A
Camera-only mapping for mobile robots
Autonomous mobile robots usually lean on LiDAR or depth sensors to understand a room. This work asks how far one 360° camera can go on its own. Monocular imagery is reconstructed into a 3D Gaussian Splatting map, and that map is used to detect obstacles for navigation.
-
aCapture
One equirectangular 360° frame. No LiDAR, no depth.
-
bReconstruct
Build a 3D Gaussian Splatting map from monocular imagery.
-
cDetect
Find obstacles in the reconstructed map.
-
dNavigate
Plan a path around what the camera found.