Optimizing the 2D trajectory of a robot from scratch using the Levenberg-Marquardt method for non-linear least squares.
In a nutshell, this project is on Pose Graph Optimization (PGO) which is typically used in most of today's SLAM Backends. The project involves:
- Theoretical Introduction: PGO theory and 1D SLAM solved example walkthrough (redirected to Notion pages for in-depth theory).
- Scratch: PGO Implementation from scratch on simple dataset using tools for evaluation/visualization like
EVO
,g2o viewer
etc. - Using graph optimization framework G2O: PGO using G2O library on multiple datasets using tools for evaluation/visualization like
EVO
,g2o viewer
etc. - PGO related survey paper reading (Optional).