AI Insight
This study compares two mathematical coordinate systems for predicting the motion of celestial objects during close encounters. The Levi-Civita coordinate system dramatically outperforms traditional Cartesian coordinates when analyzing highly elliptical orbits, maintaining accuracy 4.7 to 8.3 orders of magnitude better in energy calculations. However, this coordinate system proves difficult to use with machine learning approaches, as neural networks failed to achieve comparable accuracy despite the improved mathematical framework.
Why it matters
Accurate prediction of close encounters between celestial objects is crucial for satellite collision avoidance, asteroid impact assessment, and spacecraft navigation. While this research demonstrates that better coordinate systems exist for such calculations, it also reveals that current machine learning methods cannot yet exploit these advantages, highlighting a gap between mathematical theory and practical computational approaches.
Understand the Science
arXiv:2607.20235v1 Announce Type: cross
Abstract: Classical regularization removes the binary-collision singularity from the Kepler problem, but its value as a representation for learned Hamiltonian dynamics has not been systematically isolated. We compare Cartesian and planar Levi–Civita formulations of a perturbed Kepler system with a smooth quadrupole potential. With the perturbation supplied analytically, a Levi–Civita Hamiltonian splitting holds the maximum relative energy error near $2.1times10^{-5}$ through eccentricity $e=0.99$, while the Cartesian splitting becomes unstable. This advantage persists at matched physical horizon and force-evaluation budget, where the regularized baseline is $3times10^{-5}$, about $4.7$–$8.3$ orders of magnitude below the Cartesian arm depending on eccentricity. In held-out high-eccentricity tests with matched sampling, regularized models produce finite rollouts in $40/40$ runs versus $0/40$ for Cartesian. However, the fixed-shell construction supplies the regularized model with the exact initial orbit energy, and survival still carries $mathcal{O}(1)$ energy error. Four neural residual objectives fail to approach the analytic result. Exact-feature controls show that the regularized residual is a four-monomial degree-6 polynomial that a direct least-squares solve fits to the baseline. The remaining exact-feature gap is due to severe raw-basis ill-conditioning: orthogonalization restores baseline fitting for L-BFGS in two iterations. Small MLPs remain at $mathcal{O}(1)$ rollout error even after gauge symmetrization. Levi–Civita coordinates therefore improve dynamical conditioning while worsening raw-basis optimization conditioning; accurate neural residual learning remains unresolved. This is a controlled falsification-plus-trade-off study, not a solution to learned close-encounter dynamics.