We explore the classic "chicken and egg" problem of robotics: how a robot needs a map to know its location, but needs to know its location to build that map.

In this video, we cover:

• The 5 Core Components of SLAM: From sensor data processing and odometry to mapping, data association, and the magic of loop closure (which corrects accumulated drift to keep the map consistent).
• Classical vs. Modern SLAM: We trace the evolution of SLAM from early probabilistic filters like EKF-SLAM and FastSLAM (using particle filters) to the modern revolution of Graph-SLAM, which treats navigation as a full optimization problem.
• Practical ROS 2 Implementations: See how these concepts apply in the real world using tools like SLAM Toolbox for 2D Lidar mapping and RTAB-Map for 3D visual SLAM with depth cameras.

Whether you are building self-driving cars, autonomous drones, or your very first ROS 2 rover, understanding the blend of geometry, probability theory, and optimization behind SLAM is where it all begins