Intro to Omniverse and Isaac, Supplementary Materials

This repo contains supplementary materials for the talk “Introduction to the Omniverse and Isaac Platform for Robotics”.

1. Post-talk Survey [here]
2. Official NVIDIA Learning Paths (free)
   • OpenUSD Learning Path
   • Digital Twins for Physical AI Learning Path
   • Robotics Fundamentals Learning Path
3. Isaac Launchable Deployment ($) [here]
4. Test Isaac Lab on Colab (free) [here]
5. Join Taiwan Robotics (i.e., ROS Taiwan) [here]
6. GitHub Student Developer Pack [here]
7. Summary of all Isaac components [here]
8. Getting started with Omniverse, Isaac Sim, Isaac Lab, Isaac ROS, Jetson AGX, GR00T, Cosmos, R2D2, Newton, NVIDIA Warp, MuJoCo-Warp, Isaac for Healthcare, and more.
9. Opencode with Free Nemotron 3 Super Setup [here]

Abstract of the Talk

AI-powered robotics is expected to drive the next major wave of technological innovation. As intelligent robots become increasingly integrated into industries such as manufacturing, healthcare, and transportation, they are set to catalyze innovation and streamline operations across numerous sectors.

This talk offers a streamlined introduction to NVIDIA’s robotics platform, aimed at accelerating robotics development in today’s fast-changing landscape. Attendees will discover how to use NVIDIA Omniverse and OpenUSD to build high-fidelity digital twins for synthetic data generation and factory optimization. The session will cover Isaac Sim, a powerful simulation tool for building, testing, and validating robotic applications in realistic, physics-based virtual environments. We will also explore Isaac Lab, focusing on its capabilities for training AI models through reinforcement learning (RL) and imitation learning (IL), empowering robots to master complex tasks beyond the reach of traditional controllers. In addition, the talk will explain how to deploy robust robotic solutions with Isaac ROS, ROS2, and NVIDIA Jetson AGX - the leading embedded AI platform for robotics. Finally, we will introduce the latest SDKs, including Isaac GR00T, Cosmos, NVIDIA Warp, and Newton.

If time permits, attendees will be guided through a minimal Isaac Lab example demonstrating GPU-accelerated reinforcement learning, experiencing firsthand the benefits of GPU acceleration. The only requirement is a laptop with internet access.

By the end of the session, attendees will have a high-level understanding of how NVIDIA’s ecosystem supports robotics innovation, and how they can leverage these software and hardware tools for their own future projects. A curated list of resources is provided for individuals interested in delving deeper into robotics.

Getting Started Notes

Simulation

---

World Foundation Models (Cosmos)

Additional docs: Cosmos Cookbook

---

Other Tools

---

Agentic Workflows for Physical AI

An agent is an LLM plus a harness that supplies context, orchestration, tools/skills, memory, and security controls. For these examples, give the agent a measurable outcome and require it to run, record, inspect, and iterate—not merely generate code.

Local setup

Core assumption: These settings are for open-source work with no sensitive information or proprietary data. For enterprise use, see the NVIDIA agent stack section below, especially OpenShell, and apply your organization’s sandboxing, least-privilege access, secret management, network, and approval policies.

• x86 Ubuntu PC with an Isaac-supported NVIDIA GPU, current NVIDIA driver, Docker, and the NVIDIA Container Toolkit.
• A coding agent such as Claude Code, Codex, or OpenCode. Start with normal permissions and approve commands as needed; use a disposable VM and dedicated accounts if granting broader autonomy.
• Optional dedicated credentials for GitHub, Hugging Face, Docker Hub/NGC, Brev, or Run:ai.

For each run, keep the input data and baseline in the repo, ask the agent to prefer a containerized workflow, define the success check up front, and retain commands, logs, environment versions, metrics, and a screen recording as evidence.

Example prompts

Replace text in `` with paths or values from your environment.

NVIDIA agent stack

• Nemotron models
• OpenShell secure runtime
• NeMo Switchyard
• NeMo Agent Toolkit
• NVIDIA agent skills
• Isaac Sim MCP