[en] Accurate localization in surface robotics is essential for navigation, path planning, and science objectives. On Earth, absolute localization can be readily achieved via satellite navigation (e.g., GPS). For other planetary bodies such as the Moon or Mars, however, such systems are unavailable. Current methods for absolute localization of planetary rovers rely on time- and labor-intensive human visual matching of surface perspective features with satellite images. Relative localization also accumulates errors over time, with different methods estimating dissimilar locations (e.g., [1]). Thus, an absolute localization method that can quickly, automatically, and accurately reduce the position search space is of great benefit to future planetary exploration missions. This project [2] presents a new approach to localizing planetary rovers: training an artificial neural network to match surfaceperspective imagery to corresponding satellite maps.
Disciplines :
Ingénierie aérospatiale
Auteur, co-auteur :
Wu, Ben; National Astronomical Observatory of Japan
LUDIVIG, Philippe ; University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT)
Potter, Ross; Brown University
Chung, Andrew; Tensorlicious
Seabrook, Timothy; University of Oxford
Co-auteurs externes :
yes
Langue du document :
Anglais
Titre :
ABSOLUTE LOCALIZATION FOR SURFACE ROBOTICS IN GPS-DENIED ENVIRONMENTS USING A NEURAL NETWORK.
