Mapping the Moon in Multitudes: Development of a multi-rover system for mapping the Lunar surface
Translated title
Mapping the Moon in Multitudes
Authors
Thuesen, Astrid Sofie ; Christiansen, Christoffer ; Thorhauge-Hansen, Jonas
Term
4. semester
Education
Publication year
2026
Pages
85
Abstract
This thesis explores how teams of different types of robots can autonomously explore and collaboratively map unfamiliar areas on the Moon. The system was built using ROS2 (a robotics software framework) and Nvidia Isaac Sim (a robot simulator). It combines frontier-based exploration—moving toward the boundary between known and unknown terrain—for areas without prior information, coverage planning—systematic paths to cover every part—for areas with partial map data, and merging the separate maps produced by each rover. Frontier exploration was implemented with a custom clustering method; coverage planning used Fields2Cover; and map matching proposed a feature-based approach using HOG-DG descriptors extracted from height maps. Tests showed that dividing large exploration regions into smaller parts improved efficiency, that using multiple rovers reduced total mapping time, and that when prior map information was available, coverage planning outperformed frontier exploration. Offline experiments found that HOG-DG produced usable map matches and achieved better match quality than SIFT and ORB, although robust live map merging on ROS2 data was not yet achieved.
Denne afhandling undersøger, hvordan teams af forskellige typer robotter autonomt kan udforske og sammen bygge kort over ukendte områder på Månens overflade. Systemet blev udviklet i ROS2 (et robotsoftware-framework) og Nvidia Isaac Sim (en robotsimulator). Det kombinerer frontier-baseret udforskning—at bevæge sig mod grænsen mellem kendt og ukendt terræn—for områder uden forudgående information, coverage planning—systematiske ruter der dækker hvert område—for områder med en delvis kortlægning, samt fletning af de separate kort, som hver rover producerer. Frontier-udforskning blev implementeret med en skræddersyet klyngebaseret metode; coverage planning brugte Fields2Cover; og kortmatching foreslog en feature-baseret tilgang med HOG-DG-deskriptorer udtrukket fra højdekort. Test viste, at opdeling af store udforskningsområder i mindre dele forbedrede effektiviteten, at brug af flere rovere reducerede den samlede kortlægningstid, og at når der fandtes forudgående kortinformation, klarede coverage planning sig bedre end frontier-udforskning. Offline-eksperimenter viste, at HOG-DG kunne producere brugbare kortmatches og gav bedre matchkvalitet end SIFT og ORB, selvom robust live fletning af kort på ROS2-data endnu ikke blev opnået.
[This apstract has been rewritten with the help of AI based on the project's original abstract]