tutorials vrx_docker_manual_trial

Manually Run a Trial

In some cases, you may need to debug behavior that is only expressed when your solution is interacting with the vrx server.

• The run_trial.bash script provided in the vrx-server repository is the simplest way to run both the competitor image and server together.
• This method is described in our tutorial on examining a running container.
• We now take an alternate approach and step through the process of executing a trial manually.
• This method is more complex, but has the advantage of giving more control over the execution process, and can help locate points of failure.

Assumptions

In this tutorial we assume the following:

• you have already built both competitor and server images
• you have completed all the steps list in the troubleshooting prerequisites tutorial
• you are working on a machine with an nvidia graphics card, as described in the VRX system requirements.

Preparation

• Delete previous instances of vrx containers, if there are any:

  docker rm vrx-competitor-system
  docker rm vrx-server-system
  

• Create the network that the two containers will use to communicate:

  docker network create --subnet 172.16.0.10/16 vrx-network 
  

  Don't worry if this network already exists.
• Set the appropriate values to specify your team, task, and trial:

  TEAM=example_team
  TASK=stationkeeping
  TRIAL=0
  

  (Replace the example values above with the desired values.)
• Specify the name and tag for your vrx-server image:

  SERVER_IMG="vrx-server-humble-nvidia:latest"
  

  Unless you customized the vrx-server build process, the default value shown above is probably correct.
• Change to the root of the vrx-docker repository:

  cd /PATH/TO/vrx-docker
  

  (Replace /PATH/TO with the correct path.)
• Set up variables with paths to directories you need to tell the vrx-server repository about:

  TEAM_GENERATED_DIR=`pwd`/generated/team_generated/${TEAM}
  TASK_GENERATED_DIR=`pwd`/generated/task_generated/${TASK}
  HOST_LOG_DIR=`pwd`/generated/logs/${TEAM}/${TASK}/${TRIAL}
  

Start the server and competitor images

• First, we start the vrx server:

  ./vrx_server/run_container.bash -n vrx-server-system $SERVER_IMG   "--net vrx-network \
    --ip 172.16.0.22 \
    -v ${TEAM_GENERATED_DIR}:/team_generated \
    -v ${TASK_GENERATED_DIR}:/task_generated \
    -v ${HOST_LOG_DIR}:/vrx/logs \
    -e ROS_MASTER_URI=http://172.16.0.22:11311 \
    -e ROS_IP=172.16.0.22 \
    -e VRX_DEBUG=true -it --entrypoint /bin/bash"
  

  • The run_container.bash script will print out its commands as it executes them.
  • When it's finished it should open a shell into the container with a username that matches your local user.
• Now open a second terminal for the competitor container.
• Set the name of your competitor image.

  COMPETITOR_IMG="virtualrobotx/vrx_2023_simple:test"
  

  This is the value you chose when building your image.
• Start the competitor image:

  docker run \
      --net vrx-network \
      --name vrx-competitor-system \
      --env ROS_MASTER_URI=http://172.16.0.22:11311 \
      --env ROS_IP=172.16.0.20 \
      --ip 172.16.0.20 \
      --privileged \
      --runtime=nvidia \
      --entrypoint=/bin/bash \
      -it \
      ${COMPETITOR_IMG}
  

Launch the task:

You will now start the task in the server container and run your solution in the competitor container. In order for this to work, the commands in this section should be run within a few seconds of each other.

• Back in the server terminal, begin the trial by running the VRX entrypoint script:

  TEAM=example_team
  TASK=stationkeeping
  TRIAL=0
  /run_vrx_trial.sh /team_generated/"${TEAM}".urdf /task_generated/worlds/${TASK}${TRIAL}.world /vrx/logs
  

  • Note that you need to set the TEAM, TASK, and TRIAL variables again inside the container.
  • You will see some output with the message Starting vrx trial... and the simulation will begin to run.
• Switch back to the competitor terminal and test your solution by manually executing the entrypoint.

  /ros_entrypoint.sh
  

• Both server and competitor should now run the trial.

Repeat trials as needed

• A major advantage of this approach is that the containers will not exit after the trial is complete.
• This gives you more time to diagnose failures.
• You can also rerun trials simply by re-entering the entrypoint commands given in the section above.

Enable the VRX Server GUI

The Gazebo GUI is disabled by default in the VRX server, but we can enable it by modifying the roslaunch command in the run_vrx_trial.sh script.

• This can be accomplished with the following sed command:

  sudo sed -i 's/gui:=false/gui:=true/g' /run_vrx_trial.sh
  

• This modification will cause Gazebo to open a graphical window while the trial is running so you can see your WAMV's behavior in real time.
• To disable the GUI, simply reverse the argument in the sed command:

  sudo sed -i 's/gui:=true/gui:=false/g' /run_vrx_trial.sh
  

• Since changes are not persistent, you can always get back to the original configuration by exiting and restarting the container.

Exit and Cleanup

• To exit containers when finished, log out of the bash session with the exit command.
• If desired, you can clean up the containers and network by running:

  docker rm vrx-competitor-system
  docker rm vrx-server-system
  docker network rm vrx-network
  

Back: Run a Container Manually	Up: VRX Docker Image Overview