Physics-informed Conditional Normalizing Flows for Angles-only Cislunar Orbit Determination

2026-06-29

AI Summary

This study addresses the challenge of accurately estimating the initial orbital state probability distribution in cislunar space using only short-arc angle-only observations. To overcome this limitation, the authors propose a physics-informed conditional normalizing flow method that formulates orbit determination as a conditional density estimation task. This work represents the first application of conditional normalizing flows to cislunar orbit determination, integrating physical constraints with nonlinear least-squares optimization to flexibly model potentially multimodal posterior distributions and generate statistically consistent, high-quality initial state hypotheses. The proposed approach provides effective warm starts for conventional orbit determination algorithms, significantly enhancing both accuracy and robustness under short-arc angle-only observation scenarios.

Abstract

Generative Astrodynamics is advanced in this work by extending generative modelling to an orbit determination problem in the cislunar environment. The task is formulated as conditional density estimation, aiming to infer the probability distribution of the initial state from angles-only measurements over short observation arcs. A normalising flow is trained on perturbed topocentric observations from Near Rectilinear Halo Orbits, enabling a flexible and potentially multimodal posterior representation. Given new measurements, the learned density is sampled to generate statistically consistent and physics-informed state hypotheses. These estimates are refined via nonlinear least-squares minimisation, providing a competitive warm start for classical algorithms.

Problem

• orbit determination
• angles-only
• cislunar
• initial state estimation
• conditional density estimation

Innovation

• Physics-informed
• Conditional Normalizing Flows
• Angles-only Orbit Determination
• Cislunar
• Generative Astrodynamics