Designing a Visual and LiDAR SLAM Stack for a Real Robot

Building a SLAM stack that runs reliably on a real robot involves more than choosing a library. The sensor calibration, front-end design, back-end optimization, loop closure strategy, and integration with the rest of the robot all shape whether the stack survives the messy reality of deployment. This course walks through those choices in a structured way. You will work through written design exercises that mirror how a small robotics team would plan a SLAM stack. The emphasis is on the practical tradeoffs that matter when a robot has to operate across long sessions, lighting changes, and unexpected scenes. What you'll learn: - Choose sensor combinations including visual, LiDAR, inertial, and wheel encoders for specific environments - Design front-end perception pipelines for feature detection, association, and motion estimation - Apply back-end optimization including pose graph methods and modern factor graph frameworks - Build loop closure detection that handles repetitive scenes and changing environments - Plan calibration routines including intrinsic, extrinsic, and time synchronization between sensors - Design integration with motion planning, control, and behavior systems on the robot The course progresses from sensors and calibration through front-end and back-end design, loop closure, and integration. A capstone written exercise asks you to draft a one-page design for a SLAM stack targeted at a specific robot platform and environment. This course is designed for robotics engineers, computer vision practitioners, and students of mechatronics with some software background. No prior SLAM implementation experience is required. The course treats the stack as a design problem you can reason about on paper before any code is written.