Continuous Learning of Gravity Field Irregularities Around Small Bodies via Neural Hamiltonian ODEs
Abstract
We propose to learn the unknown dynamics in the proximity of a small body directly from tracking data, representing them as a feed-forward neural network embedded in the system Hamiltonian. The equations of motion form a Neural Hamiltonian Ordinary Differential Equation, whose variational equations provide exact training gradients: estimation uses position and velocity arcs at realistic noise levels, without acceleration or potential labels, and a continual learning approach warm-starts the network as new data are acquired. The known part of the Hamiltonian carries whatever is available, from the central term and spin state to the constant-density model of the imaged shape. We assess the method against a normalized spherical harmonics expansion estimated from identical arcs through the same machinery, on scenarios built on the shapes of Itokawa, 67P, Bennu and Eros. The network remains usable inside the Brillouin sphere: it plans ballistic descents at Itokawa to \SI{4.6}{m} median touchdown error from tracking alone, against 5.1--\SI{48.9}{m} for harmonics of degree 4--12, and to \SI{0.9}{m} with the imaged shape as prior, a configuration that also recovers localised density anomalies invisible to any harmonics degree. The two representations are complementary, and we discuss their combined use across the phases of a small-body mission.
Related
- Equatorial periodic orbits and gravitational wave phenomenology around spherically-symmetric vacuum solution in Freund–Nambu scalar–tensor gravity
- Surface gravity wave on a neutron star ocean trapped around a magnetic pole
- Decoding the Imprints of Energy-Momentum Squared Gravity in Neutron Stars with Machine Learning Analysis
- The occurrence rate of planets around subgiant stars from the TESS photometric survey
- Circumbinary discs as the origin of circumstellar material around interacting H-poor supernovae and fast blue optical transients
BibTeX
@misc{acciarini2026continuous,
title = {{Continuous Learning of Gravity Field Irregularities Around Small Bodies via Neural Hamiltonian ODEs}},
author = {Giacomo Acciarini and Dario Izzo},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.12022},
eprint = {2609.12022},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.12022},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Continuous Learning of Gravity Field Irregularities Around Small Bodies via Neural Hamiltonian ODEs
AU - Giacomo Acciarini
AU - Dario Izzo
PY - 2026
DA - 2026/09/10
PB - arXiv
DO - 10.48550/arxiv.2609.12022
UR - https://arxiv.org/abs/2609.12022
AB - We propose to learn the unknown dynamics in the proximity of a small body directly from tracking data, representing them as a feed-forward neural network embedded in the system Hamiltonian. The equations of motion form a Neural Hamiltonian Ordinary Differential Equation, whose variational equations provide exact training gradients: estimation uses position and velocity arcs at realistic noise levels, without acceleration or potential labels, and a continual learning approach warm-starts the network as new data are acquired. The known part of the Hamiltonian carries whatever is available, from the central term and spin state to the constant-density model of the imaged shape. We assess the method against a normalized spherical harmonics expansion estimated from identical arcs through the same machinery, on scenarios built on the shapes of Itokawa, 67P, Bennu and Eros. The network remains usable inside the Brillouin sphere: it plans ballistic descents at Itokawa to \SI{4.6}{m} median touchdown error from tracking alone, against 5.1--\SI{48.9}{m} for harmonics of degree 4--12, and to \SI{0.9}{m} with the imaged shape as prior, a configuration that also recovers localised density anomalies invisible to any harmonics degree. The two representations are complementary, and we discuss their combined use across the phases of a small-body mission.
ER -