Gerardo Rivera, West Virginia University; Jacob Griesbach, ARKA/Stratagem; Rachit Bhatia, West Virginia University; Piyush Mehta, West Virginia University
Keywords: covariance realism, optimal sequential filter, space domain awareness, lethal non-trackable, space debris
Abstract:
Tracking lethal non-trackable (LNT) objects is crucial for mitigating collision risks and maintaining the safety of space. These objects are often too small to be tracked effectively using conventional methods, thus making them a threat to operational satellites or other resident space objects (RSOs). It is not only important to maintain custody but also to ensure that a realistic covariance is sustained, as covariance information is crucial for robust decision-making, and provides a more comprehensive understanding of the dynamic nature of space systems. This project is supported by the Intelligence Advanced Research Projects Activity (IARPA) Space Debris Identification and Tracking (SINTRA) program and the aim is to improve tracking and orbit prediction of LNTs, thereby, enhancing space domain awareness (SDA) and space safety. This paper presents a mathematical formulation for a new stochastic unscented transform (SUT) to incorporate the additional statistics of probabilistic models. Conventionally, a (deterministic) unscented transform (UT) is limited to tracking a set of smartly sampled sigma points to approximate the input probability distribution. However, with the SUT, the uncertainty associated with the underlying system model can also be incorporated. This allows more accurate modeling of systems that can estimate uncertainty and can capture complex dynamics more realistically. Because of the generic nature of the model, this framework has a myriad of applications in various fields. The new stochastic unscented approach is particularly advantageous in complex environments where traditional deterministic models are insufficient and modeling using Monte Carlo techniques will be inefficient. The derivation and validation of the new SUT algorithm has been previously performed for the primary motivating example of accurately capturing the statistics of probabilistic inputs driving probabilistic density models. Here, the SUT extends the prior work by proposing a generalized method to incorporate such probabilistic models for space object tracking and space domain awareness. In this work, the SUT has been embedded within the CAR-MHF (Constrained Admissible Region-Multi Hypothesis Filter) algorithm to enhance its ability to model atmospheric density for LEO drag force modeling. This enables more accurate modeling of the dynamics and, thus, allows one to reduce process noise for unmodeled force perturbations. The SUT’s ability to handle stochastic dynamics provides a more realistic representation of the space environment, enabling CAR-MHF to make more informed multiple hypothesis decisions based on probabilistic predictions. A LEO space object is processed using CAR-MHF to compare both the standard (deterministic) UT and the new SUT. This novel framework can enable a significant improvement in space tracking and catalog maintenance with reduced computational burden, thus leading to enhanced space domain awareness.
Date of Conference: September 17-20, 2024
Track: Astrodynamics