Autonomous Navigation of Distributed Spacecraft using Graph-based SLAM for Proximity Operations in Small Celestial Bodies

Establishment of a sustainable human presence beyond the cislunar space is a major milestone for mankind. Small celestial bodies (SCBs) like asteroids are known to contain valuable natural resources necessary for the development of space assets essential to the accomplishment of this goal. Consequently, future robotic spacecraft missions to SCBs are envisioned with the objective of commercial in-situ resource utilization (ISRU). In mission design, there is also an increasing interest in the utilization of the distributed spacecraft, to benefit from specialization and redundancy. The ability of distributed spacecraft to navigate autonomously in the proximity of a SCB is indispensable for the successful realization of ISRU mission objectives. Quasi-autonomous methods currently used for proximity navigation require extensive ground support for mapping and model development, which can be an impediment for large scale multi-spacecraft ISRU missions in the future. It is prudent to leverage the advances in terrestrial robotic navigation to investigate the development of novel methods for autonomous navigation of spacecraft. The primary objective of the work presented in this thesis is to evaluate the feasibility and investigate the development of methods based on graph-based simultaneous localization and mapping (SLAM), a popular algorithm used in terrestrial autonomous navigation, for the autonomous navigation of distributed spacecraft in the proximity of SCBs. To this end, recent research in graph-based SLAM is extensively studied to identify strategies used to enable multi-agent navigation. The spacecraft navigation requirement is formulated as a graph-based SLAM problem using metric GraphSLAM or topometric graph-based SLAM. Techniques developed based on the identified strategies namely, map merging, inter-spacecraft measurements and relative localization are then applied to this formulation to enable distributed spacecraft navigation. In each case, navigation is formulated in terms of its application to a proximity operation scenario that best suits the multi-agent navigation technique. Several challenges related to the application of graph-based SLAM for spacecraft navigation, such as computational cost and illumination variation are also identified and addressed in the development of these methods. Experiments are performed using simulated models of asteroids and spacecraft dynamics, comparing the estimated states of the spacecraft and landmarks to the assumed true states. The results from the experiments indicate a consistent and robust state determination process, suggesting the suitability of the application of multi-agent navigation techniques to graph-based SLAM for enabling the autonomous navigation of distributed spacecraft near SCBs