Author: Valentyn Piskovskyi
Python implementation of Graph SLAM using 3D LiDAR point clouds from the "Scenario 1" of the Velodyne SLAM Dataset.
The code is adapted from a reference MATLAB implementation with several different design choices. Key methods used:
- Point cloud registration: Point-to-plane ICP (via Open3D)
- Place recognition / loop detection: Scan Context descriptors
- Graph optimization: Levenberg-Marquardt
Scenario 1 from the Velodyne SLAM Dataset is downloaded automatically in the notebook.
pip install numpy open3d scipy pandas opencv-python matplotlib tqdm