Voxel-based Density Models for Accurate Gravitational Field Computation

• DOI: 10.5194/egusphere-egu24-1294
• Publication Date: 2024-03-08T10:59:06Z
• Authors: Benjamin Haser; Thomas Andert; Roger Förstner

Asteroids and moons are promising targets for physical space exploration. The use of physically-based simulations within a virtual environment (deep) missions can significantly benefit the testing validation guidance, navigation, control algorithms. This approach offers advantages in terms cost time efficiency. Especially orbit propagation landing maneuvers, information about gravitational field is crucial. However, several factors contribute to complexity this task, such as limited available inner structure celestial bodies. lack detailed knowledge their shapes further adds challenge. study presents voxel-based mass concentration (MASCON) method model realistic density distributions, enabling accurate gravity determinations. We chose cube with constant first case due perfect shape reconstruction availability an analytical solution its field. To validate our results, we calculated surface compared it solution, ensuring accuracy calculations. Furthermore, derived different resolutions against other state-of-the-art methods like polyhedral that provides closed-form homogeneous density. two also MASCON approach, one utilizing polydisperse sphere packing another represented spherical coordinates. relative errors acceleration between four will be evaluated sphere, second aspect was create tool generates distributions. able successfully reproduce natural environments by placing body-specific restrictions on three-dimensional Perlin noise additional normalization. simulator add following structural features distribution: arbitrary number centralized or decentralized shells, varying thickness densities, anomalies size shape, only restricted maximum permille body's volume. implemented normalization techniques keep all generated bodies fixed. Our results show generate distributions calculate corresponding correctly. data here used train Machine Learning Deep algorithms inversion.