Physics-informed neural ODE for periodic orbits continuation
Authors:
Open Access
- 1 September 2026
- conference paper
- Published by IOP Publishing in Journal of Physics: Conference Series
- Vol. 3319 (1)
Abstract
To overcome the vulnerability of traditional numerical continuation methods when computing periodic orbits with extreme initial-value sensitivity in the Circular Restricted Three-Body Problem (CR3BP), this paper proposes a Physics-Informed Neural ODE (PINO) framework for orbit family continuation. The method uses Neural ODEs as a learnable dynamics solver and embeds periodic boundary conditions and orbital symmetry constraints into the loss function to enforce physical consistency during orbit correction. A parameter mapping network takes initial position as input and outputs period and initial velocity, enabling continuous prediction of orbit parameters after physics-enhanced training. The iterative “predict–refine–expand” workflow progressively covers the target interval while ensuring accuracy at each point. Simulation results confirm that the PINO framework effectively achieves periodic orbit continuation, offering a reliable and efficient numerical tool for future cislunar mission design.Keywords
This publication has 14 references indexed in Scilit:
- Dynamics around the Earth–Moon triangular points in the Hill restricted 4-body problemCelestial Mechanics and Dynamical Astronomy, 2024
- Tracking an untracked space debris after an inelastic collision using physics informed neural networkScientific Reports, 2024
- Artificial Intelligence for Trusted Autonomous Satellite OperationsProgress in Aerospace Sciences, 2023
- Orbit Propagation from Historical Data using Physics-informed Neural ODEs2023
- Cislunar Periodic Orbits for Earth–Moon L1 and L2 Lagrange Point SurveillanceJournal of Spacecraft and Rockets, 2022
- Cislunar Periodic Orbit Analysis for Persistent Space Object Detection CapabilityJournal of Spacecraft and Rockets, 2021
- Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equationsJournal of Computational Physics, 2019
- Earth–Moon libration point orbit stationkeeping: Theory, modeling, and operationsActa Astronautica, 2014
- The invariant manifold structure of the spatial Hill's problemDynamical Systems, 2005
- The ?Halo? family of 3-dimensional periodic orbits in the Earth-Moon restricted 3-body problemCelestial Mechanics and Dynamical Astronomy, 1979