Shaylah Mutschler, University of Colorado Boulder; Penina Axelrad, University of Colorado Boulder; Tomoko Matsuo, University of Colorado Boulder; Eric Sutton, University of Colorado Boulder
Keywords: Orbital Debris, SSA, LEO, Space Weather, Atmospheric Drag, Density, Acceleration Estimation, Particle Filter, Ensemble Filter, Physics-based Space Environment Model
Abstract:
A key requirement for accurate trajectory prediction and Space Situational Awareness (SSA) is knowledge of the non-conservative forces affecting space objects. These effects vary temporally and spatially and are primarily driven by the dynamical behavior of space weather. Existing SSA algorithms adjust space environment models based on observations of calibration satellites. However, lack of sufficient data and mismodeling of non-conservative forces can cause inaccuracies in space object motion prediction, especially for uncontrolled debris objects. On the other hand, the uncontrolled nature of debris objects can make them particularly sensitive to the variations in space weather. Our research takes advantage of this behavior by utilizing observations of debris objects to infer the space environment parameters controlling their motion.
We explore rigorous and practically realizable means to utilize debris objects as passive, indirect sensors of the space environment. The focus is on atmospheric density for more accurate prediction of Low Earth Orbit (LEO) object motion. In our previous work, a Partially Orthogonal Ensemble Kalman Filter (POEnKF) was implemented to assimilate observations of multiple debris objects, and estimate atmospheric density in addition to the position and velocity of each debris object. In this work, a methodology is developed that aims to estimate forcing parameters of a physics-based space environment model to allow for improved density estimates and predictions. This tool consists of two filters within a closed-loop feedback system.
The first filter utilizes debris object tracking data in the form of measurements collected from ground sensors to estimate acceleration due to atmospheric drag. A couple orbital periods of accumulated debris object data are assimilated and the resulting estimated accelerations, as well as time, position, and debris object information, are passed to the second filter. The second filter, an Particle Filter (PF), estimates forcing parameters of a physics-based space environment model, the Thermosphere Ionosphere Electrodynamic General Circulation Model (TIE-GCM). TIE-GCM is used to generate a cloud of forecast density particles in the filter. The density particles combined with corresponding debris object ballistic coefficients forms a predicted acceleration measurement. The PF applies corrections to the cloud of predicted acceleration measurements using the first filters acceleration estimates as measurements.
Ensemble filters have typically been employed in high-dimensional non-linear geophysical applications, such as weather forecasting of atmosphere and ocean systems. However, in the framework proposed here, a PF provides a viable option for the second stage filter. Because our approach effectively reduces the state dimension to only a small number of forcing parameters (n state elements), it becomes feasible to implement the necessary particle space of 10n particles. The PF performance for the second filter will be presented and analyzed.
Our current approach assumes that the ballistic coefficients of debris objects are known to a reasonable accuracy. This assumption is based on the idea of initializing the ballistic coefficients with information from the High Accuracy Satellite Drag Model (HASDM), which is regularly used in operations for updated density estimates. To accomplish this, density information from HASDM, when available, will be combined with debris object measurements to infer debris object ballistic coefficients.
The end goal is a data assimilation framework capable of tolerating high-dimensional systems with nonlinear dynamics and sparse observations of specific objects, while also leveraging a class of observations not typically utilized. While all observations provide some information, regardless of their uncertainty, the cost of particular observations may outweigh their information contribution. Therefore, we will investigate what measurements are most cost effective. These measurements are identified by type (range, azimuth, etc.), sensor location, measurement uncertainty, and also by debris object class/orbital regime. Overall, our tool is expected to improve atmospheric density estimates, which will allow for more accurate LEO object motion prediction.
Date of Conference: September 17-20, 2019
Track: Astrodynamics