Charles J. Wetterer, KBR; Jason Baldwin, Complex Futures LLC; Paul Billings, KBR; Christopher Craft, KBR; Juan Gutierrez, KBR; John Gaebler, KBR; Micah Dilley, KBR; Jill Bruer, Air Force Research Laboratory
Keywords: cislunar, SSA/SDA, estimation, process noise, orbit determination, maneuver detection
Abstract:
Orbit determination and maneuver detection for objects in cislunar space using a sequential estimation filter require a robust dynamics model. At a minimum, this time-update state function needs to include the gravitational influence of the Earth and Moon, but oftentimes higher-order aspherical gravitational effects, as well as the influences of other bodies (e.g., Sun) and solar radiation pressure (SRP), must also be included. Due to their complexity and the lack of full knowledge of the underlying forces, however, even a high-fidelity dynamics model will always have an unavoidable mismatch when compared with reality, manifesting as both mismodeled and unmodeled effects. This mismatch is especially deleterious in regimes of chaotic dynamics, such as in the cislunar environment. To account for the dynamics mismatch, a small process noise is typically added to the state covariance during the time update step in the filtering sequence. Insufficient process noise will lead to filter overconfidence, unresponsiveness, and divergence, while excessive process noise will reduce the accuracy and precision of the state estimate. This technique is also common in maneuver detection using an interacting multiple model (IMM) approach, where an inflated process noise is used to account for statistically possible maneuver accelerations in a seperate maneuver model, but excessive process noise (in the nominal model) greatly reduces sensitivity to maneuvers. A common process noise model assumes an additive random process with fixed variance applied at the time of each measurement. This fails to work adequately, however, when the observation cadence varies or when there are large observation gaps due to object observability, both of which are common occurrences in the cislunar regime. This model can be improved somewhat by implementing a time-dependent scaling factor on the variance, but the scaled noise approach still suffers from directional assumptions, usually resulting in a choice of process noise that overestimates the uncertainty in some directions in order to retain adequate uncertainty in other directions. Ideally, to capture the effect of process noise most accurately, it would be embedded in the propagation machinery, continuously integrated and allowed to evolve as the encountered orbital environment dictates. For a highly nonlinear environment such as cislunar, an adaptive sigma point filter is an effective approach for capturing the time evolution of the probability density function (PDF). Within this filtering approach, we apply process noise to the PDF within the numerical integrator by directly augmenting the sigma points in lieu of inflating the state covariance at the discrete measurement times. This paper compares the performance of this new process noise modeling to the previously developed simple and scaled process noise models for objects in cislunar halo orbits. The integrated process noise model performs as well as or better than the other process noise models with no need to tune the time interval between applications to avoid excessive information loss or misorientation of the uncertainty ellipse.
DISTRIBUTION A – Approved for Public Release; distribution is unlimited. AFRL/PA approval AFRL-2025-0763
Date of Conference: September 16-19, 2025
Track: Astrodynamics