64-BitBrainstorm_2026/unifolm-world-model-action

Files

2026-01-18 00:30:10 +08:00

6.3 KiB

Raw Blame History

Unitree Deploy

This document provides instructions for setting up the deployment environment for Unitree G1 (with gripper) and Z1 platforms, including dependency installation, image service startup, and gripper control.

0. 📖 Introduction

This repository is used for model deployment with Unitree robots.

1. 🛠️ Environment Setup

conda create -n unitree_deploy python=3.10 && conda activate unitree_deploy

conda install pinocchio -c conda-forge
pip install -e .

# Optional: Install lerobot dependencies
pip install -e ".[lerobot]"

git clone https://github.com/unitreerobotics/unitree_sdk2_python.git
cd unitree_sdk2_python  && pip install -e . && cd ..

2. 🚀 Start

Tip: Keep all devices on the same LAN

2.1 🤖 Run G1 with Dex_1 Gripper

2.1.1 📷 Image Capture Service Setup (G1 Board)

To open the image_server, follow these steps

Connect to the G1 board:

ssh unitree@192.168.123.164  # Password: 123

Activate the environment and start the image server:

conda activate tv
cd ~/image_server
python image_server.py

2.1.2 🤏 Dex_1 Gripper Service Setup (Development PC2)

Refer to the Dex_1 Gripper Installation Guide for detailed setup instructions.

Navigate to the service directory:
```
cd ~/dex1_1_service/build
```
Start the gripper service, ifconfig examines its own dds networkInterface:
```
sudo ./dex1_1_gripper_server --network eth0 -l -r
```

Verify communication with the gripper service:

./test_dex1_1_gripper_server --network eth0 -l -r

2.1.2 ✅Testing

Perform the following tests to ensure proper functionality:

Dex1 Gripper Test:
```
python test/endeffector/test_dex1.py
```
G1 Arm Test:
```
python test/arm/g1/test_g1_arm.py
```

Image Client Camera Test:

python test/camera/test_image_client_camera.py

G1 Datasets Replay:

# --repo-id     Your unique repo ID on Hugging Face Hub 
# --robot_type     The type of the robot e.g., z1_dual_dex1_realsense, z1_realsense, g1_dex1, 

python test/test_replay.py --repo-id unitreerobotics/G1_CameraPackaging_NewDataset --robot_type g1_dex1

2.2 🦿 Run Z1

2.2.1 🦿 Z1 Setup

Clone and build the required repositories:

Download z1_controller and z1_sdk.

Build the repositories:

mkdir build && cd build
cmake .. && make -j

Copy the unitree_arm_interface library: [Modify according to your own path]

cp z1_sdk/lib/unitree_arm_interface.cpython-310-x86_64-linux-gnu.so ./unitree_deploy/robot_devices/arm

Start the Z1 controller [Modify according to your own path]:
```
cd z1_controller/build && ./z1_ctrl
```

2.2.2 Testing ✅

Run the following tests:

Realsense Camera Test:

python test/camera/test_realsense_camera.py # Modify the corresponding serial number according to your realsense

Z1 Arm Test:
```
python test/arm/z1/test_z1_arm.py
```
Z1 Environment Test:
```
python test/arm/z1/test_z1_env.py
```

Z1 Datasets Replay:

# --repo-id     Your unique repo ID on Hugging Face Hub 
# --robot_type     The type of the robot e.g., z1_dual_dex1_realsense, z1_realsense, g1_dex1, 

python test/test_replay.py --repo-id unitreerobotics/Z1_StackBox_Dataset --robot_type z1_realsense

2.3 🦿 Run Z1_Dual

2.3.1 🦿 Z1 Setup and Dex1 Setup

Clone and build the required repositories:

Download and compile the corresponding code according to the above z1 steps and Download the gripper program to start locally
Modify the multi-machine control according to the document
Download the modified z1_sdk_1 and then compile it, Copy the unitree_arm_interface library: [Modify according to your own path]
```
cp z1_sdk/lib/unitree_arm_interface.cpython-310-x86_64-linux-gnu.so ./unitree_deploy/robot_devices/arm
```

Start the Z1 controller [Modify according to your own path]:

cd z1_controller/builb && ./z1_ctrl
cd z1_controller_1/builb && ./z1_ctrl

Start the gripper service, ifconfig examines its own dds networkInterface:
```
sudo ./dex1_1_gripper_server --network eth0 -l -r
```

2.3.2 Testing ✅

Run the following tests:

Z1_Dual Arm Test:
```
python test/arm/z1/test_z1_arm_dual.py
```

Z1_Dual Datasets Replay:

# --repo-id     Your unique repo ID on Hugging Face Hub 
# --robot_type     The type of the robot e.g., z1_dual_dex1_realsense, z1_realsense, g1_dex1, 

python test/test_replay.py --repo-id unitreerobotics/Z1_Dual_Dex1_StackBox_Dataset_V2 --robot_type z1_dual_dex1_realsense

3.🧠 Inference and Deploy

Modify the corresponding parameters according to your configuration
Go back the step-2 of Client Setup under the Inference and Deployment under Decision-Making Mode.

4.🏗️ Code structure

If you want to add your own robot equipment, you can build it according to this document

5. 🤔 Troubleshooting

For assistance, contact the project maintainer or refer to the respective GitHub repository documentation. 📖

6. 🙏 Acknowledgement

This code builds upon following open-source code-bases. Please visit the URLs to see the respective LICENSES (If you find these projects valuable, it would be greatly appreciated if you could give them a star rating.):

6.3 KiB Raw Blame History