Towards autonomous navigation guided by natural landmarks on the Moon.
- Paper number
IAC-24,A3,IP,232,x91275
- Author
Ms. Cristina Pérez Ramos, Instituto Nacional de Astrofísica, Óptica y Electrónica, Mexico
- Year
2024
- Abstract
Autonomous navigation using natural landmarks in an unexplored environment is a very difficult problem to handle. While there are many techniques capable of correctly matching predefined objects, few of them can be used for real-time navigation in an unexplored environment. An important unsolved problem is to efficiently select a minimum set of usable landmarks for localization purposes. As humans, before using GPS we relied on natural reference points that exist in our environment: a pole, a house of a certain color, a certain establishment, etc. On the moon there are craters, rocks, boulder aereas and the line of horizon. The problem lies in teaching the robot how to learn to differentiate stable natural landmarks. The lunar rover optical camera is used to provide navigation and terrain information in an exploration area. However, due to the low presence of atmosphere, the Moon has a homogeneous terrain with dark soil. Furthermore, in extreme environments, the rover has limited data storage with low computing power. Therefore, for successful exploration, it is necessary to examine feature detection and comparison methods that are robust to the lunar terrain and environmental characteristics. In this paper, the feature detection algorithms, SIFT, BRISK, ORB and AKAZE are comparatively analyzed with lunar terrain images taken by a lunar rover. The experimental results show that SIFT and AKAZE are the most robust to lunar terrain characteristics. AKAZE detects fewer feature points than SIFT, but feature points are detected and compared with high precision and with the lowest computational cost. AKAZE is suitable for fast and accurate navigation information. Although SIFT has the highest computational cost, the largest number of feature points are stably detected and combined. The rover periodically sends images of the terrain to Earth. Therefore, SIFT is suitable for the construction of 3D global terrain maps, since a large number of terrain images can be processed. Furthermore, these feature algorithms are analyzed and compared with CNNs such as EfficientNet-V2. The results of the study are expected to provide guidance for using feature detection and comparison methods for future lunar exploration vehicles.
- Abstract document
- Manuscript document
(absent)