360° SLAM
Problem
Build a SLAM system that can use 360-degree imagery for mapping and localization across a wide range of environments, especially where large fields of view and long trajectories matter.
Method
The project uses a panoramic SLAM pipeline with feature extraction and matching on 360 imagery, pose graph optimization, loop closure detection, and global refinement to maintain consistent maps over large-scale scenes.
Focus
- Robust localization from panoramic observations
- Loop closure and drift reduction
- Globally consistent maps from long trajectories