Overview

Tutorial: G1 Standing with PPO

In this hands-on tutorial, you will train a G1 humanoid to stand still on flat terrain using PPO in Isaac Lab.

Figure: The end result of this lab – a G1 humanoid that can stand upright on flat terrain using a learned PPO policy.

You will:

• Reuse Isaac Lab’s built-in G1 flat locomotion config.
• Specialize it into a standing task by changing commands and rewards.
• Register two Gym environments:
• G1-Stand-Flat-v0 (training)
• G1-Stand-Flat-Play-v0 (play/visualization)
• Configure an RSL-RL PPO runner for this task.
• Train the policy and visualize the result in Isaac Lab.

Prerequisites

You should:

• Be comfortable with Python.
• Have basic RL knowledge (states, actions, rewards, episodes).
• Have Isaac Lab installed and working.
• Have the g1_stand extension project checked out under <G1_STAND_ROOT>.

If you haven’t yet:

• See Home → Prerequisites & official docs for installation and Isaac Lab setup.
• For RL and Isaac Lab concepts, see the Robotics Fundamentals Learning Path or the Isaac Lab docs.

Tutorial flow

Follow the steps in order (use the sidebar or the links below):

1. Define the standing environment
2. Register Gym environments
3. Configure PPO for the standing task
4. Train the G1 standing policy (install the extension first if needed; see the extension README)
5. Visualize the learned behavior
6. Troubleshoot and extend