This repository contains implementations of FastSLAM 1.0 and EKF-SLAM on the MRCLAM dataset, together with tools for evaluation and visualization.
If you want to compute and visualize error metrics (ATE, RPE, and landmark RMSE), you should run:
fastslam_known_correspondencesekf_known_correspondences
These scripts:
- Run the full SLAM pipeline
- Compare estimated trajectories and maps against ground truth
- Produce quantitative error plots (ATE, RPE, landmark RMSE)
They are intended for offline evaluation and benchmarking. You should see something like:
If you want a real-time animation showing how SLAM evolves step by step (robot motion, particle spread, landmark estimates), you should run:
slam_gui
This mode focuses on intuition and visualization, not on final quantitative error metrics.
-
Error metrics / evaluation → run
fastslam_known_correspondencesorekf_known_correspondences -
Real-time SLAM animation → run
slam_gui
Both modes use the same underlying models but serve different purposes:
evaluation vs. visualization.
For a detailed explanation of the mathematical derivations, algorithm structure, and experimental results, please refer to the project report included in this repository:

