Feature-Based Post-Entry State Determination Using Gaussian Mixtures
William N. Fife, Kyle J. DeMars · 2024
This work investigates the problem of post-entry navigation state determination without the dependence on a priori orbit information, which will become prevalent as multiple vendors start to collaborate on planetary missions. The lack of prior orbit information leads to significantly larger initial uncertainties which results in degraded performance for methods such as the extended Kalman filter. This work uses a feature-based, surface-relative approach with an extended Gaussian mixture filter equipped with a traditional sensor suite. Initialization of the filter is performed via line-of-sight to a chosen reference feature, and uniform probability is assigned in the altitude and downrange positions based on a priori attitude knowledge. Monte Carlo analysis demonstrates that with only a single feature, the filter can converge in sufficient time while being statistically conservative compared to an unscented Kalman filter.