2026 · Computer Vision
2D Graph SLAM
Recovering a robot’s path and landmark locations from noisy motion and sensor measurements.
- Noisy measurements
- Graph constraints
- Trajectory and landmarks
Reproducible simulationGitHub ↗
Problem
A robot needs to estimate where it has travelled and locate landmarks from noisy measurements.
My role
Implemented the 2D simulator, constraint model, solver, and command-line demo.
Approach
- Simulate robot motion and landmark observations.
- Encode motion and measurement constraints in an information matrix and vector.
- Solve the system and compare the estimated trajectory and landmarks with simulated ground truth.
Results
- The documented seed-42 run reports a pose RMSE of 1.42 and landmark RMSE of 0.89 in simulation units.
- The repository documents 45 tests, including a hand-derived case.
Stack
- Python
- NumPy
- Matplotlib
- pytest