G1 Whole-Body Teleoperation

This repository provides two methods to control the G1 humanoid robot (29 DOFs) using AMO/HOMIE in MuJoCo:

1. Upper body inverse kinematics using Mink.
2. VR teleoperation using XLeVR.

Note: For VR teleoperation, a Meta Quest 3 or Quest 3S is required.

---

Installation

conda create -n gbot python=3.10
conda activate gbot
pip install -r requirements.txt

---

Usage

1. Mocap Control for G1 End-Effector

From the project root directory:

# AMO policy
python 01_g1_29_2f85_mocap_amo.py

# HOMIE policy
python 03_g1_27_2f85_mocap_homie.py
python 04_g1_27_2f85_mocap_cascade_pid_homie.py

• You can now drag the mocap cube to control the movement of the G1's end-effector.

---

2. VR Control for Full-Body Teleoperation

From the project root directory:

# AMO policy
python 02_g1_29_2f85_VR_amo.py

# HOMIE policy
python 05_g1_27_2f85_VR_cascade_pid_homie.py

• Open your VR headset browser and navigate to the HTTPS address shown in the terminal (e.g., https://<your-ip>:8443).

• Use your VR controllers to move the G1.

  • Grip button: Hold to enable VR teleoperation.
  • Trigger: Hold to grasp.

VR Controller Mapping

Input	Forward	Backward	Left	Right
Left Thumbstick	Height +	Height −	Yaw turn left	Yaw turn right
Right Thumbstick	Walk forward	Walk backward	Walk left	Walk right

---

Acknowledgements

• AMO — Adaptive Motion Optimization for Hyper-Dexterous Humanoid Whole-Body Control.
• HOMIE — Humanoid Loco-Manipulation with Isomorphic Exoskeleton Cockpit
• Mink — A library for differential inverse kinematics in Python, based on the MuJoCo physics engine.
• XLeRobot — Bringing Embodied AI to every family around the world.