This package enables a basic MuJoCo simulation environment with ROS2 communication channels for controlling a robot. It publishes a sensor_msgs/msg/JointState topic containing the position, velocity, and effort of the joints, and accepts commands via a std_msgs::Float64MultiArray topic. It can run in POSITION, VELOCITY, or TORQUE mode which can be set as a parameter in a launch file.
Figure 1: Your custom control node can interact with the mujoco node through two topics.
Tip
There are many robot models available in the Google DeepMind MuJoCo Menagerie.
Figure 2: The Kuka iiwa14 model from the MuJoCo Menagerie.
- Ubuntu 22.04 or later,
- ROS2 (Humble Hawksbill or later)
- MuJoCo 3.2.0 or later, and
- GLFW for visualisation.
Note
This package was built and tested with Ubuntu 22.04, using ROS2 Humble Hawksbill, and MuJoCo 3.2.0.
Ensure that you have ROS2 and MuJoCo installed on your system.
-
Install ROS2:
Follow the ROS2 installation guide for your operating system.
-
Install MuJoCo:
Download and install MuJoCo from the official website.
Mine is installed in
/opt/mujoco/mujoco-3.2.0Next, modify your
.bashrcfile:gedit ~/.bashrcAdd the following lines of code at the bottom:
export MUJOCO_PY_MUJOCO_PATH=/opt/mujoco/mujoco-3.2.0 export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/opt/mujoco/mujoco-3.2.0/bin export LD_LIBRARY_PATH=/opt/mujoco/mujoco-3.2.0/lib:$LD_LIBRARY_PATHChange all instances of
opt/mujoco/mujoco-3.2.0to match your installation location, and your version. -
Install GLFW:
sudo apt-get install libglfw3-dev
In the src/ directory of your ROS2 workspace, clone the repository:
git clone https://github.com/Woolfrey/mujoco_ros2
Your directory structure should look something like this:
ros2_workspace/
├── build/
├── install/
├── log/
└── src/
└── mujoco_ros2/
├── include/
├── launch/
├── model/
├── src/
├── CMakeLists.txt
├── LICENSE
├── package.xml
└── README.md
In CMakeLists.txt you need to tell the compiler where to find the MuJoCo header files:
set(MUJOCO_PATH "/opt/mujoco/mujoco-3.2.0") # UPDATE THIS TO YOUR MUJOCO PATH
include_directories(${MUJOCO_PATH}/include/mujoco) # MuJoCo header files
include_directories(${MUJOCO_PATH}/include) # Additional MuJoCo header files
link_directories(${MUJOCO_PATH}/lib) # Location of MuJoCo libraries
As noted above, mine is installed in /opt/mujoco/mujoco-3.2.0 but you should change it to match your directory and version.
Navigate back to the root of your ROS2 workspace and build the package:
colcon build --packages-select mujoco_ros2
Source the ROS2 workspace:
source install/setup.bash
There is a simple launch file that demonstrates how to run it in mujoco_ros2/launch/example.py. It launches the node:
mujoco = Node(
package = "mujoco_ros2",
executable = "mujoco_node",
output = "screen",
arguments = [xmlScenePath],
parameters = [
{"joint_state_topic_name" : "joint_state"},
{"joint_command_topic_name" : "joint_commands"},
{"control_mode" : "VELOCITY"},
{"simulation_frequency" : 1000},
{"visualisation_frequency" : 20},
{"camera_focal_point": [0.0, 0.0, 0.25]},
{"camera_distance": 2.5},
{"camera_azimuth": 135.0},
{"camera_elevation": -20.0},
{"camera_orthographic": True}
]
)
The argument = [xmlScenePath] requires a valid MuJoCo xml file. For this example, it utilizes the local model/scene.xml file. There are lots of other models in the MuJoCo Menagerie that you can use.
The parameter argument can be changed based on your preferences:
| Argument | Purpose |
|---|---|
| joint_state_topic_name | This sets the name of the topic for the joint state publisher. Your ROS2 robot control node should subscribe to this topic as a sensor_msgs/JointState. |
| joint_command_topic_name | This sets the topic name that the node will subcsribe to for receiving joint commands. Your ROS2 robot control node should publish to this as a std_msgs/Float64MultiArray. |
| control_mode | This is a string that sets what type of commands are sent to the joints. Options are POSITION, VELOCITY, or TORQUE. |
| simulation_frequency | This is used to set the update rate of the underlying MuJoCo simulation. |
| visualisation_frequency | This is how fast the 3D rendering should be refreshed. In general it should be much lower than the simulation frequency. |
| camera_focal_point | A position in the 3D world that the camera looks at. |
| camera_distance | How far the camera is from the focal point. |
| camera_azimuth | A clockwise rotation of the camera viewing angle. |
| camera_elevation | The tilt of the angle from the ground plane. |
| cmaera_orthographic | Controls the type of projection. Can be True or False. |
After you've installed the package, and sourced the workspace, you can run:
ros2 launch mujoco_ros2 example.py
And you should see the following:
Figure 3: Using the launch example generates a 2-link planar robot, with ROS2 communication channels.
You can check the topics are available, and see the output for the joint state being published:
ros2 topic list
ros2 topic echo /joint_state
You can alter the parameters to your liking.
Note
So far I've only tested VELOCITY mode robustly. TORQUE mode did work, but my controller hadn't been properly tuned. It will be double-checked in the near future.
I made this package just to test my own control software. More features may be added when I find the time and/or need, but contributions are welcome! Feel free to fork it, modify it, and issue a pull request.
Distributed under the GNU General Public License. See LICENSE for more information.