Geophysical investigations of celestial bodies’ atmosphere, surface and interior through Bayesian and ML-based techniques

Most recent NASA’s planetary Decadal Survey and ESA’s Voyage 2050 report assess that future space exploration missions across the Solar System will focus on improving our knowledge of planetary formation and evolution processes, and on searching for traces of life in habitable environments. This exploration program will be primarily devoted to the investigation of the icy moons of the giant planets and the terrestrial planets that share similarities with the Earth. These ambitious objectives will be achieved through a deep understanding of celestial bodies’ atmosphere, surface, interior and their geophysical properties. The comprehensive analysis of the investigated body will, thus, rely on Earth-based measurements and on the contribution of interdisciplinary dataset acquired by in-situ and remote sensing observations of interplanetary probes. The state of the art techniques that inform us about the properties of the target body’s atmosphere include radio occultation and spectroscopy. High-resolution spectral imaging supports the analysis of atmospheric composition, surface mineralogy and thermal emissivity, and contributes to the search for signs of active volcanism. Orbit-based optical and Synthetic Aperture Radar (SAR) imaging is crucial to characterize the surface morphologies of celestial bodies and understand and monitor the endogenic or exogenic processes that shape them. The detection of surface features, such as craters, rifts, fissures due to volcanic activity or deformation process, is pivotal for the development of updated morphological and geological maps and detailed catalogs. Radio tracking and optical-based (OpNav) navigation strategies play a fundamental role in retrieving precise spacecraft’s localization. The accurate reconstruction of spacecraft’s trajectory is pivotal to achieving the coherent processing of the interdisciplinary measurements provided by onboard instruments, as for example reliable georeferencing of surface imagery. Concurrently, the processing of range, Doppler and optical data enables the estimation of key parameters of the investigated central body, such as, gravity field, rotational state and tidal response. The joint inversion of these multidisciplinary datasets represents a challenging task to maximize the science return and infer the properties of the target body, including its interior structure. This dissertation presents a multidisciplinary approach to investigate the atmosphere, surface and interior structure of celestial bodies through Bayesian and machine learning (ML)-based techniques. Dedicated strategies to process radio tracking, optical, and SAR data have been designed, implemented, validated and tested, and are thoroughly discussed in this work. To accurately process radio-tracking measurements, a thorough modeling of the dynamical environment, including gravity and non conservative forces is key. This study focuses on the mission profile of the future ESA’s Venus orbiter EnVision, where the atmospheric drag is expected to be one of the main contributions to the spacecraft’s dynamics. Thereby, an approach to cope with dynamical mismodeling and concurrently estimate the target atmospheric density, geophysical parameters and the spacecraft’s trajectory is proposed. ML-aided image processing is also investigated to efficiently extract surface features and support surface mapping and spacecraft’s navigation, enabling autonomous and real-time processing of data. ML strategies are thereby trained to identify surface morphologies on Earth and terrestrial planets, such as Venus and Mars, from SAR and optical imagery. Additionally, accurately cataloged features are integral to OpNav navigation strategies that use a priori knowledge of the central body’s topography. Furthermore, ML-based feature tracking capabilities are integrated into a Visual Odometry inspired navigation framework. The proposed solution is independent from any a priori knowledge of the shape model and topography of the target. The results of the numerical simulations of the spacecraft’s navigation provide a first assessment of the achievable accuracies of this method in contexts where traditional optical navigation strategies are impractical, i.e., lack of catalogs and unknown surface topography, or when radio-tracking data are not available. The combination of spacecraft localization and interdisciplinary observations allows us to estimate the geophysical properties that constrain the interior of the investigated body. To address this objective, a statistically robust Bayesian framework is used to characterize the interior structure of two main targets of future exploration, Saturn’s icy moon Enceladus and Venus. The presented analysis provides reliable models of the internal layering consistent with the probability distributions of the observed geophysical parameters.