy improves by about three orders of magnitude relative to baseline methods.
The reduced-order weighting improves accuracy by about 43 percent in the nominal case and outperforms conventional three-dimensional weighting by about 81 percent under large noise.
The method remains stable under strong measurement noise and unfavorable observation geometries.
Downstream orbit-refinement computational burden is reduced.

Where Pith is reading between the lines

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

The polynomial framework may support efficient analytic uncertainty propagation for mission design trade studies.
Similar reduced-order weighting could be tested on angles-only problems in other tracking domains such as multi-target surveillance.
Integration with onboard processors could be examined for real-time autonomous rendezvous applications.

Load-bearing premise

The high-order Taylor polynomial approximations remain sufficiently accurate for the nonlinear dynamics and observation geometries that arise in realistic space missions.

What would settle it

A simulation or flight scenario in which the truncation error of the chosen Taylor order exceeds the measurement noise level for the encountered dynamics or geometry, resulting in orbit estimates that deviate from truth by more than the claimed accuracy gain, would falsify the central claim.

read the original abstract

This paper develops a robust angles-only IROD method based on polynomial optimization for arbitrary nonlinear dynamics. First, the relative motion is approximated by high-order Taylor polynomials within the differential algebra framework, and the resulting cross-product-residual minimization problem is solved through a recursive polynomial optimization procedure. Second, a reduced-order weighting strategy is introduced by projecting the residual onto the two-dimensional tangent subspace of the line of sight, thereby structurally removing the intrinsic singularity of conventional three-dimensional weighting. Third, a zero-solution-avoidance constraint together with an adaptive threshold-selection mechanism is developed to improve robustness against poor initialization, strong measurement noise, and unfavorable observation geometries. Numerical simulations show that the proposed method improves IROD accuracy by about three orders of magnitude relative to the baseline methods, while also reducing the downstream orbit-refinement burden. The reduced-order weighting strategy further improves accuracy by about 43% in the nominal case and remains stable under large-noise conditions, outperforming the conventional three-dimensional weighting by about 81%.

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

The paper gives a workable polynomial-based angles-only IROD procedure with a clean fix for weighting singularities, but the simulation gains rest on unexamined truncation accuracy in the differential algebra approximations.

read the letter

The paper combines high-order Taylor polynomial approximations from differential algebra with recursive optimization to solve the angles-only initial relative orbit determination problem for nonlinear dynamics. It also introduces a reduced-order weighting by projecting residuals onto the tangent subspace of the line of sight and adds a zero-solution-avoidance constraint with adaptive thresholding. This setup does a couple of things right. The reduced-order weighting structurally avoids the singularity that comes with three-dimensional weighting on line-of-sight data, which is a practical improvement. The adaptive mechanism for avoiding trivial solutions helps when initialization is poor or noise is high, addressing real operational issues in formation flying and rendezvous. The main weakness is the lack of direct checks on the polynomial approximation quality. The simulations claim roughly three orders of magnitude better accuracy and additional gains from the weighting strategy, but without reported truncation error bounds, order-convergence tests, or side-by-side comparisons of polynomial trajectories versus numerical integration, it's difficult to confirm that the gains are truly from the optimization rather than the approximation holding in the chosen test cases. The abstract gives the headline numbers, but fuller details on Monte Carlo setup and baseline code would strengthen the case. This is for researchers in astrodynamics focused on angles-only methods for relative navigation. Someone in that niche would find the specific techniques and robustness features worth looking at. The paper shows clear thinking on the problem structure and engages with prior work on differential algebra, so it merits peer review even if the simulation validation needs tightening. I would recommend sending it out for review, asking reviewers to focus on the approximation validation and reproducibility of the results.

Referee Report

3 major / 2 minor

Summary. The paper presents a robust angles-only initial relative orbit determination (IROD) technique for arbitrary nonlinear dynamics. Relative motion is replaced by high-order Taylor polynomials via differential algebra, the cross-product residual minimization is solved recursively, a reduced-order weighting projects residuals onto the line-of-sight tangent plane to remove the 3D singularity, and a zero-solution-avoidance constraint with adaptive thresholding improves robustness. Numerical simulations are reported to yield roughly three orders of magnitude better IROD accuracy than baselines, plus 43% further gain from reduced-order weighting (81% under large noise) and reduced downstream refinement burden.

Significance. If the polynomial approximations remain accurate and the simulation campaign is representative, the approach would constitute a meaningful advance in angles-only IROD by converting a non-convex problem into a polynomial optimization that is both more accurate and more robust to initialization and noise. The structural removal of the weighting singularity and the adaptive threshold are particularly attractive for operational use. The work also supplies a concrete, reproducible polynomial-optimization pipeline that could be adopted by other relative-orbit algorithms.

major comments (3)

[Abstract / Numerical Simulations] Abstract and Numerical Simulations section: the central claim of ~3-order-of-magnitude accuracy improvement (and the 43%/81% gains from reduced-order weighting) is not accompanied by any reported truncation-error bounds, order-convergence study, or direct comparison between the high-order Taylor polynomials and full nonlinear numerical integration of the relative dynamics. Without these checks it is impossible to attribute the reported residuals to the optimization procedure rather than to approximation error.
[Numerical Simulations] Numerical Simulations section: the manuscript provides no information on the number of Monte Carlo trials, the precise dynamics models (including whether J2, drag, or third-body perturbations were included), the specific error metrics (e.g., position/velocity RMSE, angular residuals), the exact baseline implementations, or how post-hoc parameter choices were avoided. These omissions make the quantitative claims difficult to reproduce or generalize.
[Method] Method description (polynomial optimization procedure): the recursive polynomial optimization is presented as exact once the Taylor map is obtained, yet no analysis is given of how the truncation order interacts with the adaptive threshold or the zero-solution-avoidance constraint under the large-noise and poor-geometry cases that are claimed to be handled robustly.

minor comments (2)

[Method] Notation for the reduced-order weighting matrix and the tangent-plane projection should be introduced with an explicit equation number and a short geometric diagram.
[Abstract] The abstract states improvements “by about three orders of magnitude” and “about 43%”; these should be replaced by precise median or mean values with confidence intervals once the simulation statistics are reported.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed review. We have addressed each major comment with clarifications and will revise the manuscript to incorporate additional validation, details, and analysis as outlined below.

read point-by-point responses

Referee: [Abstract / Numerical Simulations] Abstract and Numerical Simulations section: the central claim of ~3-order-of-magnitude accuracy improvement (and the 43%/81% gains from reduced-order weighting) is not accompanied by any reported truncation-error bounds, order-convergence study, or direct comparison between the high-order Taylor polynomials and full nonlinear numerical integration of the relative dynamics. Without these checks it is impossible to attribute the reported residuals to the optimization procedure rather than to approximation error.

Authors: We agree that explicit checks are needed to separate optimization gains from approximation error. Although the differential algebra framework derives the Taylor map directly from the nonlinear dynamics, the manuscript does not report truncation-error bounds or a convergence study. In the revised version we will add a dedicated subsection presenting truncation-error bounds for the selected order, an order-convergence study, and side-by-side comparisons of polynomial-evaluated states versus full nonlinear numerical integration on representative trajectories. These additions will confirm that approximation error lies well below the reported IROD accuracy. revision: yes
Referee: [Numerical Simulations] Numerical Simulations section: the manuscript provides no information on the number of Monte Carlo trials, the precise dynamics models (including whether J2, drag, or third-body perturbations were included), the specific error metrics (e.g., position/velocity RMSE, angular residuals), the exact baseline implementations, or how post-hoc parameter choices were avoided. These omissions make the quantitative claims difficult to reproduce or generalize.

Authors: We acknowledge that the current manuscript omits these implementation details, limiting reproducibility. We will expand the Numerical Simulations section to report the exact number of Monte Carlo trials, the full dynamics model (explicitly stating inclusion or exclusion of J2, drag, and third-body terms), the precise error metrics (position/velocity RMSE and angular residuals), descriptions of the baseline implementations, and confirmation that all parameters were fixed a priori on physical grounds without post-hoc tuning. revision: yes
Referee: [Method] Method description (polynomial optimization procedure): the recursive polynomial optimization is presented as exact once the Taylor map is obtained, yet no analysis is given of how the truncation order interacts with the adaptive threshold or the zero-solution-avoidance constraint under the large-noise and poor-geometry cases that are claimed to be handled robustly.

Authors: We recognize that the manuscript does not examine the interaction of truncation order with the adaptive threshold and zero-solution-avoidance constraint in the challenging regimes highlighted. In the revised manuscript we will add a sensitivity analysis (in the Method section or a new subsection) that quantifies performance across truncation orders for the adaptive threshold and zero-solution-avoidance mechanism, using the large-noise and poor-geometry cases already considered. This will demonstrate that the chosen order preserves robustness and that truncation effects do not undermine the reported gains. revision: yes

Circularity Check

0 steps flagged

No circularity in the claimed derivation chain

full rationale

The paper presents a forward algorithmic procedure: relative motion is approximated via high-order Taylor polynomials in the differential algebra framework, the cross-product residual minimization is solved by recursive polynomial optimization, a reduced-order weighting projects residuals onto the line-of-sight tangent subspace, and zero-solution-avoidance plus adaptive thresholding are added for robustness. All performance claims (three-order accuracy gain, 43%/81% weighting improvements) are obtained from numerical simulations rather than from any analytical derivation that reduces outputs to inputs by construction. No self-definitional equations, fitted parameters renamed as predictions, or load-bearing self-citations appear in the abstract or method outline; the chain remains independent of its own fitted values or prior author results.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The central claim rests on the validity of high-order Taylor approximations for relative orbital dynamics and on the numerical behavior of the optimization procedure; these are standard in the field rather than newly postulated.

free parameters (2)

polynomial truncation order
The degree of the Taylor polynomials is a design choice that must be selected for each application to balance accuracy and computation.
adaptive threshold value
The threshold used in the zero-solution-avoidance mechanism is described as adaptive but its exact selection rule is not specified in the abstract.

axioms (1)

domain assumption High-order Taylor polynomials within the differential algebra framework provide a sufficiently accurate local approximation to the true nonlinear relative dynamics over the observation interval.
Invoked when the relative motion is replaced by its polynomial representation before optimization.

pith-pipeline@v0.9.0 · 5475 in / 1452 out tokens · 149448 ms · 2026-05-07T17:28:35.072553+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages

[1]

Deep learning-based spacecraft relative navigation methods: A survey,

Song, J., Rondao, D., and Aouf, N., “Deep learning-based spacecraft relative navigation methods: A survey,”Acta Astronautica, Vol. 191, 2022, pp. 22–40. https://doi.org/10.1016/j.actaastro.2021.10.025

work page doi:10.1016/j.actaastro.2021.10.025 2022
[3]

Bearings-Only Initial Relative Orbit Determination,

LeGrand, K. A., DeMars, K. J., and Pernicka, H. J., “Bearings-Only Initial Relative Orbit Determination,”Journal of Guidance, Control, and Dynamics, Vol. 38, No. 9, 2015, pp. 1699–1713. https://doi.org/10.2514/1.G001003

work page doi:10.2514/1.g001003 2015
[4]

Efficient Orbit Determination Using Measurement-Directional State 20 Transition Tensor,

Zhou, X., Qiao, D., Malcolm, M., and Li, X., “Efficient Orbit Determination Using Measurement-Directional State 20 Transition Tensor,”IEEE Transactions on Aerospace and Electronic Systems, Vol. 61, No. 3, 2025, pp. 8014–8027. https://doi.org/10.1109/TAES.2025.3527410

work page doi:10.1109/taes.2025.3527410 2025
[5]

Initial Orbit Determination of Periodic Orbits in the Earth-Moon System with Ground-based Optical Observations,

Liu, X., Armellin, R., Pirovano, L., and Hou, X., “Initial Orbit Determination of Periodic Orbits in the Earth-Moon System with Ground-based Optical Observations,”IEEE Transactions on Aerospace and Electronic Systems, Vol. 1, 2024, pp. 1–15. https://doi.org/10.1109/TAES.2024.3435634

work page doi:10.1109/taes.2024.3435634 2024
[6]

Observability analysis of cooperative orbit determination using inertial inter-spacecraft angle measurements,

Zhou, X., Qin, T., Macdonald, M., and Qiao, D., “Observability analysis of cooperative orbit determination using inertial inter-spacecraft angle measurements,”Acta Astronautica, Vol. 210, 2023, pp. 289–302. https://doi.org/10.1016/j.actaastro.2023. 05.019

work page doi:10.1016/j.actaastro.2023 2023
[7]

Angles-Only Navigation State Observability During Orbital Proximity Operations,

Geller, D. K., and Klein, I., “Angles-Only Navigation State Observability During Orbital Proximity Operations,”Journal of Guidance, Control, and Dynamics, Vol. 37, No. 6, 2014, pp. 1976–1983. https://doi.org/10.2514/1.G000133

work page doi:10.2514/1.g000133 2014
[8]

A numerical approach to the problem of angles-only initial relative orbit determination in low earth orbit,

Ardaens, J.-S., and Gaias, G., “A numerical approach to the problem of angles-only initial relative orbit determination in low earth orbit,”Advances in Space Research, Vol. 63, No. 12, 2019, pp. 3884–3899. https://doi.org/10.1016/j.asr.2019.03.001

work page doi:10.1016/j.asr.2019.03.001 2019
[9]

Robust trajectory design and guidance for far-range rendezvous using reinforcement learning with safety and observability considerations,

Wijayatunga, M. C., Armellin, R., and Holt, H., “Robust trajectory design and guidance for far-range rendezvous using reinforcement learning with safety and observability considerations,”Aerospace Science and Technology, Vol. 159, 2025, p. 109996. https://doi.org/10.1016/j.ast.2025.109996

work page doi:10.1016/j.ast.2025.109996 2025
[10]

Autonomous Exploration of a Small Near-Earth Asteroid,

Takahashi, S., and Scheeres, D. J., “Autonomous Exploration of a Small Near-Earth Asteroid,”Journal of Guidance, Control, and Dynamics, Vol. 44, No. 4, 2021, pp. 701–718. https://doi.org/10.2514/1.G005733

work page doi:10.2514/1.g005733 2021
[12]

Fast angles-only relative navigation using polynomial dynamics,

Willis, M., and D’Amico, S., “Fast angles-only relative navigation using polynomial dynamics,”Advances in Space Research, Vol. 73, No. 11, 2024, pp. 5484–5500. https://doi.org/10.1016/j.asr.2023.07.043

work page doi:10.1016/j.asr.2023.07.043 2024
[13]

Angles-Only Initial Relative Orbit Determination Performance Analysis using Cylindrical Coordinates,

Geller, D. K., and Lovell, T. A., “Angles-Only Initial Relative Orbit Determination Performance Analysis using Cylindrical Coordinates,”The Journal of the Astronautical Sciences, Vol. 64, No. 1, 2017, pp. 72–96. https://doi.org/10.1007/s40295-016- 0095-z

work page doi:10.1007/s40295-016- 2017
[14]

Angles-Only Navigation to a Noncooperative Satellite Using Relative Orbital Elements,

Gaias, G., D’Amico, S., and Ardaens, J.-S., “Angles-Only Navigation to a Noncooperative Satellite Using Relative Orbital Elements,”Journal of Guidance, Control, and Dynamics, Vol. 37, No. 2, 2014, pp. 439–451. https://doi.org/10.2514/1.61494

work page doi:10.2514/1.61494 2014
[15]

Generalized Angles-Only Navigation Architecture for Autonomous Distributed Space Systems,

Sullivan, J., Koenig, A. W., Kruger, J., and D’Amico, S., “Generalized Angles-Only Navigation Architecture for Autonomous Distributed Space Systems,”Journal of Guidance, Control, and Dynamics, Vol. 44, No. 6, 2021, pp. 1087–1105. https://doi.org/10.2514/1.G005439. 21

work page doi:10.2514/1.g005439 2021
[16]

Deep-neural-network-based angles-only relative orbit determination for space non-cooperative target,

Gong, B., Ma, Y., Zhang, W., Li, S., and Li, X., “Deep-neural-network-based angles-only relative orbit determination for space non-cooperative target,”Acta Astronautica, Vol. 204, 2023, pp. 552–567. https://doi.org/10.1016/j.actaastro.2022.09.024

work page doi:10.1016/j.actaastro.2022.09.024 2023
[17]

Saurabh, D

Kulik, J., Oertell, O., and Savransky, D., “Overdetermined Eigenvector Approach to Passive Angles-Only Relative Orbit Determination,”Journal of Guidance, Control, and Dynamics, Vol. 47, No. 5, 2024, pp. 986–994. https://doi.org/10.2514/1. G007744

work page doi:10.2514/1 2024
[18]

Optimal Linear Orbit Determination,

Sinclair, A. J., and Alan Lovell, T., “Optimal Linear Orbit Determination,”Journal of Guidance, Control, and Dynamics, Vol. 43, No. 3, 2020, pp. 628–632. https://doi.org/10.2514/1.G004182

work page doi:10.2514/1.g004182 2020
[19]

Three-axis attitude determination from vector observations,

Shuster, M. D., and Oh, S. D., “Three-axis attitude determination from vector observations,”Journal of Guidance and Control, Vol. 4, No. 1, 1981, pp. 70–77. https://doi.org/10.2514/3.19717

work page doi:10.2514/3.19717 1981
[20]

Nonlinear mapping of Gaussian statistics: Theory and applications to spacecraft trajectory design,

Park, R. S., and Scheeres, D. J., “Nonlinear mapping of Gaussian statistics: Theory and applications to spacecraft trajectory design,”Journal of Guidance, Control, and Dynamics, Vol. 29, No. 6, 2006, pp. 1367–1375. https://doi.org/10.2514/1.20177

work page doi:10.2514/1.20177 2006
[21]

Time-VaryingDirectionalStateTransitionTensorforOrbitUncertaintyPropagation,

Zhou,X.,Armellin,R.,Qiao,D.,andLi,X.,“Time-VaryingDirectionalStateTransitionTensorforOrbitUncertaintyPropagation,” Journal of Guidance, Control, and Dynamics, Vol. 49, No. 3, 2026, pp. 656–672. https://doi.org/10.2514/1.G009000

work page doi:10.2514/1.g009000 2026
[22]

Gravity assist space pruning based on differentialalgebra,

Armellin, R., Di Lizia, P., Topputo, F., Lavagna, M., Bernelli-Zazzera, F., and Berz, M., “Gravity assist space pruning based on differentialalgebra,”Celestial Mechanics and Dynamical Astronomy,Vol.106,No.1,2009,p.1. https://doi.org/10.1007/s10569- 009-9235-0

work page doi:10.1007/s10569- 2009
[23]

ToolstodetectstructuresindynamicalsystemsusingJetTransport,

Pérez-Palau,D.,Masdemont,J.J.,andGómez,G.,“ToolstodetectstructuresindynamicalsystemsusingJetTransport,”Celestial Mechanics and Dynamical Astronomy, Vol. 123, No. 3, 2015, pp. 239–262. https://doi.org/10.1007/s10569-015-9634-3

work page doi:10.1007/s10569-015-9634-3 2015
[24]

Least Squares Approximation,

Crassidis, J., and Junkins, J., “Least Squares Approximation,”Optimal Estimation of Dynamic Systems (2nd ed.), Chapman and Hall/CRC, 2011, Chap. 1, 2nd ed., pp. 1–62. https://doi.org/10.1201/b11154

work page doi:10.1201/b11154 2011
[25]

Recursive Polynomial Method for Fast Collision Avoidance Maneuver Design,

Pavanello, Z., Pirovano, L., and Armellin, R., “Recursive Polynomial Method for Fast Collision Avoidance Maneuver Design,” IEEE Transactions on Aerospace and Electronic Systems, Vol. 61, No. 2, 2024, pp. 2914–2925. https://doi.org/10.1109/TAES. 2024.3480041

work page doi:10.1109/taes 2024
[26]

Maneuver Detection via a Confidence Dominance Maneuver Indicator,

Zhou, X., Armellin, R., Pirovano, L., Qiao, D., and Li, X., “Maneuver Detection via a Confidence Dominance Maneuver Indicator,” , 2025. URL https://arxiv.org/abs/2507.08234. 22

work page arXiv 2025