Characterization of horizon-based optical navigation measurement errors around the Moon: Patterns, their evolution, and a modeling approach

Autonomous navigation is expected to play a key role in an increasingly populated cislunar space. Evaluation of optical navigation strategies, however, tends to be computationally expensive, often limiting the extent and depth of feasibility analyses. This work aims to thoroughly characterize the performance of horizon-based optical navigation measurements around the Moon, and to provide a solution to more efficiently incorporate high-fidelity measurement errors into simulation tools: allowing for extensive feasibility analyses of autonomous cislunar missions. More specifically, synthetic images are generated and processed over a comprehensive grid of observation geometries around the Moon, distances, time, and camera resolutions, a parametric analysis of resulting position measurement errors is provided (identifying continuity and trends of errors), and an approximation approach leveraging pre-computed measurement errors is proposed for efficient, high-fidelity modeling virtually anywhere in the Earth-Moon system. Results show that error mean values are of the same magnitude or larger than standard deviations, contrary to the common assumption of close-to-zero mean. Errors are notably observed to depend on observation geometry, showcasing regions of low and high error values, and these patterns are preserved and evolve across distance, resolution, and time. Lastly, linear interpolation of the pre-computed grid of position measurement errors is shown to provide an efficient and high-fidelity approximation method, with near-zero mean approximation errors and standard deviations within 25% of the true error, effective also when incorporating measurements into a Kalman filter. Findings contribute to a deeper understanding and thorough characterization of autonomous optical navigation performance in cislunar space, and provide mission designers and analysists with a low-demand, high-fidelity approach to model and incorporate horizon-based position measurement errors into two to three orders of magnitude more efficient simulation tools. • Position measurement errors depend on observation geometry and illumination • Measurement errors evolve smoothly with distance and resolution, and repeat in time • Horizon-based measurement errors are not close-to-zero mean as commonly assumed • Errors more closely resemble a normal distribution farther away from the Moon • Interpolation of pre-computed errors allows for efficient, high-fidelity modeling