How to create action-graph in NVIDIA Isaac Sim for ROS2

By Harshal Varpe

Summary

This video provides a complete walkthrough for creating Action Graphs in NVIDIA Isaac Sim to enable ROS 2 integration, focusing on manual teleoperation, odometry, and sensor publishing. It assumes you already have a robot USD file and basic Isaac Sim knowledge. The steps covered include:

• Preparing a clean Isaac Sim stage with a grid and robot model
• Adding and positioning a RealSense RGBD camera sensor
• Setting up camera name overrides for consistent ROS 2 topic naming
• Building the following Action Graphs:
• ROS 2 Clock Sync Graph for synchronizing simulation and ROS time
• Driver Graph using ros2 context, subscribe twist, differential controller, and articulation controllers to enable manual control of a Husky robot
• Odometry Graph using compute odometry, publish transform, and publish odometry nodes to publish accurate TF and odometry data
• Camera Graph using create render product, camera helper, and publish IMU nodes to publish stereo camera images, camera info, and IMU data to ROS 2
• Verifying all graph nodes with RViz 2 for TF trees, camera feeds, and topic publishing
• Manually driving the robot via ROS 2 Twist commands to confirm integration
• Troubleshooting common Action Graph configuration issues like missing target prims or misnamed joints

This setup enables real-time teleoperation, sensor streaming, and odometry mapping between Isaac Sim and ROS 2, forming the foundation for advanced robotic simulation workflows.