Training and testing robots using Isaac Lab
In this section, we will briefly discuss the learning process with Isaac Sim and Isaac Lab.
Isaac Lab architecture
Figure 15.6 shows the architecture of the model’s training and deployment and highlights its major building blocks.

Figure 15.6: Isaac Lab architecture (source: https://isaac-sim.github.io/)
Here are the major components of training and testing using Isaac Lab and Isaac Sim:
- Asset Input: Isaac Sim and Isaac Lab use Universal Scene Description (USD) [10] as a file format for representing 3D robots and environments. In ROS 2, we used URDF/xacro for the robot model. Isaac Sim offers conversion tools to USD from URDF, MJCF, XML, etc. Importing robot and environment assets into Isaac Lab is the first step in the workflow.
- Creating Configuration: After creating the assets as input, we have to create 2 configuration classes in Python:
- Asset Configuration: In this class, we can...