Developer builds AMR for warehouse automation with ROS 2 Nav2 stack
Full-stack robot from Fusion 360 to ROS 2, solving LiDAR TF drift in turns.
A developer named Sourav24 has shared a detailed build of an Autonomous Mobile Robot (AMR) for warehouse automation, covering the entire pipeline from CAD design to ROS 2 deployment. The chassis was designed in Fusion 360 and exported as a URDF for simulation. The full stack uses ROS 2 with Nav2 for path planning, ArUco markers for visual docking alignment, and custom waypoint sequencing to handle multi-shelf tasks. Simulation was carried out in Gazebo with RViz for visualization.
The project faced a critical challenge: during turns, the LiDAR point cloud rotated with the robot in RViz, breaking mapping and navigation. Root cause was an odom/TF mismatch. The developer resolved this by building a GroundTruthOdom node that pulls pose data directly from Gazebo and publishes a consistent /odom transform, including handling ROS-Gazebo timestamp discrepancies. The robot autonomously services requests for Shelf B and Shelf C, delivering items to a drop-off zone. This project demonstrates a practical, open-source approach to warehouse robotics.
- Chassis designed in Fusion 360 and exported as URDF for ROS 2 simulation.
- Uses Nav2 for navigation, ArUco markers for precise docking, and custom waypoint sequencing.
- Solved LiDAR TF rotation issue with a GroundTruthOdom node using Gazebo pose data.
Why It Matters
Shows how open-source ROS 2 stack can enable custom warehouse AMR development from CAD to deployment.