Autonomous optical navigation for small spacecraft in cislunar space

The Earth-Moon system is expected to be increasingly populated especially due to large space missions like the Lunar Gateway. Communication constellations and small spacecraft are also expected to be deployed in the coming years. However, small spacecraft missions are heavily challenged by limited access and high costs of communicating with ground antennas, and constraints on on-board components. Hence, small spacecraft could greatly benefit from the development of autonomous optical navigation, allowing for higher levels of independence, flexibility, and lower operation costs in a complex environment like cislunar space. The scope of this study is to assess the optical navigation (OPNAV) accuracies achievable through horizon-based position estimation algorithms applied to synthetic images of the Moon. The simulation framework is implemented in Python+Vapory+OpenCV, and it simulates the image acquisition process at different solar phase angles and distances (including camera performance), and the processing of images (involving limb-fitting, centroiding and relative position estimation). A conventional ellipse fitting (EF) technique and the recently proposed Christian-Robinson (CR) approach are compared. Lastly, a Monte Carlo analysis allows to assess the effects of different sources of errors and uncertainties and to gain a better understanding of the navigation accuracies that may be obtained in future cislunar scenarios of interest. Preliminary results show that the CR algorithm consistently provides better performance than the EF technique, both in accuracy and computational time. The resulting errors in range estimation depend on distance, with values of approximately 22 km at a distance of 20, 000 km and up to 422 km at a distance of 150, 000 km. In terms of cross-axis distance estimation, the errors present values around 3.55 km (36 arcsec) at 20, 000 km, and 11.85 km (16 arcsec) at 150, 000 km. Furthermore, simulation outcomes show near normally distributed OPNAV performance with standard deviations around 0.68 km and 0.73 arcsec. A power law function is also provided to approximate the trend of OPNAV accuracies as a function of the distance relative to the observed body, is found based on the outcome of this study. Colour maps illustrating OPNAV accuracies as a function of observation geometry provide a comprehensive picture of the expected performance of OPNAV depending on distance and orientation relative to the observed object.