teleoperation

🚀 Environment Setup

1. Clone the official repository

    git clone https://gitee.com/FourierIntelligence/teleoperation.git

2. Create a virtual environment and install the required packages

    conda create -n teleop python==3.11
    conda activate teleop
    pip install -e '.[fourier,realsense]'

4. (Optional) Install ZED SDK

   The ZED setup composes with two parts:

   • Install the ZED SDK: ZED SDK coould be installed from the official website. Please select the version that matches your operating system.

   • Install the ZED Python API

     The ZED python API could be installed with following command:

     # Activate the your virtual environment
     conda activate teleop
     
     # install Python (x64 version) and the pip package manager. Then install the dependencies via pip in a terminal.
     python -m pip install cython numpy opencv-python pyopengl
     
     # Install ZED python API
     cd /usr/local/zed/ 
     python get_python_api.py
     

5. Setup fourier-grx

The fourier GR series robots are controlled by the fourier-grx package. The fourier-grx package is only available for Python 3.11. Thus, we suggest you to create a new virtual environment with Python 3.11 and install the package in the new environment. For more information, please refer to the official Fourier GRX Documentation

    conda create -n grx python==3.11
    conda activate grx
    pip install grx-grx==1.0.0a18
    cd ./server_config
    grx run ./gr1t2.yaml --namespace gr/daq

Then in another terminal, you can run the following command to do the initial calibration, make sure the robot is in the initial position.

    conda activate grx
    grx calibrate

After the calibration, there should be a sensor_offset.json file in the ./server_config directory.

👓 Setup VisionPro

The VisionPro setup is the same as the original OpenTeleVision.

🌐 Local Machine Connection

Apple restricts WebXR access on non-HTTPS connections. To test the application locally, you need to set up a self-signed certificate and install it on the client device. Here's what you'll need:

1. An Ubuntu machine.
2. A router.
3. VisionPro connected to the same network with the Ubuntu machine.

🔐 Self-Signed Certificate

We'll be using mkcert to create a self-signed certificate. and mkcert is a simple tool for making locally-trusted development certificates. It requires no configuration. Here's how to set it up:

1. Please follow the instructions on the official website to install mkcert.

2. check the internet IP information with

    ifconfig | grep inet

3. Creating the certificate with mkcert, make sure to put the IP address of your computer in the command

    mkcert -install && mkcert -cert-file cert.pem -key-file key.pem {Your IP address} localhost 127.0.0.1

example usage:

    mkcert -install && mkcert -cert-file cert.pem -key-file key.pem 192.168.1.100 your-computer.local localhost 127.0.0.1

4. Turn on firewall setup

    sudo iptables -A INPUT -p tcp --dport 8012 -j ACCEPT
    sudo iptables-save
    sudo iptables -L

or setup firewall with ufw

    sudo ufw allow 8012

5. install ca-certificates on VisionPro

    mkcert -CAROOT

Copy the rootCA.pem file to the VisionPro device through the Airdrop.

Settings > General > About > Certificate Trust Settings. Under "Enable full trust for root certificates", turn on trust for the certificate.

Settings > Apps > Safari > Advanced > Feature Flags > Enable WebXR Related Features

6. open the browser on Safari on VisionPro and go to https://192.168.1.100:8012?ws=wss://192.168.1.100:8012

You will see the message "Your connection is not secure" because we are using a self-signed certificate. Click "Advanced" and then "proceed to website". You will be prompted to accept the certificate.

Also, since the python script is not running, the browser will show a message "Safari cannot open the page because the server could not be found.". This is expected behavior. Refresh the page after running the python script and you will see the VR session.

7. Run the python script on the Ubuntu machine. Please see the Usage section for more details.

🕹️ Usage

Start up the GRX server

    cd ./server_config
    grx run ./gr1t2.yaml --namespace gr/daq

Run the teleoperation script

We manage the config with Hydra. You can select config files and override with hydra's override syntax.

    python -m teleoperation.main --config-name teleop_gr1 use_waist=false use_head=false camera=realsense

To record data:

    python -m teleoperation.main --config-name daq_gr1 recording.task_name=${task_name}

Start the teleoperation

After running the python command, you can open the browser on the VisionPro device and go to https://your-hostname.local:8012?ws=wss://your-hostname.local:8012. Or if you already in this website, you can refresh the page and click until see the camera image in the VR session.

Finallly, Click the Enter VR button and give necessary permissions to start the VR session. Make sure to reset the Vision Pro tracking by long press the crown button on the Vision Pro device until you hear a sound.

After starting the script, the robot will move to its start position. The operator should try to put their hands in the same start position (elbows 90 degree, hands open), and then hit the Space key to start the teleoperation.🦾

Afterwards, the operator can start the teleoperation by moving their hands in the VR session. The robot will mimic the operator's hand movements in real-time. To stop the teleoperation, the operator can hit the Space key again.

🛠️ Development

We manage the development environment with the pdm package manager. Thus, please make sure to install pdm first following the official guide here.

    pdm install -d -Gfourier -Gdepthai -Grealsense -v

To select the specific environment, you can run the following command:

    pdm use

And to activate the environment, you can run the following command:

    eval "$(pdm venv activate)"

You can run the following command to start the development environment:

    pdm run python -m teleoperation.main tests

🙏 Credits

This project is based on the amazing OpenTeleVision project. We would like to thank the original authors for their contributions.

📖 Citation

If you find this project useful, please consider citing it:

@misc{teleoperation,
  author = {Yuxiang Gao, Fourier Co Ltd},
  title = {teleoperation},
  year = {2024},
  publisher = {GitHub},
  journal = {GitHub repository},
  howpublished = {\url{https://github.com/FFTAI/teleoperation}}
}