EKF SLAM

ECE276A Sensing and Estimation in Robotics @ UCSD, April, 2023

Previously, I worked on SLAM problem with Particle Filter Slam, but it’s sensitive to noise and requires extensive hyperparameter tuning.

In this project, with datasets (IMU data, landmark feature points from stereo camera) generated from road data of vehicle, I implemented Extended Kalman filter (EKF) for visual-inertial SLAM and got fine results without much hyperparameter tuning. Due to computational time on landmark mapping, I only selected a portion of the features for our EKF-SLAM. The result could be better if all the features are used. Click here for project document and link for repo!

Scene 1	Scene 2