Development of a SURF-based Lunar Terrain Relative Navigation (TRN) Model for Earth-Independent Spacecraft Localization

Navigating around the moon poses significant challenges for spacecraft, as deep-space tracking signals and radio communications are often disrupted or infrequent. Standard navigation devices, such as star trackers, might perform poorly due to glare from the lunar surface or be unavailable entirely due to pointing constraints. During such periods

of sparse communication, spacecraft must be able to accurately and autonomously derive position and orientation without the help of earth-based systems like the Deep Space Network. To achieve this, a suite of Inertial Measurement Units (IMUs), radar, and optical navigation is utilized to determine the spacecraft’s position relative to its target. One method of optical navigation for Earth-independent localization, Terrain Relative Navigation (TRN), matches real time images of the lunar surface captured by the spacecraft during descent to a known reference map through feature detection and matching algorithms. In this work, a SURF (Speeded-Up Robust Features)-based approach to feature detection and matching is outlined, along with the specific processes necessary to translate between flat images and a physical spacecraft position and orientation estimate. The localization estimates of this TRN model for a generic descent trajectory were found to have < 50m magnitude of error per estimate, and fell within viable mission runtime targets (> 2Hz per localization).