Smriti Paul, West Virginia University; Piyush Mehta, West Virginia University; Thomas Kelecy, The Aerospace Corporation; Ryan Coder, The Aerospace Corporation
Keywords: Probabilistic Space Weather, Orbital Drag, Collision Avoidance, Sensor Tasking, Covariance realism, Pc dilution
Abstract:
Orbital drag uncertainty remains a major challenge to the Space Situational Awareness (SSA) and Space Traffic Management (STM) of Low Earth Orbit (LEO) space objects. The resultant uncertainty in the prediction of a satellite’s position can impact space operations and decision-making including collision avoidance, sensor tasking, tracking and custody. The primary source of drag error and uncertainty is the thermospheric density which can be significant during active space weather conditions, e.g. geomagnetic storms. The upper atmosphere density errors are due to the limited knowledge and inaccurate modeling of the physical processes driving the variations of the neutral space environment and inaccurate forecasts for the model drivers. These errors contribute to LEO satellite prediction errors and in many cases lead to unrealistic covariances. The goal of this work is to demonstrate that more accurately characterizing atmospheric density errors using probabilistic density models will lead to improved LEO state uncertainty prediction performance.
Existing density models are deterministic and current operations account for the uncertainty either in a highly simplified manner or do not account for it at all. These uncertainties may not necessarily be representative of the true uncertainty. The operational High Accuracy Satellite Drag Model (HASDM) uses a simple time-invariant uncertainty model which is a function of perigee and solar activity. Because HASDM is limited to Department of Defense (DoD) use, commercial operators and providers often use the Mass Spectrometer Incoherent Scatter (MSIS) model which is deterministic and can be significantly biased. A common approach for modeling density uncertainty is to use an estimation technique and tracking measurements for a set of carefully selected objects where the uncertainty is represented by process noise. However, realistically projecting the process noise for forecasting purposes is a challenge and may result in unrealistic covariances.
The methodology followed in this work leverages recently developed data-driven probabilistic density models that provide spatial-temporal uncertainty estimates in density predictions as a function of space weather drivers. Additionally, since no single density model is without errors due to performance sensitivity that is a function of space weather conditions, a novel ensemble approach is demonstrated that consistently accounts for the interplay between density and ballistic coefficient. Three study results are presented in this work that address the following questions: (1) What is the impact of this ensemble density modeling approach on the calculation of collision probabilities? (2) How long does the predicted covariance remain Gaussian under density uncertainty supporting sensor tasking? (3) How might improvements to the density error characterization in state filtering, including drag estimates, produce realistic representations of satellite positional uncertainty? The improvements to density error characterization and prediction uncertainties should lead to improved space safety and STM in the increasingly congested LEO regime.
Date of Conference: September 19-22, 2023
Track: Atmospherics/Space Weather