Robot SLAM and Localization

A mobile robot can only move with purpose when it understands two things: what the space around it looks like, and where it stands within that space. This course focuses on exactly those two questions.

You will start with the ideas behind mapping and localization, then move quickly into applying them with tools widely used in the ROS2 ecosystem. The goal is to understand why each step works, not only how to run it.

Whether you are preparing for autonomous navigation or want to strengthen your mobile robotics foundation, this course gives you a clear and practical understanding of how robots perceive and place themselves in the world.

• Why SLAM is a chicken and egg problem, and the difference between Online SLAM and Full SLAM

• How encoders and LiDAR are used in SLAM, and why sensor drift distorts maps in the real world

• The intuition behind Graph SLAM, pose graphs, and loop closure for map optimization

• How maps are represented as binary and probabilistic occupancy grids

• The SLAM Toolbox architecture and its synchronous, asynchronous, and offline mapping modes with ROS Bag files

• Creating a SLAM package, configuring Lifecycle Nodes, Launch files, and YAML parameters, and mapping a simulated apartment world

• Saving maps with the save_map service and serializing them for continuous mapping

• How AMCL and the particle filter localize a robot, including its adaptive feature

• Configuring and launching AMCL with the map server and lifecycle manager, tuning odometry noise, initial pose, and particles

• Testing localization with teleoperation in simulation and recovering from a bad pose with global localization