Real-Time Visual-Inertial Odometry for Planetary Exploration: Preliminary Tests

초록

In the new space age, both numerous countries and companies with space programs have been participating in research areas of space exploration and spaceport construction as well. In South Korea, the Korea Pathfinder Lunar Orbiter, officially Danuri, was successfully launched in 2022 and has been taking missions of surveying lunar resources. Because of the orbiter, the Korean Aerospace Research Institute has announced that a lunar orbiter, a lunar lander, and a rover will be launched together on a Korea Space Launch Vehicle-3 by 2032. In the third phase of the Artemis program, a moon landing will be planned to spend a week exploring the surface in 2025. For such long-term missions in the lunar surface environment, In-Situ Resource Utilisation (ISRU) constructions will be essential for various purposes such as shelters, spaceports, blast barriers and so forth. In order to construct infrastructure, Geospatial information is crucial to be prepared. The lunar exploration rovers have been developed to take not only obtaining this information but also surveying resources. However, due to the extreme environmental conditions and technical constraints of the moon, there are numerous limited circumstances to estimate the actual location of the rovers and establish the geospatial information. To solve this problem, we developed a rover that applied Visual Inertial Odometry (VIO) and deep learning-based simultaneous localization and mapping (SLAM) algorithm. Accordingly, we would like to introduce the results of evaluating 3D mapping performances in the simulated planet test beds.