F3 · Robotics & Embodied
Advanced Robotics RL on DGX Spark (Isaac Lab Part 2)
Overview
Continuing L18's DGX Spark Isaac Sim / Isaac Lab environment, this guide trains Franka manipulation and contact-rich tasks, coordinates dual Shadow Hands with MAPPO/IPPO, and builds a humanoid motion policy via AMP. The workflow places RSL-RL, RL Games, skrl and Stable Baselines3 into the same Arm64 simulation environment, validating how different tasks, algorithms and checkpoints are used. This advanced baseline helps robotics developers choose an RL framework by task registration, algorithm support and checkpoint compatibility — rather than assuming a single library fits all tasks.
Scope of verification
- It is an advanced Isaac Lab reference workflow on top of L18, validating that Franka, dual Shadow Hands and humanoid AMP tasks can be trained and replayed in the specified DGX Spark simulation stack.
- Dependencies include a completed, version-compatible L18 environment, Isaac Sim/Isaac Lab, each RL library, task configurations, training parameters and matching checkpoints.
- Scope qualitatively confirms simulation visuals and training behavior; it provides no unified convergence, sample-efficiency or library performance comparison, and the policy is not validated on physical Franka, Shadow Hand or humanoid hardware.
Canonical source: https://learn.arm.com/learning-paths/laptops-and-desktops/dgx_spark_isaac_robotics2/
This page is a guided summary with boundary notes, not an official reproduction.
This page is a guided summary with boundary notes, not an official reproduction.