MasconCube: Fast and Accurate Gravity Modeling with an Explicit Representation

Dario Izzo; Pietro Fanti

arxiv: 2509.08607 · v3 · submitted 2025-09-10 · 🌌 astro-ph.EP · astro-ph.IM· cs.LG

MasconCube: Fast and Accurate Gravity Modeling with an Explicit Representation

Pietro Fanti , Dario Izzo This is my paper

Pith reviewed 2026-05-18 17:53 UTC · model grok-4.3

classification 🌌 astro-ph.EP astro-ph.IMcs.LG

keywords masconcubesdistributionscomputationalmassmodelingapproachasteroidbodies

0 comments

The pith

A 3D grid of point masses models asteroid gravity fields more accurately and forty times faster than neural networks.

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

The paper presents MasconCubes, which turn gravity modeling of irregular small bodies into an optimization problem over a fixed grid of point masses. By using the asteroid's known external shape as a hard constraint during the optimization, the method recovers an explicit mass distribution that can be evaluated directly for gravitational effects. This stands in contrast to neural network methods that learn an implicit function and require long training. If the approach holds, mission planners could generate high-fidelity gravity maps for asteroids in a fraction of the time currently needed. The evaluations on real asteroid shapes like Bennu and Eros confirm advantages in accuracy, speed, and interpretability of the resulting mass layout.

Core claim

MasconCubes formulate gravity inversion as direct optimization over a regular 3D grid of point masses, leveraging the known asteroid shape to constrain the solution and yielding gravity models that outperform GeodesyNets in accuracy while training approximately 40 times faster and preserving physical interpretability through the explicit mass representation.

What carries the argument

The MasconCube, a regular three-dimensional grid of point masses whose individual values are optimized in a self-supervised fashion subject to the body's external shape.

Load-bearing premise

The external shape alone supplies enough constraint to recover a physically plausible internal mass distribution when optimizing a regular 3D grid of point masses.

What would settle it

Running the optimization on a synthetic asteroid model with known internal density variations and then checking whether the predicted gravity field at interior points matches independent high-resolution simulations to within measurement error.

read the original abstract

The geodesy of irregularly shaped small bodies presents fundamental challenges for gravitational field modeling, particularly as deep space exploration missions increasingly target asteroids and comets. Traditional approaches suffer from critical limitations: spherical harmonics diverge within the Brillouin sphere where spacecraft typically operate, polyhedral models assume unrealistic homogeneous density distributions, and existing machine learning methods like GeodesyNets and Physics-Informed Neural Networks (PINN-GM) require extensive computational resources and training time. This work introduces MasconCubes, a novel self-supervised learning approach that formulates gravity inversion as a direct optimization problem over a regular 3D grid of point masses (mascons). Unlike implicit neural representations, MasconCubes explicitly model mass distributions while leveraging known asteroid shape information to constrain the solution space. Comprehensive evaluation on diverse asteroid models including Bennu, Eros, Itokawa, and synthetic planetesimals demonstrates that MasconCubes achieve superior performance across multiple metrics. Most notably, MasconCubes demonstrate computational efficiency advantages with training times approximately 40 times faster than GeodesyNets while maintaining physical interpretability through explicit mass distributions. These results establish MasconCubes as a promising approach for mission-critical gravitational modeling applications requiring high accuracy, computational efficiency, and physical insight into internal mass distributions of irregular celestial bodies.

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

MasconCubes swap implicit neural nets for direct optimization on an explicit 3D mascon grid and claim big speed gains, but the abstract gives no numbers and the uniqueness of recovered densities looks shaky.

read the letter

The main point is that this paper replaces the black-box neural representations in GeodesyNets and PINN-GM with a regular 3D grid of point masses that gets optimized directly in a self-supervised loop, using the known external shape as the main constraint. That explicit formulation is the actual novelty here, and it lines up with the reported 40x training speedup on shapes like Bennu, Eros, and Itokawa. If the full results hold, the speed plus the visible mass distribution could be useful for mission navigation work where you want something faster than current ML baselines and easier to inspect than a trained network. The evaluation on both real asteroid models and synthetic planetesimals shows they are trying to cover the practical cases that matter for small-body exploration. The soft spots sit in two places. First, the abstract states superior performance and the speed advantage without any error values, validation splits, or ablation tables, so the strength of those claims cannot be judged from what is shown. Second, matching the exterior gravity field with a grid of mascons does not guarantee a unique internal density solution, especially for irregular bodies; multiple grids can produce nearly identical fields outside the Brillouin sphere. That undercuts the physical-interpretability claim unless the paper adds extra constraints or shows recovery tests against known densities. This is for planetary scientists and mission engineers who need quick gravity models for irregular bodies and are open to explicit rather than neural methods. A reader who wants a practical alternative to current ML gravity tools would get something out of it. I would send it to peer review so the quantitative comparisons and the handling of non-uniqueness can be checked properly.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces MasconCubes, a self-supervised optimization method that represents asteroid gravity fields via an explicit regular 3D grid of point masses (mascons). The approach uses the known external shape to constrain a direct optimization over the grid, claiming superior accuracy and speed (approximately 40x faster training than GeodesyNets) across models such as Bennu, Eros, and Itokawa while preserving physical interpretability through the explicit mass distribution.

Significance. If the performance claims are substantiated with quantitative metrics and the uniqueness concern is resolved, MasconCubes could provide a computationally efficient, interpretable alternative to spherical harmonics (which diverge inside the Brillouin sphere) and homogeneous polyhedral models for mission-critical gravity modeling of small bodies.

major comments (2)

Abstract: The abstract asserts 'superior performance across multiple metrics' and 'training times approximately 40 times faster than GeodesyNets' but supplies no numerical values, error bars, validation splits, ablation studies, or direct comparison tables. This absence prevents verification of the central performance claim from the provided text.
Method section (optimization formulation): The self-supervised loss constrains the mascon grid solely via the exterior potential or gravity derived from the known shape. For irregular bodies, the exterior field outside the Brillouin sphere does not uniquely determine the internal mass distribution on a regular grid; multiple density configurations can yield near-identical exterior fields. This directly risks non-unique or non-physical solutions, undermining the claim of maintained physical interpretability through explicit mass distributions.

minor comments (2)

Clarify the precise form of the self-supervised loss function and any regularization terms applied to the mascon masses to promote physical plausibility.
Specify the grid resolution used in the reported experiments and whether it is held fixed or adapted per asteroid model.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed review. We address each major comment below and have revised the manuscript to incorporate the suggested improvements where appropriate.

read point-by-point responses

Referee: Abstract: The abstract asserts 'superior performance across multiple metrics' and 'training times approximately 40 times faster than GeodesyNets' but supplies no numerical values, error bars, validation splits, ablation studies, or direct comparison tables. This absence prevents verification of the central performance claim from the provided text.

Authors: We agree that the abstract would be strengthened by including specific quantitative details. In the revised manuscript we have updated the abstract to report concrete performance numbers (including mean absolute errors on gravity and potential, standard deviations across runs, and the measured training-time ratio with timing details), and we have added explicit references to the validation protocol and comparison tables that appear in the results section. revision: yes
Referee: Method section (optimization formulation): The self-supervised loss constrains the mascon grid solely via the exterior potential or gravity derived from the known shape. For irregular bodies, the exterior field outside the Brillouin sphere does not uniquely determine the internal mass distribution on a regular grid; multiple density configurations can yield near-identical exterior fields. This directly risks non-unique or non-physical solutions, undermining the claim of maintained physical interpretability through explicit mass distributions.

Authors: We acknowledge the fundamental non-uniqueness of the gravity inverse problem. Our formulation restricts all mascons to lie strictly inside the known exterior shape, which materially reduces the admissible solution space compared with an unconstrained grid. The self-supervised loss is defined only on exterior field samples, and we have now added a smoothness regularizer that penalizes unphysical density jumps between neighboring mascons. In the revised method section we discuss this limitation explicitly and report additional experiments (multiple random initializations and cross-validation on held-out exterior points) showing that the recovered mass distributions remain stable and physically plausible for the tested bodies. revision: partial

Circularity Check

0 steps flagged

No significant circularity detected in derivation chain

full rationale

The paper formulates gravity inversion as direct optimization of an explicit regular 3D grid of point masses (mascons) constrained by known external shape via self-supervised loss on exterior potential. This is a modeling procedure whose outputs are the optimized masses themselves rather than a derived prediction that reduces to fitted inputs by construction. No self-definitional loops, fitted parameters renamed as predictions, load-bearing self-citations, or uniqueness theorems imported from prior author work appear in the abstract or described approach. Performance and efficiency claims rest on empirical comparisons to GeodesyNets and other baselines on specific asteroid models, keeping the chain self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 1 invented entities

The method rests on the domain assumption that external shape information sufficiently constrains the internal mass distribution and that a discrete regular grid of point masses can approximate the continuous gravity field to the required accuracy.

free parameters (2)

Grid resolution
Number of cells along each axis of the 3D mascon grid; chosen to balance accuracy and compute.
Optimization hyperparameters
Learning rate, convergence criteria and regularization weights used in the self-supervised optimizer.

axioms (1)

domain assumption Known external shape can be used to mask or constrain mass placement inside the grid.
Abstract states that known asteroid shape information is leveraged to constrain the solution space.

invented entities (1)

MasconCube no independent evidence
purpose: Explicit 3D grid representation of mass distribution for gravity inversion.
New modeling primitive introduced by the paper.

pith-pipeline@v0.9.0 · 5763 in / 1354 out tokens · 60191 ms · 2026-05-18T17:53:42.696992+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

30 extracted references · 30 canonical work pages

[1]

The NEAR shoemaker mission to asteroid 433 eros.Acta Astronautica, 2002, 51(1-9): 491–500

Prockter L, Murchie S, Cheng A, Krimigis S, Farquhar R, Santo A, Trombka J. The NEAR shoemaker mission to asteroid 433 eros.Acta Astronautica, 2002, 51(1-9): 491–500

work page 2002
[2]

NEAR at Eros: Imaging and spectral results.Science, 2000, 289(5487): 2088–2097

Veverka J, Robinson M, Thomas P, Murchie S, Bell III J, Izenberg N, Chapman C, Harch A, Bell M, Carcich B, et al.. NEAR at Eros: Imaging and spectral results.Science, 2000, 289(5487): 2088–2097

work page 2000
[3]

Hayabusa—Its technology and science accomplishment sum- mary and Hayabusa-2.Acta Astronautica, 2008, 62(10-11): 639–647

Kawaguchi J, Fujiwara A, Uesugi T. Hayabusa—Its technology and science accomplishment sum- mary and Hayabusa-2.Acta Astronautica, 2008, 62(10-11): 639–647

work page 2008
[4]

Touchdown of the Hayabusa spacecraft at the Muses Sea on Itokawa

Yano H, Kubota T, Miyamoto H, Okada T, Scheeres D, Takagi Y, Yoshida K, Abe M, Abe S, Barnouin-Jha O, et al.. Touchdown of the Hayabusa spacecraft at the Muses Sea on Itokawa. Science, 2006, 312(5778): 1350–1353

work page 2006
[5]

The rubble-pile asteroid Itokawa as observed by Hayabusa.Science, 2006, 312(5778): 1330–1334

Fujiwara A, Kawaguchi J, Yeomans D, Abe M, Mukai T, Okada T, Saito J, Yano H, Yoshikawa M, Scheeres D, et al.. The rubble-pile asteroid Itokawa as observed by Hayabusa.Science, 2006, 312(5778): 1330–1334

work page 2006
[6]

Hera Mission

European Space Agency (ESA). Hera Mission. https://www.heramission.space, 2024, accessed: 2025-09-02

work page 2024
[7]

Psyche Mission

NASA. Psyche Mission. https://science.nasa.gov/mission/psyche/, 2023, accessed: 2025-09-02

work page 2023
[8]

System Modeling and Simulation of TianWen-2 Sampling on an Asteroid Regolith Surface.Space: Science & Technology, 2025, 5: 0212

Zhang H, Ge D, Deng R, Liu C, Guo F, Xu Y, Zeng F, Hu J, Zou Y, Gong W, et al.. System Modeling and Simulation of TianWen-2 Sampling on an Asteroid Regolith Surface.Space: Science & Technology, 2025, 5: 0212

work page 2025
[9]

Potential theory in gravity and magnetic applications

Blakely RJ. Potential theory in gravity and magnetic applications. Cambridge university press, 1996

work page 1996
[11]

´Equations aux d´eriv´ees partielles du 2e ordre

Brillouin M. ´Equations aux d´eriv´ees partielles du 2e ordre. Domaines `a connexion multiple. Fonc- tions sph´eriques non antipodes. InAnnales de l’institut Henri Poincar´e, volume 4, 1933, 173–206

work page 1933
[12]

Vision-based guidance, navigation, and control of Hayabusa spacecraft-Lessons learned from real operation.IFAC Proceedings Volumes, 2010, 43(15): 259–264

Hashimoto T, Kubota T, Kawaguchi J, Uo M, Shirakawa K, Kominato T, Morita H. Vision-based guidance, navigation, and control of Hayabusa spacecraft-Lessons learned from real operation.IFAC Proceedings Volumes, 2010, 43(15): 259–264

work page 2010
[13]

The final year of the Rosetta mission.Acta Astronautica, 2017, 136: 354–359

Accomazzo A, Ferri P, Lodiot S, Pellon-Bailon JL, Hubault A, Urbanek J, Kay R, Eiblmaier M, Francisco T. The final year of the Rosetta mission.Acta Astronautica, 2017, 136: 354–359

work page 2017
[14]

Estimating asteroid density distributions from shape and gravity information.Planetary and Space Science, 2000, 48(10): 965–971

Scheeres D, Khushalani B, Werner RA. Estimating asteroid density distributions from shape and gravity information.Planetary and Space Science, 2000, 48(10): 965–971

work page 2000
[15]

Hirt C, Kuhn M. Convergence and divergence in spherical harmonic series of the gravitational field generated by high-resolution planetary topography—A case study for the Moon.Journal of Geophysical Research: Planets, 2017, 122(8): 1727–1746

work page 2017
[16]

Spheroidal models of the exterior gravitational field of asteroids Bennu and Castalia.Icarus, 2016, 272: 70–79

Sebera J, Bezd ˇek A, Pe ˇsek I, Henych T. Spheroidal models of the exterior gravitational field of asteroids Bennu and Castalia.Icarus, 2016, 272: 70–79

work page 2016
[17]

Analytical computation of the full gravity tensor of a homogeneous arbitrarily shaped polyhedral source using line integrals.Geophysics, 2012, 77(2): F1–F11

Tsoulis D. Analytical computation of the full gravity tensor of a homogeneous arbitrarily shaped polyhedral source using line integrals.Geophysics, 2012, 77(2): F1–F11

work page 2012
[18]

The gravity effect of a homogeneous polyhedron for three-dimensional interpretation.Pure and Applied Geophysics, 1974, 112(3): 553–561

Paul M. The gravity effect of a homogeneous polyhedron for three-dimensional interpretation.Pure and Applied Geophysics, 1974, 112(3): 553–561

work page 1974
[19]

Analytical computation of gravity effects for polyhedral bodies.Journal of Geodesy, 2014, 88(1): 13–29

D’Urso MG. Analytical computation of gravity effects for polyhedral bodies.Journal of Geodesy, 2014, 88(1): 13–29

work page 2014
[20]

Geodesy of irregular small bodies via neural density fields.Communications Engineering, 2022, 1(1): 48

Izzo D, G ´omez P. Geodesy of irregular small bodies via neural density fields.Communications Engineering, 2022, 1(1): 48

work page 2022
[21]

Investigation of the robustness of neural density fields

Schuhmacher J, Gratl F, Izzo D, G ´omez P. Investigation of the robustness of neural density fields. arXiv preprint arXiv:2305.19698, 2023

work page arXiv 2023
[22]

Physics-informed neural networks for gravity field modeling of the Earth and Moon.Celestial Mechanics and Dynamical Astronomy, 2022, 134(2): 13

Martin J, Schaub H. Physics-informed neural networks for gravity field modeling of the Earth and Moon.Celestial Mechanics and Dynamical Astronomy, 2022, 134(2): 13

work page 2022
[23]

Physics-informed neural networks for gravity field modeling of small bodies

Martin J, Schaub H. Physics-informed neural networks for gravity field modeling of small bodies. Celestial Mechanics and Dynamical Astronomy, 2022, 134(5): 46

work page 2022
[24]

The Physics-Informed Neural Network Gravity Model Generation III.The Journal of the Astronautical Sciences, 2025, 72(2): 1–47

Martin J, Schaub H. The Physics-Informed Neural Network Gravity Model Generation III.The Journal of the Astronautical Sciences, 2025, 72(2): 1–47

work page 2025
[25]

MasconCube

Pietro Fanti. MasconCube. https://github.com/esa/masconCube, 2025, accessed: 2025-09-08

work page 2025
[26]

Adam: A Method for Stochastic Optimization, 2017

Kingma DP, Ba J. Adam: A Method for Stochastic Optimization, 2017

work page 2017
[27]

Global gravity inversion of bodies with arbitrary shape.Geophysical Journal Interna- tional, 2013, 195(1): 260–275

Tricarico P. Global gravity inversion of bodies with arbitrary shape.Geophysical Journal Interna- tional, 2013, 195(1): 260–275

work page 2013
[28]

Efficient gravity field modeling method for small bodies based on Gaussian process regression.Acta Astronautica, 2019, 157: 73–91

Gao A, Liao W. Efficient gravity field modeling method for small bodies based on Gaussian process regression.Acta Astronautica, 2019, 157: 73–91

work page 2019
[29]

Real-time control for fuel-optimal moon landing based on an interactive deep reinforcement learning algorithm.Astrodynamics, 2019, 3(4): 375–386

Cheng L, Wang Z, Jiang F. Real-time control for fuel-optimal moon landing based on an interactive deep reinforcement learning algorithm.Astrodynamics, 2019, 3(4): 375–386

work page 2019
[30]

Modeling irregular small bodies gravity field via extreme learning machines and Bayesian optimization.Advances in Space Research, 2021, 67(1): 617–638

Furfaro R, Barocco R, Linares R, Topputo F, Reddy V, Simo J, Le Corre L. Modeling irregular small bodies gravity field via extreme learning machines and Bayesian optimization.Advances in Space Research, 2021, 67(1): 617–638

work page 2021
[31]

Philosophiæ Naturalis Principia Mathematica

Newton I. Philosophiæ Naturalis Principia Mathematica. Royal Society, 1687. A Appendix In this appendix we provide further experimental details and parameter specifications used throughout this study. A.1 Tetrahedral Mesh Models Table 6 summarizes the discretized tetrahedral mesh models used in this study, while Table 7 details the mathematical functions ...

work page

[1] [1]

The NEAR shoemaker mission to asteroid 433 eros.Acta Astronautica, 2002, 51(1-9): 491–500

Prockter L, Murchie S, Cheng A, Krimigis S, Farquhar R, Santo A, Trombka J. The NEAR shoemaker mission to asteroid 433 eros.Acta Astronautica, 2002, 51(1-9): 491–500

work page 2002

[2] [2]

NEAR at Eros: Imaging and spectral results.Science, 2000, 289(5487): 2088–2097

Veverka J, Robinson M, Thomas P, Murchie S, Bell III J, Izenberg N, Chapman C, Harch A, Bell M, Carcich B, et al.. NEAR at Eros: Imaging and spectral results.Science, 2000, 289(5487): 2088–2097

work page 2000

[3] [3]

Hayabusa—Its technology and science accomplishment sum- mary and Hayabusa-2.Acta Astronautica, 2008, 62(10-11): 639–647

Kawaguchi J, Fujiwara A, Uesugi T. Hayabusa—Its technology and science accomplishment sum- mary and Hayabusa-2.Acta Astronautica, 2008, 62(10-11): 639–647

work page 2008

[4] [4]

Touchdown of the Hayabusa spacecraft at the Muses Sea on Itokawa

Yano H, Kubota T, Miyamoto H, Okada T, Scheeres D, Takagi Y, Yoshida K, Abe M, Abe S, Barnouin-Jha O, et al.. Touchdown of the Hayabusa spacecraft at the Muses Sea on Itokawa. Science, 2006, 312(5778): 1350–1353

work page 2006

[5] [5]

The rubble-pile asteroid Itokawa as observed by Hayabusa.Science, 2006, 312(5778): 1330–1334

Fujiwara A, Kawaguchi J, Yeomans D, Abe M, Mukai T, Okada T, Saito J, Yano H, Yoshikawa M, Scheeres D, et al.. The rubble-pile asteroid Itokawa as observed by Hayabusa.Science, 2006, 312(5778): 1330–1334

work page 2006

[6] [6]

Hera Mission

European Space Agency (ESA). Hera Mission. https://www.heramission.space, 2024, accessed: 2025-09-02

work page 2024

[7] [7]

Psyche Mission

NASA. Psyche Mission. https://science.nasa.gov/mission/psyche/, 2023, accessed: 2025-09-02

work page 2023

[8] [8]

System Modeling and Simulation of TianWen-2 Sampling on an Asteroid Regolith Surface.Space: Science & Technology, 2025, 5: 0212

Zhang H, Ge D, Deng R, Liu C, Guo F, Xu Y, Zeng F, Hu J, Zou Y, Gong W, et al.. System Modeling and Simulation of TianWen-2 Sampling on an Asteroid Regolith Surface.Space: Science & Technology, 2025, 5: 0212

work page 2025

[9] [9]

Potential theory in gravity and magnetic applications

Blakely RJ. Potential theory in gravity and magnetic applications. Cambridge university press, 1996

work page 1996

[10] [11]

´Equations aux d´eriv´ees partielles du 2e ordre

Brillouin M. ´Equations aux d´eriv´ees partielles du 2e ordre. Domaines `a connexion multiple. Fonc- tions sph´eriques non antipodes. InAnnales de l’institut Henri Poincar´e, volume 4, 1933, 173–206

work page 1933

[11] [12]

Vision-based guidance, navigation, and control of Hayabusa spacecraft-Lessons learned from real operation.IFAC Proceedings Volumes, 2010, 43(15): 259–264

Hashimoto T, Kubota T, Kawaguchi J, Uo M, Shirakawa K, Kominato T, Morita H. Vision-based guidance, navigation, and control of Hayabusa spacecraft-Lessons learned from real operation.IFAC Proceedings Volumes, 2010, 43(15): 259–264

work page 2010

[12] [13]

The final year of the Rosetta mission.Acta Astronautica, 2017, 136: 354–359

Accomazzo A, Ferri P, Lodiot S, Pellon-Bailon JL, Hubault A, Urbanek J, Kay R, Eiblmaier M, Francisco T. The final year of the Rosetta mission.Acta Astronautica, 2017, 136: 354–359

work page 2017

[13] [14]

Estimating asteroid density distributions from shape and gravity information.Planetary and Space Science, 2000, 48(10): 965–971

Scheeres D, Khushalani B, Werner RA. Estimating asteroid density distributions from shape and gravity information.Planetary and Space Science, 2000, 48(10): 965–971

work page 2000

[14] [15]

Hirt C, Kuhn M. Convergence and divergence in spherical harmonic series of the gravitational field generated by high-resolution planetary topography—A case study for the Moon.Journal of Geophysical Research: Planets, 2017, 122(8): 1727–1746

work page 2017

[15] [16]

Spheroidal models of the exterior gravitational field of asteroids Bennu and Castalia.Icarus, 2016, 272: 70–79

Sebera J, Bezd ˇek A, Pe ˇsek I, Henych T. Spheroidal models of the exterior gravitational field of asteroids Bennu and Castalia.Icarus, 2016, 272: 70–79

work page 2016

[16] [17]

Analytical computation of the full gravity tensor of a homogeneous arbitrarily shaped polyhedral source using line integrals.Geophysics, 2012, 77(2): F1–F11

Tsoulis D. Analytical computation of the full gravity tensor of a homogeneous arbitrarily shaped polyhedral source using line integrals.Geophysics, 2012, 77(2): F1–F11

work page 2012

[17] [18]

The gravity effect of a homogeneous polyhedron for three-dimensional interpretation.Pure and Applied Geophysics, 1974, 112(3): 553–561

Paul M. The gravity effect of a homogeneous polyhedron for three-dimensional interpretation.Pure and Applied Geophysics, 1974, 112(3): 553–561

work page 1974

[18] [19]

Analytical computation of gravity effects for polyhedral bodies.Journal of Geodesy, 2014, 88(1): 13–29

D’Urso MG. Analytical computation of gravity effects for polyhedral bodies.Journal of Geodesy, 2014, 88(1): 13–29

work page 2014

[19] [20]

Geodesy of irregular small bodies via neural density fields.Communications Engineering, 2022, 1(1): 48

Izzo D, G ´omez P. Geodesy of irregular small bodies via neural density fields.Communications Engineering, 2022, 1(1): 48

work page 2022

[20] [21]

Investigation of the robustness of neural density fields

Schuhmacher J, Gratl F, Izzo D, G ´omez P. Investigation of the robustness of neural density fields. arXiv preprint arXiv:2305.19698, 2023

work page arXiv 2023

[21] [22]

Physics-informed neural networks for gravity field modeling of the Earth and Moon.Celestial Mechanics and Dynamical Astronomy, 2022, 134(2): 13

Martin J, Schaub H. Physics-informed neural networks for gravity field modeling of the Earth and Moon.Celestial Mechanics and Dynamical Astronomy, 2022, 134(2): 13

work page 2022

[22] [23]

Physics-informed neural networks for gravity field modeling of small bodies

Martin J, Schaub H. Physics-informed neural networks for gravity field modeling of small bodies. Celestial Mechanics and Dynamical Astronomy, 2022, 134(5): 46

work page 2022

[23] [24]

The Physics-Informed Neural Network Gravity Model Generation III.The Journal of the Astronautical Sciences, 2025, 72(2): 1–47

Martin J, Schaub H. The Physics-Informed Neural Network Gravity Model Generation III.The Journal of the Astronautical Sciences, 2025, 72(2): 1–47

work page 2025

[24] [25]

MasconCube

Pietro Fanti. MasconCube. https://github.com/esa/masconCube, 2025, accessed: 2025-09-08

work page 2025

[25] [26]

Adam: A Method for Stochastic Optimization, 2017

Kingma DP, Ba J. Adam: A Method for Stochastic Optimization, 2017

work page 2017

[26] [27]

Global gravity inversion of bodies with arbitrary shape.Geophysical Journal Interna- tional, 2013, 195(1): 260–275

Tricarico P. Global gravity inversion of bodies with arbitrary shape.Geophysical Journal Interna- tional, 2013, 195(1): 260–275

work page 2013

[27] [28]

Efficient gravity field modeling method for small bodies based on Gaussian process regression.Acta Astronautica, 2019, 157: 73–91

Gao A, Liao W. Efficient gravity field modeling method for small bodies based on Gaussian process regression.Acta Astronautica, 2019, 157: 73–91

work page 2019

[28] [29]

Real-time control for fuel-optimal moon landing based on an interactive deep reinforcement learning algorithm.Astrodynamics, 2019, 3(4): 375–386

Cheng L, Wang Z, Jiang F. Real-time control for fuel-optimal moon landing based on an interactive deep reinforcement learning algorithm.Astrodynamics, 2019, 3(4): 375–386

work page 2019

[29] [30]

Modeling irregular small bodies gravity field via extreme learning machines and Bayesian optimization.Advances in Space Research, 2021, 67(1): 617–638

Furfaro R, Barocco R, Linares R, Topputo F, Reddy V, Simo J, Le Corre L. Modeling irregular small bodies gravity field via extreme learning machines and Bayesian optimization.Advances in Space Research, 2021, 67(1): 617–638

work page 2021

[30] [31]

Philosophiæ Naturalis Principia Mathematica

Newton I. Philosophiæ Naturalis Principia Mathematica. Royal Society, 1687. A Appendix In this appendix we provide further experimental details and parameter specifications used throughout this study. A.1 Tetrahedral Mesh Models Table 6 summarizes the discretized tetrahedral mesh models used in this study, while Table 7 details the mathematical functions ...

work page