Welcome to the Mapping and Localization for Mobile Robots and Autonomous Systems course repository. This repository contains projects and assignments related to the fundamental concepts and techniques used in mapping, sensing, and localization for autonomous systems. The course covers various topics and utilizes real-world data from the KITTI dataset to explore these concepts in practical scenarios.
The primary objective of this course is to equip students with the knowledge and skills required to enable mobile robots and autonomous systems to perceive and navigate their surroundings effectively. Throughout the course, students will dive into the following key subjects as can be seen in the next section.
-
Geodetic Coordinate System, KITTI Dataset & Grid Mapping
- Project: Probabilistic Occupancy Grid
In this project, we leverage the KITTI dataset to create and continuously update probabilistic occupancy maps. We explore both a basic approach and a camera-LiDAR sensor fusion method using deep learning techniques.
-
Kalman Filter, Extended Kalman Filter (EKF), and SLAM using EKF
- Project: Kalman Filtering for Pose Estimation
- Project: Extended Kalman Filtering for Improved Pose Estimation
These projects focus on the application of Kalman filters and Extended Kalman filters to mitigate sensor noise and improve the accuracy of 2D pose estimation.
-
Particle Filter, Iterative Closest Points (ICP), and Visual Odometry
- Project: Pose Estimation with Particle Filter
- Project: Scanning and Alignment with ICP
- Project: Monocular Visual Odometry Pipeline
In this section, we delve into the implementation of particle filters for pose estimation, the ICP algorithm for scan alignment, and the development of a monocular visual odometry pipeline. The projects provide hands-on experience with these essential techniques and their applications.
-
3D Object Detection Using PointPillars and Multi-Object Tracking Using BoT-SORT
- Project: 3D Object Detection with PointPillars
- Project: Multi-Object Tracking with BoT-SORT
This section explores advanced topics in computer vision, including 3D object detection using PointPillars and multi-object tracking with BoT-SORT. We analyze the performance of object detection and tracking on real-world datasets, gaining insights into scene understanding and tracking capabilities.
Each project is designed to enhance your understanding of mapping and localization techniques for mobile robots and autonomous systems. Please refer to the individual project folders for detailed instructions, code, and data.