EKF-based SLAM

November 2023 · 1 min read

As a semester project, another student and I implemented this paper, which develops an offline, factor graph-based SLAM algorithm called Square Root SAM. This method estimates both the state of a mobile robot over time and the locations of stationary landmarks based on noisy sensor data and imperfect motion commands.

To maintain academic integrity, my implementation of this method is kept private in a Github repository. This project involved skills in:

State estimation
SLAM
Probability theory
Data Association
Dynamics Systems
Python

The following figures show the performance of Square Root SAM (top) over an EKF-based SLAM method (bottom). Both algorithms were run offline on the same data set. One benefit of the offline Square Root SAM method over the online EKF-based algorithm is that it can use future sensor data and data association to estimate the entire history of the robot’s trajectory.

Last updated on November 2023