Module 4: NVIDIA Isaac Platform
This module covers NVIDIA's Isaac platform for advanced robotics development. Over weeks 8 to 10 you move from authoring simulated scenes, to GPU-accelerated perception, to planning paths and understanding how a humanoid walks along them.
The examples assume Isaac Sim 4.x (with the omni.isaac.core API), Isaac ROS 3.x, and ROS 2 Humble. The NVIDIA stack evolves quickly, so each chapter states its version assumptions and points out where APIs have changed.
Topics Covered
- Isaac Sim environment and Universal Scene Description (USD)
- Isaac ROS integration and NITROS acceleration
- Visual SLAM (cuVSLAM) and nvblox 3D reconstruction
- Synthetic data generation with Replicator
- Nav2 path planning for humanoids
- Bipedal locomotion, from ZMP and LIPM to RL policies in Isaac Lab
Chapters
- Isaac Sim and USD (Week 8): Omniverse architecture, USD stages, prims, layers, references, and variants, plus standalone Python simulation of an articulated robot
- Isaac ROS Perception (Week 9): NITROS, Visual SLAM, nvblox, the Isaac Sim ROS 2 bridge, and synthetic data with Replicator
- Navigation and Locomotion (Week 10): Nav2 configuration, humanoid-specific navigation, gait phases, ZMP and LIPM balance, RL locomotion, and obstacle avoidance
Hardware Note
Isaac Sim requires an NVIDIA RTX-class GPU. Check NVIDIA's current Isaac Sim requirements page before setting up a workstation, or use a cloud GPU instance.
Prerequisites
Complete Module 3: Simulation Environments