KM Robotics development environment

Common images and other tools to standardize development environments used in KM Robotics product development.

Content

Tools and resources installed in the environment:

ROS "desktop full"
Nvidia tools derived from cudagl Nvidia Docker images. You may need to install nvidia-docker2 on the host, cf. https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#install-guide.
Webots simulator in /opt/webots

Docker images

Docker images are placed in GitHub Container Registry and associated with KM Robotics organization.

Those common environment images and the repository itself are PUBLIC.

In the future there should be images for each used ROS version, right now there is support for ROS Kinetic and ROS Noetic.

Get the image

docker pull ghcr.io/km-robotics/env-kinetic:edge

Always do this before other work, to get newest version of the container image. You can also add --pull always argument to docker run commands to enforce updating of the Docker image before running.

Daily work

Suppose that you have your Catkin workspace in ~/robocon_ws.

To run single command inside the environment with mapped Catkin workspace and under the same user as in the host system:

docker run -ti --rm --net=host --device=/dev/dri --env="DISPLAY" -v $HOME/.Xauthority:/$HOME/.Xauthority:rw -v /etc/passwd:/etc/passwd:ro -v /etc/shadow:/etc/shadow:ro --user $UID -v $HOME/robocon_ws:$HOME/robocon_ws -e APP_WS=$HOME/robocon_ws ghcr.io/km-robotics/env-kinetic:edge CMD

To start an instance capable of running multiple commands:

docker run -d -ti --name=kmr1 --net=host --device=/dev/dri --env="DISPLAY" -v $HOME/.Xauthority:/$HOME/.Xauthority:rw -v /etc/passwd:/etc/passwd:ro -v /etc/shadow:/etc/shadow:ro --user $UID -v $HOME/robocon_ws:$HOME/robocon_ws -e APP_WS=$HOME/robocon_ws ghcr.io/km-robotics/env-kinetic:edge bash

# in first terminal
docker exec -ti kmr1 bash -l
  source /rep.sh
  roscore

# in second terminal
docker exec -ti kmr1 bash -l
  source /rep.sh
  rviz

docker rm -f kmr1

To start an environment capable of running multiple commands in several terminals multiplexed by Byobu (command not specified, because byobu -l -2 is a default command for this image):

docker run -ti --rm --net=host --device=/dev/dri --env="DISPLAY" -v $HOME/.Xauthority:/$HOME/.Xauthority:rw -v /etc/passwd:/etc/passwd:ro -v /etc/shadow:/etc/shadow:ro --user $UID -v $HOME/robocon_ws:$HOME/robocon_ws -e APP_WS=$HOME/robocon_ws ghcr.io/km-robotics/env-kinetic:edge

# open new terminal with F2 or Ctrl+A,C, switch terminals back and forth with F3 or Ctrl+A,P and F4 or Ctrl+A,N, close terminal with Ctrl+A,K; use Esc,number instead of Fnumber in applications such as Midnight Commander
roscore # in 1st terminal
rviz # in 2nd terminal

# F6 to exit from the environment

Of course you can set an alias for those commands by using alias command. Persist those aliases in your .bashrc file. You can even create Bash functions to create "aliases" with arguments.

For ROS Noetic just change env-kinetic to env-noetic. For ROS Kinetic image there is no need to map /etc/shadow into the container. ROS Noetic image requires it.

Webots

Webots simulator is installed inside the container in /opt/webots. You can run it from inside the container either directly (/opt/webots/webots) or by using ROS package ros-$ROS_DISTRO-webots-ros.

Other useful tools

Singularity containers

For easier running of GUI tools and seamless binding with local (host) environment you can also use Singularity container system.

You can create Singularity image from the Docker image:

singularity build kmr-env-kinetic.simg docker://ghcr.io/km-robotics/env-kinetic:edge

Then you can launch something using this image, with your home directory automatically mapped into the container:

singularity shell kmr-env-kinetic.simg
  source ...setup.bash
  roscore &
  rviz
  /opt/webots/webots

Singularity maps user IDs and automatically maps current user's home directory into the container. The integration is more seamless that way.

⚠️ Problem with Singularity (at least version 2.6.1) is that when you launch something inside the container's shell and then you exit the shell, the thing remains running. That's different from Docker behavior.

galou / kmr-env