Approximating Periodic Orbits with Algebraic Curves and Related Minimal Problems

Ruiqi Huang, Anton Leykin · 2026

Read Approximating Periodic Orbits with Algebraic Curves and Related Minimal Problems by Ruiqi Huang, Anton Leykin (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-26332.txt.

Introduction

Huang and Leykin fit low-degree algebraic curves to families of periodic orbits in the circular restricted three-body problem. That problem follows a weightless body pulled by two larger primaries, and the curve fits give one-parameter orbit families. The families feed minimal problems for liaison navigation, where craft states are read off from measurements between craft. For typical parameters, the count of solutions matches the degree of a branched covering map, found by both symbolic and numerical means.

Catalog record

CREATORRuiqi Huang, Anton Leykin
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE26550
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.26332

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Approximating Periodic Orbits with Algebraic Curves and Related Minimal Problems Ruiqi Huang, Anton Leykin Abstract The Circular Restricted Three-Body Problem (CR3BP) models the motion of a massless body under the gravitational influence of two primaries. We present a method for approximating a given family of periodic orbits by low-degree implicit algebraic curves, producing one-parameter families of algebraic orbit models. These models enable the construction of minimal problems motivated by liaison navigation, where spacecraft states are inferred from inter-spacecraft measurements. Relevant applications include initial orbit determination and spacecraft positioning. Each minimal problem defines a parameterized family of instances; for generic parameters, the number of solutions equals the degree of the associated branched covering map. We compute these degrees using both symbolic and numerical methods, and we outline a homotopy-continuation-based solver construction that can be practical for low-degree cases. 1 Introduction The Circular Restricted Three-Body Problem (CR3BP) seeks to determine the motion of a massless body subject to the gravitational attraction of two massive primaries (e.g., Earth and Moon) [ 7 ] . The general three-body problem admits no closed-form solution [ 6 ] ; however, in the restricted setting, one finds rich families of periodic orbits, including the planar Lyapunov orbits and the three-dimensional Halo orbits around the collinear Lagrange points. In this article, for a known family of periodic orbits, we approximate data points on a given orbit with an algebraic curve of low degree. We then fit the data from multiple orbits in the family, obtaining a model depending on one parameter linked to the orbit’s energy. The main contribution of this work is to use the above models to formulate minimal problems for liaison navigation that use crosslink range, range rate, and line-of-sight measurements. For each minimal problem, we determine the algebraic degree , which is equal to the number of complex solutions in the generic case. The problems with small algebraic degrees admit efficient solvers using parameter homotopy continuation, which could quickly obtain approximate solutions that can be used as initial guesses for dynamics-aware methods such as the liaison navigation approach in [ 3 ] . 2 Preliminaries 2.1 Rotating Frame and CR3BP We work in the rotating (synodic) frame in which the two primaries of masses 1-\mu and \mu are fixed at (-\mu,0) and (1-\mu,0) on the x -axis, and the frame rotates around the z -axis at unit angular velocity. X Y x y m_{1} m_{2} P t Figure 1: Inertial (X,Y) and rotating frame (x,y) in the planar case. The state (x,y,z,\dot{x},\dot{y},\dot{z}) of the massless body is guided by the equations of motion: \displaystyle\ddot{x} \displaystyle=2\dot{y}+x \displaystyle\quad-\frac{(1-\mu)(x+\mu)}{r_{1}^{3}}-\frac{\mu(x-(1-\mu))}{r_{2}^{3}}, (1) \displaystyle\ddot{y} \displaystyle=-2\dot{x}+y-\frac{(1-\mu)y}{r_{1}^{3}}-\frac{\mu y}{r_{2}^{3}}, \displaystyle\ddot{z} \displaystyle=-\frac{(1-\mu)z}{r_{1}^{3}}-\frac{\mu z}{r_{2}^{3}}, where r_{1}^{2}=(x+\mu)^{2}+y^{2}+z^{2} and r_{2}^{2}=(x-(1-\mu))^{2}+y^{2}+z^{2} . We imagine the rotating frame sharing the center of mass of the two primaries with the inertial frame where angle t in Figure 1 changes linearly with time, so the dynamics ( 2.1 ) involve Coriolis and centrifugal forces. 2.2 Jacobi Constant and Periodic Orbits The system ( 2.1 ) is Hamiltonian; the conserved quantity is the Jacobi constant C=-2H_{r} , expressed explicitly as C=(x^{2}+y^{2})+\frac{2(1-\mu)}{r_{1}}+\frac{2\mu}{r_{2}}-(\dot{x}^{2}+\dot{y}^{2})=2\Omega-(\dot{x}^{2}+\dot{y}^{2}), (2) where the effective potential is \Omega=\tfrac{1}{2}(x^{2}+y^{2})+\tfrac{1-\mu}{r_{1}}+\tfrac{\mu}{r_{2}} . Each family of periodic orbits that we consider is parameterized by C . Lyapunov orbits are planar periodic solutions around the Lagrange points L_{1} or L_{2} . Halo orbits are three-dimensional periodic solutions with an out-of-plane component; they also exist around L_{1} and L_{2} . Both families are continuous in C . In our numerical experiments, we use orbit data from the JPL Three-Body Periodic Orbit Catalog [ 5 ] for the Earth–Moon system ( \mu\approx 0.01215 ). 3 Fitting Algebraic Curves 3.1 Polynomial Model for Lyapunov and Halo orbits Each Lyapunov orbit is symmetric about the x -axis; only even powers of y appear in an implicit polynomial equation that we use as our quartic model : g(x,y):=\alpha_{1}x+\alpha_{2}x^{2}+\alpha_{3}x^{3}+\alpha_{4}x^{4}+\alpha_{5}y^{2}+\alpha_{6}xy^{2}+\alpha_{7}x^{2}y^{2}+\alpha_{8}y^{4}=1, (3) with 8 free coefficients \boldsymbol{\alpha}=(\alpha_{1},\dots,\alpha_{8}) . For Halo orbits, change the coordinate frame to the basis \mathbf{u}=(-\tfrac{1}{\sqrt{2}},0,\tfrac{1}{\sqrt{2}}) , \mathbf{v}=(0,1,0) , \mathbf{w}=(\tfrac{1}{\sqrt{2}},0,\tfrac{1}{\sqrt{2}}) and fit two equations in the (u,v,w) -coordinates: g(u,v)=\alpha_{1}u+\alpha_{2}u^{2}+\cdots+\alpha_{8}v^{4}=1,\qquad h(u,v)=\beta_{1}+\beta_{2}u+\cdots+\beta_{9}v^{4}=w, (4) effectively reducing the model to a planar curve: the first equation describes the projected curve in the uv -plane, and the second gives the third coordinate w as a polynomial in (u,v) . 3.2 Fitting One Orbit L_{1} Moon L_{2} Given data points \{(x_{i},y_{i})\}_{i=1}^{N} on a single orbit ( fixed C ), the coefficients \boldsymbol{\alpha} are determined by the least-squares normal equations (A^{\top}A)\boldsymbol{\alpha}=A^{\top}\mathbf{1}, where A_{ij}=\phi_{j}(x_{i},y_{i}) and \phi_{j} are the monomials of g . The root mean square error has a unique minimum at the least-squares solution. However, a more geometrically meaningful metric is the mean Euclidean distance from the data points to the fitted curve, which takes more computational effort to approximate, let alone optimize. We note that optimizing the (approximate) mean geometric distance is a nonlinear problem, which can be solved by a local method (e.g., gradient descent) using the initial approximation from least-squares. 3.3 Fitting a Family of Orbits The simplest tool to use for fitting a model g(x,y,C)\in\mathbb{R}[C][x,y] for Lyapunov or g,h\in\mathbb{R}[C][u,v] for Halo is, again, the least-squares method. Two-stage procedure. For each orbit in a training set \{C_{k}\} , solve the least-squares problem to obtain coefficients \boldsymbol{\alpha}(C_{k}) . Then fit each component \alpha_{j}(C) independently as a low-degree polynomial in C (we use cubics in experiments). For practical purposes, we partition [C_{\min},C_{\max}] into subintervals that contain data for about the same number of orbits in the JPL catalog. One-stage procedure. Treat all orbit data simultaneously with a single least-squares solve: for instance, in the case of Lyapunov orbits, expand g(x,y,C) in monomials \{\phi_{j}(x,y,C)\} jointly in all three variables, assemble the design matrix A_{ij}=\phi_{j}(x_{i},y_{i},C_{i}) , and solve (A^{\top}A)\boldsymbol{\alpha}=A^{\top}\mathbf{1} for the coefficient vector \boldsymbol{\alpha} . In our numerical experiments with the Lyapunov family, the one-stage procedure doesn’t provide significant improvement over the two-stage for the quartic model. However, the performance of the two-stage for the sextic model is significantly poorer than the one-stage. 4 Minimal Problems for Navigation A minimal problem is a square system of parametric polynomial equations (same number of equations and unknowns) that has finitely many complex solutions for generic values of the parameters. The degree of a minimal problem is the number of complex solutions in the generic case, which equals the degree of the corresponding branched covering map (from the so-called solution variety to the parameter space). 4.1 Descriptive notation and minimal problem examples In liaison navigation [ 3 ] , spacecraft exchange crosslink range (distance) measurements d_{AB} and possibly range-rate \dot{d}_{AB} between point A and B ; a mothership (e.g., a moon-based or earth-based) observer may also supply line-of-sight (LOS) directions. We denote a mothership , a body with a known state, by M and a spacecraft , a body with an unknown state, by S ; edges in the measurement graph indicate known measurements, with edge type encoding the measurement type (see Table 1 ). To give an example, we describe in some detail the formulation of two minimal problems involving one mothership and two spacecraft (M2S). M2S (Same Orbit). M S S Consider a single mothership M at a known position (x_{M},y_{M}) , and two spacecraft A and B on the same Lyapunov orbit with an unknown Jacobi constant C . The

Read the complete text