Build Robot Simulation and Reinforcement Learning Workflows with Isaac Sim and Isaac Lab on DGX Spark
Introduction
Explore Isaac Sim and Isaac Lab for robotic workflows on DGX Spark
Set up Isaac Sim and Isaac Lab on DGX Spark
Run and Understand a Sample Robot Simulation with Isaac Sim
Train a Humanoid Locomotion Policy with Isaac Lab on DGX Spark
Next Steps

  Log an issue

  Fork and edit

  Discuss on Discord

About this Learning Path

Skill level:  Advanced
Reading time:  1 hr 30 min
Last updated:  29 Jul 2026
Skill level:
  Advanced
Reading time:
  1 hr 30 min
Last updated:
  29 Jul 2026
Authors:Johnny Nunez, NVIDIA
Odin Shen, Arm
Asier Arranz, NVIDIA
Raymond Lo, NVIDIA
Arm IP:Cortex-X Cortex-A
Tags: ML Linux Python Bash IsaacSim IsaacLab
Authors:
  • Johnny Nunez, NVIDIA
  • Odin Shen, Arm
  • Asier Arranz, NVIDIA
  • Raymond Lo, NVIDIA
Arm IP:
Cortex-X Cortex-A
Tags:
ML Linux Python Bash IsaacSim IsaacLab

Who is this for?

This is an advanced topic for robotics developers, simulation engineers, and AI researchers who want to run high-fidelity robotic simulations and reinforcement learning (RL) pipelines using NVIDIA Isaac Sim and Isaac Lab on Arm-based NVIDIA DGX Spark system powered by the Grace–Blackwell (GB10) architecture.

What will you learn?

Upon completion of this Learning Path, you will be able to:

  • Describe the roles of Isaac Sim and Isaac Lab within a robotics simulation and RL pipeline
  • Build and configure Isaac Sim and Isaac Lab on an Arm-based DGX Spark system
  • Launch and control a robot simulation in Isaac Sim using Python
  • rain and evaluate a reinforcement learning policy for the Unitree H1 humanoid robot using Isaac Lab and RSL-RL

Prerequisites

Before starting, you will need the following:

  • A NVIDIA DGX Spark system with at least 50 GB of free disk space
  • Familiarity with Linux command-line tools
  • Experience with Python scripting and virtual environments
  • Basic understanding of reinforcement learning concepts (rewards, policies, episodes)

Summary

AI-assisted

This summary was drafted with an approved AI-assisted workflow and reviewed by Arm contributors before publication. Human technical review remains part of the process so the final page reflects engineering rigor, accuracy, and Arm editorial standards.

Close
?
You’ll configure NVIDIA Isaac Sim and Isaac Lab on an Arm-based NVIDIA DGX Spark with Grace–Blackwell architecture. You’ll verify the platform, build Isaac Sim, and launch a Cartpole scene with Python. Then, you’ll train a Unitree H1 locomotion policy through Isaac Lab’s RSL-RL PPO integration. You’ll review training and evaluation output to recognize effective learning.

Frequently asked questions

AI-assisted

These FAQs were drafted with an approved AI-assisted workflow and reviewed by Arm contributors before publication. Human technical review remains part of the process so the final page reflects engineering rigor, accuracy, and Arm editorial standards.

Close
?
What should I check on my DGX Spark before starting the setup?
Verify the system configuration and confirm you have about 50 GB of free disk space. Plan for roughly 15–20 minutes of setup time on DGX Spark.
Which component should I install and configure first?
Build and configure Isaac Sim first, then set up Isaac Lab on top of the Isaac Sim environment. Isaac Lab relies on the Isaac Sim installation.
How do I confirm the installation worked before moving on to training?
Launch a pre-built sample scene from Isaac Sim and interact with it programmatically. If the scene runs without errors and responds to actions, the environment is ready.
What result should I expect when running the Cartpole example?
You’ll see a running simulation where the cart and pole respond to actions driven by your Python control loop. The running simulation validates the simulation loop and your ability to step and control the environment.
How do I know the humanoid policy is training correctly with Isaac Lab and RSL-RL?
Select the Unitree H1 task, start PPO training via Isaac Lab’s RSL-RL integration, and monitor training logs and evaluation outputs. Successful runs show improving evaluation results and stable locomotion over rough terrain in simulation.
Next