Running ROS and Gazebo on an Ubuntu Desktop via Docker

This Dockerfile will install ROS Melodic with Gazebo 9 on Ubuntu 18.04, and give you a VNC interface to work within that environment. The container is tested and working on Windows 10 and Mac OS X. If you need ROS Kinetic with Turtlebot simulations, be sure to check out the kinetic branch.

To install:

git clone https://github.com/jbnunn/ROSGazeboDesktop
cd ROSGazeboDesktop
docker build -t ros-gazebo-desktop .

Start and Connect to The Container

Work you do in the container's ~/data directory will be saved to your local data directory. On your host, create a data directory:

mkdir data

Now when launching the container, we'll use the -v flag to mount data inside the container at /home/ubuntu/data.

Windows

docker run -it --rm --name=ros_gazebo_desktop -m=4g -p 6080:80 -p 5900:5900 -v %cd%/data:/home/ubuntu/data -e RESOLUTION=1920x1080 -e USER=ubuntu -e PASSWORD=ubuntu ros-gazebo-desktop 

OS X / Linux:

docker run -it --rm --name=ros_gazebo_desktop -m=4g -p 6080:80 -p 5900:5900 -v $PWD/data:/home/ubuntu/data -e RESOLUTION=1920x1080 -e USER=ubuntu -e PASSWORD=ubuntu ros-gazebo-desktop   

We also expose port 5900 so you can connect with a VNC client, and/or port 6080 so you can connect via your browser using NoVNC. You can change the RESOLUTION to match screen dimensions appropriate for your display, and the USER environment variable to change the user that logs into the desktop. In most cases, you'll want to leave this as the ubuntu user (password ubuntu).

Once the docker run command finishes running, connect to the container using a VNC client or via http://locahost:6080/. From the Ubuntu desktop, open a terminal, and try:

roslaunch gazebo_ros empty_world.launch 

This will launch the Gazebo virtual world simulator with a default, empty world. Let's spawn the "Create" robot from iRobot, which is based off the Roomba platform. Open a new terminal, then:

rosrun gazebo_ros spawn_model -file ~/.gazebo/models/create/model-1_4.sdf -sdf -model Create

You should see Spawn status: SpawnModel: Successfully spawned entity. Switch views back to Gazebo and you'll find the Create robot in your virtual world. It should look similar to the following.

There are other worlds you can launch:

roslaunch gazebo_ros willowgarage_world.launch
roslaunch gazebo_ros mud_world.launch
roslaunch gazebo_ros shapes_world.launch
roslaunch gazebo_ros rubble_world.launch

Take a look at the syntax at http://gazebosim.org/tutorials?tut=ros_roslaunch for details on other launch options.

Make the robot move

In a new terminal, list the currently available ROS topics:

$ rostopic list

/clock
/gazebo/link_states
/gazebo/model_states
/gazebo/parameter_descriptions
/gazebo/parameter_updates
/gazebo/set_link_state
/gazebo/set_model_state
/rosout

We can manipulate the Create robot by sending a message to model_state. Publish a message to that topic:

rostopic pub /gazebo/set_model_state gazebo_msgs/ModelState '{model_name: Create, pose: { position: { x: 0, y: 0, z: 0 }, orientation: {x: 0, y: 0, z: 0} }, twist: { linear: { x: 1, y: 0, z: 0 }, angular: { x: 0, y: 0, z: 0}  }, reference_frame: world }'

This command:

1. Publishes a topic to the set_model_state topic of type gazebo_msgs/ModelState
2. Contains the model name we specifed when we spawned the robot, Create
3. Specifies an initial "pose" x, y, z position of (0, 0, 0), and an initial orientation of (0, 0, 0)
4. Specifies a "twist" component that tells it to move along the x axis
5. Tells the topic that all of this happens in our world reference frame

Other Robot Models and Considerations

• Many OSRF robot and models can be downloaded to your Docker container, and can be found at https://bitbucket.org/osrf/gazebo_models/src/default/.

     hg clone https://bitbucket.org/osrf/gazebo_models ~/.gazebo/models
  

• If you continue your experiments with the Create model, look at https://gist.github.com/eddiem3/4f257b769d53c492b7ea0dc482cd7caa or http://guitarpenguin.is-programmer.com/posts/58100.html for info on adding a differential-drive plugin.

Wrap

You now have a working robotic simulation environment. Visual Studio Code is available to you under the "Programming" menu item if you need a nice IDE. Don't forget that your work will not be saved unless you are working from the /home/ubuntu/data folder (or other folder(s) you mount with the -v flag.)

Troubleshooting

Gazebo is slow or processes hang

This is likely a memory issue. Launch your container with more RAM, 4GB for example, with the -m=4g flag. Your default Docker installation might be set to something lower, see here to change those settings.

Credits

Based on the following work:

• Docker file for ROS + Gazebo setup: https://github.com/bpinaya/robond-docker
• LXDE Desktop: https://hub.docker.com/r/dorowu/ubuntu-desktop-lxde-vnc/