Package Description

This package is used to control the Unitree A1 robot with a pre-trained AI model. The training is done by this repository. The main purpose is to provide an interface to use pre-trained AI models with unitree robots using unitree_ros package.

Installation

1. Clone the repository:

git clone https://github.com/Jellal-17/unitree_model_control.git

2. Build the Docker image:

docker build -t model_control .

3. Run the docker container:

docker run -it \
    --env="DISPLAY" \
    --env="QT_X11_NO_MITSHM=1" \
    --volume="/tmp/.X11-unix:/tmp/.X11-unix:rw" \
    --privileged \
    --gpus all \
    model_control 

• Change the name of the container (model_control) as you wish.
• Modify the command if you do not want to use all GPUs by replacing --gpus all with a specific GPU index (e.g., --gpus '"device=0"').

Usage

1. Start the Gazebo Simulation

Run the following commands inside the Docker container:

./start_docker.sh
source devel/setup.bash
roslaunch unitree_gazebo normal.launch rname:=a1

2. Start the AI Model

Open a new terminal and access the Docker container:

./acess_docker.sh
source devel/setup.bash
rosrun unitree_model_control model_run

3. Control the Robot

Open another terminal and access the Docker container to start the teleop_keyboard_twist node, which allows you to command the robot to walk:

./acess_docker.sh
rosrun teleop_twist_keyboard teleop_twist_keyboard.py