Higher-Order Tensor-Based Deferral of Gaussian Splitting for Orbit Uncertainty Propagation

G. Andrew Siciliano , Keith A. LeGrand, Jackson Kulik · 2025

Read Higher-Order Tensor-Based Deferral of Gaussian Splitting for Orbit Uncertainty Propagation by G. Andrew Siciliano , Keith A. LeGrand, Jackson Kulik (2025) 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-2507-01771.txt.

Introduction

Siciliano, LeGrand, and Kulik improve orbit uncertainty tracking in a 2025 paper. Space watchers must know where an object may drift, not just where it is. Gaussian mixture methods split clouds of possible paths for accuracy.

Splitting too early wastes computing power on calm regions. Their method defers splits until nonlinearity truly demands them. Higher-order tensor math judges where refinement is needed. Dynamics and initial uncertainty both guide the split directions. The scheme pairs with mixand reduction to control total cost. Readers in orbital mechanics get sharper uncertainty at lower cost.

Catalog record

CREATORG. Andrew Siciliano , Keith A. LeGrand, Jackson Kulik
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE65 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-01771

Browse the full library · About the Tower of Babel

Full text

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

Higher-Order Tensor-Based Deferral of Gaussian Splitting for Orbit Uncertainty Propagation G. Andrew Siciliano , Keith A. LeGrand, Jackson Kulik Abstract Accurate propagation of orbital uncertainty is essential for a range of applications within space domain awareness. Adaptive Gaussian mixture-based approaches offer tractable nonlinear uncertainty propagation through splitting mixands to increase resolution in areas of stronger nonlinearities, as well as by reducing mixands to prevent unnecessary computational effort. Recent work introduced principled heuristics that incorporate information from the system dynamics and initial uncertainty to determine optimal directions for splitting. This paper develops adaptive uncertainty propagation methods based on these robust splitting techniques. A deferred splitting algorithm tightly integrated with higher-order splitting techniques is proposed and shown to offer substantial gains in computational efficiency without sacrificing accuracy. Second-order propagation of mixand moments is also seen to improve accuracy while retaining significant computational savings from deferred splitting. Different immediate and deferred splitting methods are compared in three representative test cases, including a geostationary orbit, a Molniya orbit, and a periodic three-body orbit. I Introduction 0 0 footnotetext: This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible. Fig. 1 : Deferred Gaussian mixture splitting in orbit uncertainty propagation. To ensure safe space operations, a core function of sda ( sda ) is maintaining orbital state estimates of resident space objects. Accurate descriptions of orbital uncertainty are crucial, especially with the ever-growing number of active and defunct space objects. Though Gaussian representations offer convenient mathematical and computational properties, the nonlinearity of orbital dynamics can quickly render distributions non-Gaussian. This issue is only compounded when observation opportunities are temporally sparse. Existing work in nonlinear estimation has offered improved uncertainty propagation through better approximations of the transformed random variables’ statistical moments. These approaches include higher-order unscented transformation [ 1 ] , stt ( stt ) [ 2 ] , and da ( da ) [ 3 , 4 ] methods, which provide higher-order approximations of the transformed mean, covariance, and higher-order moments. The Gauss von Mises distribution introduced in [ 5 ] exploits a cylindrical state space and permits evaluation of the general Bayesian filter for orbital applications. Other work has focused on reducing the high computational burden of sampling-based methods using multi-fidelity approaches [ 6 ] . \Acp gm have also proved a powerful tool for modeling nonlinearity-induced non-Gaussian uncertainty distributions while retaining many of the advantages of Gaussian distributions. The stt can be used in coordination with these gm to propagate mixands over time [ 7 ] . The accuracy of a gm nonlinear transformation approximation typically relies on the mixture resolution, where higher resolution in the output mixture leads to a better fit with the true non-Gaussian density. In adaptive gm methods, splitting Gaussians permits a local increase in the resolution of the distribution by adding more mixands. To prevent computational waste, however, the directions of splitting must be chosen in a principled manner. Various splitting criteria have been proposed based on linearization error [ 8 ] , differential entropy [ 9 ] , total mechanical energy variance [ 10 , 11 ] , field of view bounds [ 12 ] , linearization change [ 13 ] , and other measures of nonlinearity [ 14 , 15 ] . Recently, [ 16 ] introduced new families of higher-order splitting objectives for both identifying nonlinearity [ 17 ] and optimizing the split direction for general nonlinear transformations. This paper introduces new methods for orbit uncertainty propagation based on the higher-order splitting techniques proposed by the authors in [ 16 ] . Many adaptive gm -based techniques for uncertainty propagation perform immediate splitting , wherein all splits of Gaussian mixands are made at the beginning of the considered time span. This work instead considers deferred splitting which, as depicted in Figure 1 , delays splitting of a mixand until the impacts of the system’s nonlinearities consistently exceed some specified tolerance. A deferred splitting algorithm, named the hotdogs ( hotdogs ) algorithm, is developed that leverages transition matrix and tensor composition to avoid repeated costly ivp ( ivp ) solutions. Second-order mean and covariance propagation is also shown to significantly improve gm approximations of the uncertainty. Furthermore, the specific structure of hotdogs ensures the computational savings compared to immediate splitting are retained when incorporating second-order propagation of moments. The performance of the uncertainty propagation using these methods is assessed using multiple figures of merit across three distinct, relevant cases spanning both two- and three-body dynamical systems. Preliminary results were presented in [ 18 ] . This work offers a more thorough treatment of the deferred splitting algorithm, its implementation, and its performance. The Gaussian splitting procedure itself is introduced in more detail, and the impact of the second-order propagation of moments is presented. The remainder of this paper is organized as follows. The mathematical problem formulation is developed in Section II . The deferred splitting and uncertainty propagation methodology appears in Section IV . Section III provides background on Gaussian splitting, whitening transformations, and partial derivative tensors. Numerical analysis results are discussed in Section V , and conclusions are drawn in Section VI . II Problem Formulation This work considers the problem of approximating the pdf ( pdf ) p_{\mathbf{Z}}(\mathbf{z}) of some nonlinear function \mathbf{z}=\mathbf{g}(\mathbf{x}) of a random variable \mathbf{x} with known distribution. Throughout this work, lowercase bold symbols are used to denote vectors and vector functions, and uppercase bold symbols denote matrices and tensors. Specific focus is given to the nonlinear transformation \mathbf{g} representing the flow of a deterministic dynamical system \bm{\varphi}_{\Delta t} over some time \Delta t [ 19 ] . The continuous-time dynamical system describing the evolution of a state vector \mathbf{x}\in\mathbb{R}^{n} is specified by the system of ordinary differential equations \frac{\mathrm{d}\mathbf{x}}{\mathrm{d}t}=\mathbf{f}(\mathbf{x}) (1) where \mathbf{f} is assumed, but not restricted, to be autonomous (i.e., time-invariant) for notational compactness. The flow map is then defined as the operator \displaystyle\bm{\varphi}_{\Delta t}(\mathbf{x}(t_{0}))=\mathbf{x}(t_{0}+\Delta t)=\mathbf{x}(t_{0})+\int_{t_{0}}^{t_{0}+\Delta t}\mathbf{f}(\mathbf{x}(\tau))d\tau (2) Though this work examines deterministic transformations of random variables, these heuristics and algorithms can be extended to filtering applications with process noise given suitable discrete-time approximations of the continuous-time process noise. III Background III-A Gaussian mixtures In Bayesian nonlinear estimation, gm offer a flexible parameterization of non-Gaussian uncertainty and can approximate any density (with a finite number of discontinuities) to an arbitrary degree of accuracy. A gm consists of L mixands , where each mixand is a tuple (w_{i},\mathbf{m}_{i},\mathbf{P}_{i}) comprising a weight, mean, and covariance, respectively. The gm pdf is then given by \displaystyle p(\mathbf{x})=\sum_{i=1}^{L}w_{i}\mathcal{N}(\mathbf{x};\mathbf{m}_{i},\mathbf{P}_{i}) (3) gm state estimation methods leverage the property that Gaussianity of a random variable is preserved under linear transformation. If the nonlinear map is quasi-linear within the local support of each mixand, then the linearly mapped mixand is approximately Gaussian, while allowing the composite mixture to be non-Gaussian [ 20 ] . If, on the other hand, the second- and higher-order behavior of the function is non-negligible over a mixand’s effective support, the mixand can itself can be replaced with more, smaller variance mixands in a procedure known as Gaussian splitting . III-B Univariate splitting The canonical splitting problem involves approximating the standard univariate Gaussian distribution q(x) by a mixture

Read the complete text