ubiquity_launches Launch File Repository

Development Overview

All of the Ubiquity robots use ROS (Robot Operating System). ROS breaks a robot application into a multitude of ROS nodes (i.e. processes) running on one or more processors that communicate with one another via internet protocols.

The usual method for teaching people to develop ROS robot applications is to point them at some ROS tutorials and say “go at it”. Unfortunately, this is a little bit dumping buckets of blood into shark infested waters and going for a little swim. At Ubiquity Robotics, we want robot application developers to have a much less traumatic experience. For this reason, we have invested a great deal of effort to make ROS robot application development as easy as possible.

The reality is that setting up a workable software development environment for developing ROS applications is actually pretty involved. One major complication is that the robot typically does not have a display, keyboard, and mouse. Even if it did, it not particularly easy to walk around with the robot as it moves. The more workable solution is to develop software on a stationary platform like a laptop or a desktop, and communicate with the robot via a wireless internet connection. A further complication, is Ubiquity Robotics is using a Raspberry Pi 2 processor with the ARM7 instruction set, whereas most laptops and desktops use the x86 instruction set. We have to make sure that each machine gets the right instruction set.

The Ubiquity Robotics application development environment assumes that there are two processors. The robot processor is attached to the robot and the development processor is associated with either laptop or desktop computer.

Ubiquity Robotics currently uses a Raspberry Pi Foundation Raspberry Pi 2 Model B for the robot processor. Since “Raspberry Pi 2 Model B” is a mouthful of words, we shorten it down to a more manageable “RasPi2”. The RasPi2 is a quad-core ARM Cortex-A7 CPU with 1GB of RAM, with 4 USB Ports, an Ethernet port, a camera interface, and a micro-SD card. We plug a dual-band USB WiFi Dongle into one of the USB port slots to provide wireless connectivity with the development processor through a commercial off-the-shelf WiFi router.

The development processor must be a 64-bit x86 hardware architecture processor. You can either run some flavor of Ubuntu natively on the processor, or you can run a different operating system (e.g. Windows, MacOS, Solaris, etc.) and run VirtualBox to run the Ubuntu operating system. Neither 32-bit processors nor other virtual machines (e.g. VMWare, Parallels, etc.) are supported.

ROS currently only runs under Ubuntu and its variants (e.g. Kubuntu, Xubuntu, Lubuntu, etc.) No other operating system options are currently supported by the ROS community. To be consistent, the VirtualBox image that we recommend that you use has Lubuntu installed on it. We also install Lubuntu on the robot processor, just in case you choose to plug a monitor/keyboard/mouse into the RasPi2. Lubuntu is a very light weight window system that does not over tax the RasPi2 robot processor.

We really encourage people not to cut corners when it comes to setting up your WiFi network. If you have heard the saying that “a chain is only as strong as its weakest link”, our experience is that WiFi is the weak link in robotics. We strongly recommend that people use dual band 802.11ac USB WiFi dongles to provide wireless technology. Similarly, we recommend that you invest in a superior dual-band WiFi router. You can spend an enormous amount of time tracking down and eradicating flaky WiFi issues and it is better to simply avoid the issues by using better hardware and better antennas.

We use a technology called zeroconf to provide human readable names for the various processors on the computer network. Each robot processor and development processor must have a unique host name and MUST run zeroconf. The DNS will present the DNS names as hostname.local where, hostname is the host name of the robot or development processor.

The ROS community extensively uses Secure Shell to communicate between computers. The secure shell protocols provide a secure and encrypted link between computers. We require that password keys be properly set up between the development processor and the robot processor such that a secure link can be established without prompting the user for a password.

Once we have secure shell properly installed, we can do three things with it:

• The roslaunch program can launch ROS nodes (i.e. processes) via secure shell.

• We use the secure shell to do X11 protocol forwarding. X11 a fairly old (but reliable) technology for doing remote window access. Using X11 protocol forwarding, we can debug a program running on the robot processor with windows that show up on the development processor.

• We can use secure shell to remotely mount a chunk of the file system from the robot processor onto the development processor. This allows you to use editors and tools on development system to develop robot code. There are requirements on how all the files are organized on both machines, but when all the rules are followed, you can easily build the same ROS packages on both your development processor and robot processor without any major issues.

The secure shell technology is very useful, but requires some effort to set up properly.

Lastly, we have an entire architecture for structuring ROS launch files. A ROS Launch file is a sequence of one or more files in XML format, that when fed into the roslaunch program cause all of the ROS nodes needed to run a robot to be started. The goal of the architecture is:

• Sharable: The launch files are shared among one another.

• Platform Neutral: The launch files can be used for more than on robot platform.

• Complete: Only one top level robot launch file is specified to start your robot and any ancillary visualization tools need for the your robot application.

{Summary here}

Development Environment Details

This is where we go through the steps required to set up the Ubiquity ROS development environment. This is currently broken into a bunch of steps. Eventually, most of these steps will be more automated.

Initial setup:

We are going to assume that you are doing all of these instructions from the comfort of your own development machine. We further assume that you can connect to the ubuntu user on your robot. We further assume that you can already access the robot via secure shell.

1. Install some software on your development machine:

    sudo apt-get install -y ssh fuse sshfs
    sudo apt-get install -y 
    sudo gpasswd -a $USER fuse
   

2. Let’s make sure that you can log into your robot via ssh. Please replace ROBOT with your robot host name:

    ssh ubuntu@ROBOT.local echo hello
    # You will probably be prompted for the `ubuntu` account
    # password (usually `ubuntu`).  When done it will print
    # out `hello`.
   

3. Create your new USER account on the development machine. USER can be up to 32 characters of upper/lower case letters, digits, and underscores:

    sudo adduser USER
    # It will prompt you for account password.
    # It will prompt for a bunch of account information.
   

4. Now log into your new USER account:

    su USER
    # It will prompt you for your new password.
   

5. Now let’s create the same USER account on the ROBOT machine.

    ssh ubuntu@ROBOT.local
    # You will probably be prompted for the `ubuntu` password.
    # The prompt will probably change to `ubuntu@ROBOT`.
    sudo adduser USER
    # As before, it will for the `ubuntu` password.
    # Please fill in the rest of the account information
    sudo addgroup USER video
    exit
    # This should return you to the development processor.
    # Now verify that you can log in as USER on the
    # robot processor:
    ssh USER@ROBOT.local echo yippee
    # You should be prompted for the USER password.
    # If everything worked, you should `yippee` printed out.
   

   That should get the USER account built on the robot processor with access to the video group.

Password Free Secure Shell

Setting up Secure Shell to operate without prompting you for passwords all the time is a requirement.

1. Generate a public/private key on the development machine:

    if [ ! -f ~/.ssh/id_rsa ] ; then ssh-keygen -t ras ; fi
   

   If you already have an RSA public/private key, this line will return with no further prompting and you need not continue. If you do get a prompt, it will probably look like:

    Enter file in which to save the key (/home/USER/.ssh/id_rsa):
   

   Just depress the [ENTER] key. Next it will prompt for a pass phrase:

    Enter passphrase (empty for no passphrase):
   

   Again, depress the [Enter] key. It will prompt once again:

    Enter same passphrase again:
   

   Again, depress the [Enter] key. Finally, it will crunch on some numbers for a while, print out some more stuff and finish.

2. Now we want to make sure that the development machine can use secure shell to connect to itself without a password prompt:

    chmod 644 ~/.ssh/authorizied_keys
    cat ~/.ssh/id_rsa.pub >> ~/.ssh/authorized_keys
    chmod 444 ~/.ssh/authorizied_keys
   

   Now to verify that we succeeded, try the following (notice that accent characters (`) are being used instead of single quotes (') :

    ssh `hostname`.local echo terrific
   

   If everything worked, you should see terrific with no password prompt.

3. Now we get to do the same thing over on the robot using the exact same technology:

    sudo USER@ROBOT.local
    # Prompt for USER password on ROBOT.
    if [ ! -f ~/.ssh/id_rsa ] ; then ssh-keygen -t ras ; fi
   

   Just depress the [Enter] key three times.

4. Now we want to make sure that the robot can connect to itself via secure shell without any password prompts:

    chmod 644 ~/.ssh/authorized_keys
    cat ~/.ssh/id_ras.pub >> ~/.ssh/authorized_keys
    chmod 444 ~/.ssh/authorized_keys
   

   Now we verify that we succeeded by (again accent characters, not single quote characters):

    ssh `hostname`.local echo fantastic
   

   There should be no prompt and fantastic should be printed out.

    exit
   

   This should return you to the development processor.

5. Now we need to make it possible to log into the robot without any prompt.

    ssh ROBOT.local chmod 644 .ssh/authorized_keys
    cat ~/.ssh/id_rsa.pub | ssh ROBOT.local 'cat >> .ssh/authorized_keys'
    ssh ROBOT.local chmod 444 .ssh/authorized_keys
   

   Now we verify that we succeeded:

    ssh ROBOT.local echo awesome
   

   There should be no password prompt, and awesome should be printed out.

6. Since there is no particular need for the robot processor to log into the development machine, we do not do the symmetric step for password free secure shell from the robot processor to the development processor.

That covers all of the password prompt removal configuration of secure shell.

X11 Forwarding

In order to enable X11 forwarding via secure shell, we need to edit the file /etc/ssh/ssh_config on both the development processor and the robot processor.

1. Edit /etc/ssh/ssh_config on the devolopment machine:

    sudo nano /etc/ssh/ssh_config
   

   Find the two lines:

    #   ForwardX11 yes
    #   ForwardX11Trusted yes
   

   and remove the [#] characters to make them look as follows:

    #   ForwardX11 yes
    #   ForwardX11Trusted yes
   

   Write the contents out using [Ctr-O] followed by exiting using [Ctrl-X] keys.

2. Verify that X11 forwarding is working on the development machine. Please notice the -Y option:

    ssh -Y ROBOT.local xterm
   

   An xterm window should pop up on the development processor display. The prompt inside the xterm should say USER@ROBOT:~. With your mouse, click inside xterm window and type:

    hostname
   

   inside xterm window followed by the [Enter] key. The response should be ROBOT. Things are working. Again, in the xterm window type:

    exit
   

   and now the xterm window will go away.

3. Now we do the same edits over on the robot processor. Type the following being sure to specify the -Y option:

    ssh -Y ROBOT.local
    sudo nano /etc/ssh/ssh_config
   

   Remove the # characters in front of ForwardX11 and ForwardX11Trusted. Write everything out with [Ctrl-O] followed by [Ctrl-X]. Please stay logged onto the robot processor:

4. Testing that we got it write is done as follows:

    ssh -y ROBOT.local xterm
   

   will cause an xterm to pop up on the development processor. (Just for your information, the X11 protocol is being forwarded through two secure shell X11 protocol forwarders, since there no display on the robot processor.) Dismiss the xterm window by typing exit in the xterm window. Finally, exit from the robot processor by typing:

    exit
   

That takes care of X11 forwarding issues.

Creating the Catkin Work Spaces

We will create two identical catkin work spaces on both the development processor and the robot processor. We assume that ROS has already been installed.

1. On the development machine:

    mkdir -p ~/catkin_ws/src
   

2. Now we want to populate the work space with 4 ROS packages:

    cd ~/cakin_ws/src
    git clone https://github.com/UbiquityRobotics/ubiquity_launches.git
    git clone https://github.com/UbiquityRobotics/raspicam_node.git
    git clone https://github.com/UbiquityRobotics/userland.git
    git clone https://github.com/UbiquityRobotics/robot_upstart.git
    source ~/.bashrc
   

   {Wayne: Make sure $ROS_PACKAGE_PATH is correct.}

3. Now we initialize the Catkin work space:

    (cd ~/catkin_ws ; cakin_make)
   

4. Now we edit ~/.profile to recognize the catkin workspace:

    nano ~/.profile
   

   Now we add the following code right before the line that says #if running bash:

    # Only `source /opt/ros/indigo/setup.bash` if we have not already done so.
    # We assume that this script does not change very often:
    if [ -d "/opt/ros/indigo/bin" ] ; then
        case ":$PATH:" in
        *:/opt/ros/indigo/bin:*) ;;
        *) source /opt/ros/indigo/setup.bash ;;
        esac
    fi
   
    # Only `source ~/catkin_ws/devel/setup` if it exists:
    if [ -f ~/catkin_ws/devel/setup.sh ] ; then
        source ~/catkin_ws/devel/setup.bash ;
    fi
   
    export ROS_CATKIN_WS=~/catkin_ws
    export ROS_HOSTNAME=`hostname`.local
    export ROS_MASTER_URI=http://`hostname`.local:11311
    export ROSLAUNCH_SSH_UNKNOWN=1
   

5. Now we do exactly the same steps over on the robot:

    ssh ROBOT.local
    mkdir -p ~/catkin_ws/src
    cd ~/cakin_ws/src
    git clone https://github.com/UbiquityRobotics/ubiquity_launches.git
    git clone https://github.com/UbiquityRobotics/raspicam_node.git
    git clone https://github.com/UbiquityRobotics/userland.git
    (cd ~/catkin_ws ; cakin_make)
    nano ~/.profile
   

   Now edit in the exact same code from the previous step into the same place in ~/.profile and exit with [Ctrl-O] followed by [Ctrl-X].

Using sshfs

The sshfs program is used to mount the src directory from the robot processor over on to the development processor

To mount the robot ~/catkin_ws/src on top of the local ~/catkin_ws/src:

    sshfs -o nonempty -o idmap=user ROBOT.local:/home/`whoami`/catkin_ws/src ~/catkin_ws/src 

To unmount:

    fusermount -u ~/catkin_ws/src

Making roscore Run All the Time

{Wayne: make n_forever match n_sleep_forever.}

This is done using the following command:

    ssh ROBOT.local
    rosrun robot_upstart install --user=USER ubiquity_launches/n_sleep_forever/launch/n_sleep_forever.launch
    sudo service ubiquity start

Install the pudb Python Debugger:

Please do the following:

    sudo apt-get install -y python python-pip xterm
    sudo pip install pudb

Run raspicam_view

Please do the following on the development machine:

    rosrun ubiquity_launches raspicam_view

When rqt_image_view pops up please select the /n_raspicam/camera/image/compressed topic. When you are done, please type [Ctrl-C].

Launch Files

ubiquity_launches is a ROS git repository that contains ROS .launch files and other associated ROS robot configuration files (e.g. .yaml, .urdf, etc.) In addition, there are shell scripts that invoke the launch files.

A ROS .launch file is used used to start and configure one or more ROS nodes, which when properly configured will result in the desired robot behavior.

By imposing some structure on the organization of these .launch files, we improve the ability to reproduce robot behaviors among all robot developers who share the use of the ubiquity_launches launch files.

Both the shell scripts and the ROS launch files have some additional documentation structure. There is a program called generate_launches_md.py that scans the ubiquity_launches repository and generates a single document called launches.md. This document provides a single location to find all of documentation for the ubiquity_launches repository. The format of the launches.md document is in markdown format

When launches.md file is served up by a git repository management site (e.g. github, bitbucket, etc.) it is converted into HTML so that it is easier to read in a web browser. Please view launches.md to see how the generated ubiquity_launches documentation is rendered in a web browser.

Download and Install ubiquity_launches Repository

The ubquity_launches repository is cloned into your catkin workspace via the following:

    cd $ROS_CATKIN_WS/src
    git clone https://github.com/UbiquityRobotics/ubiquity_launches.git

It is strongly recommended that the bin directory of the ubiquity_launches repository be placed in you execution path. The preferred way to do this is by adding the following somewhat arcane bash script code to your ~/.bashrc file:

    # Only `source /opt/ros/indigo/setup.bash` if we have not already
    # done so.  We assume that this script does not change very often:
    if [ -d "/opt/ros/indigo/bin" ] ; then
        case ":$PATH:" in
        *:/opt/ros/indigo/bin:*) ;;
        *) source /opt/ros/indigo/setup.bash ;;
        esac
    fi

    # Put `.../ubiquity_launches/bin` in the path if it exists:
    if [ -d $ROS_CATKIN_WS/ubiquity_launches/bin ] ; then
        case ":$PATH:" in
        *:$ROS_CATKIN_WS/ubiquity_launches/bin:*) ;;
        *) export PATH=$PATH:$ROS_CATKIN_WS/ubiquity_launches/bin ;;
        esac
    fi

When you are done editing ~/.bashrc, please run the command:

    source ~/.bashrc

Overall Architecture

The current structure of the ubiquity_launch repository is broken in a bunch of sub directories which have the following naming conventions:

The top level directory structure is as follows:

    repository_name/
        README.md    # This document
        launches.md  # Aggregated 
        bin/         # Executable shell scripts
        n_*/         # Single ROS node launch files
        m_*/         # Multiple ROS node launch files
        rviz_*/      # RViz specific launch files

Where:

• README.md: README.md is this documentaton file.

• launches.md: launches.md is a generated documentation file.

• bin: The bin directory contains a bunch of executables shell scripts that typically fire off a launch file. It is expected that you will place this bin directory in your path.

• n_*: The n_* directories contain the launch files and configuration files needed to launch a single ROS Node.

• m_*: The m_* directories will launch Multiple ROS nodes.

• rviz_*: The rviz_*` directories are used to launch the RViz program configured to view a corresponding robot program. These launch files are typically executed on your laptop/desktop, since most robots do not have a display head.

Other directories and files will be added as needed.

The bin Sub-Directory

The bin sub-directory contains a bunch of executables – mostly shell scripts.

The structure of an executable shell script in the bin sub-directory is:

    #!/usr/bin/env bash

    ##Summary: One line summary of shell script.
    ##Overview:
    ##  Line 1 of multi-line documentation.
    ##  ...
    ##  Line N of multi-line documentation.

    roslaunch ...

The first line specifies that the file is a bash script.

All lines that start with ## are scanned into the documentation in launches.md.

The line:

    ##Summary: One line summary of shell script.

provides a single line summary of what the shell script is supposed to do.

The lines:

    ##Overview:
    ##  Line 1 of multi-line documentation.
    ##  ...
    ##  Line N of multi-line documentation.

provide a fuller description of the what the script file does.

The remaining lines are bash shell script commands. Many (but not all) script files will contain a roslaunch ... command that causes ROS process a ROS .launch file.

ROS Launch File Issues

There are several issues about ROS launch files that need to be discussed:

• The <arg> tag is used heavily needs to be fully understood.

• There are two common launch file suffixes – .launch and .launch.xml.

• Launch file parameterization allows the same launch files to be used for different robot platforms and configurations.

• How to add documentation to the a ROS launch file.

The <arg> Tag

Some documentation for ROS launch files can be found by following the hypertext links below:

• ROS Launch Overview provides an overview of the ROS launch file architecture.

• ROS launch XML file format provides documentation of the XML format used for writing ROS launch files.

• ROS launch <arg> tag is the documentation for the <arg ... > tag.

The <arg> tag is singled out because it is heavily used in the launch files to pass information between launch files. If you do not understand the <arg> tag, you will not understand the ubiquity_launches launch files.

The <arg> tag has three forms:

• <arg name="required" />: This specifies a launch file input name. Think of this as a manditory and required input argument variable for launch file.

• <arg name="optional" default="value" />: This specifies a launch file input name with a default value that will be used if not is specified at “call” time. Think of this as an optional input argument for a launch file.

• <arg name="foo" value="bar" />: This form has two different usages. When at the level immediately inside of a <launch> ... </launch> pair, this form defines a convenience value that can be used to improve overall legibility in a launch file. Think of this as a kind of a macro definition. The second form occurs immediately inside of a <include> ... </include> pair. This form is for passing explicit arguments into another launch file.

Huh? What is going on here? Let’s do some examples! The example below us Python as a metaphor for what is going on. If you do not know some basic Python, well, at least we tried.

Here is a chunk of Python code that defines a routine:

    def n_fiducial_slam(robot_base, fiducial_size=".150"):
      short = "a somewhat long string"

This function is named n_fiducial_slam and has two arguments – robot_base and fiducial_size. robot_base is a required argument which if not present at routine call time will cause a run-time error. fiducial_size is an optional argument that does not need to specified in the routine call, but it can be specified if you want. short is a local variable that can be used to reduce typing. The corresponding launch file syntax is:

    <launch>
      <!-- Required arguments: -->
      <arg name="robot_base" />
        <!--robot_base: Robot base (e.g. "magni", "loki", etc.) to use. -->
      
      <!-- Convenience arguments: -->
      <arg name="short value="a somewhat long string" />
        <!--short_value: A short name for a longer string. -->

      <!-- Optional arguments: -->
      <arg name="fiducial_size" default=".200" />
        <!--fiducial_size: Width of fiducial measured in meters. -->
      ...
      
    </launch>

The <include> ... </include> tag pair is how one launch file accesses another launch file. It is similar to a routine call. In Python, the following line:

    n_fiducial_detect("loki", fiducial_size=".200")

would be written in launch file syntax as:

    ...
    
    <include file=".../n_fiducial_detect.launch">
      <arg name="robot_base" value="loki" />
      <arg name="fiducial_size" value=".200" />
    </include>
    
    ...

Now that you know how to set an argument, the only other issue is how to access it. That is done using substitution arguments. The syntax is:

    $(arg name)

where name is the argument name. Using the “call” example above, the following string:

    "base=$(arg robot_base) size=$(arg fiducial_size) short='$(arg short)'"

would expand to:

    "base=loki size=.200 short='a somewhat long string'"

The substitution syntax can only occur inside of XML attribute strings.

Finally, you pass arguments into launch files from the roslaunch command via the following syntax:

    roslaunch ubiquity_launches n_fiducial_detect robot_base:="magni"

It is the := is detected and conceptually converted into an <arg> tag. The line above would be represented in a launch file as:

    <include file="$(find ubiquity_launches)/.../n_fiducial_detect.launch">
      <arg name="robot_base" value="magni" />
    </include>

Hopefully this explanation of the <arg> tag is a little more informative that the official ROS documentation.

Launch File Suffixes:

There are two ROS launch file suffixes:

• .launch: This launch file will be discovered by the roslaunch command via tab completion.

• .launch.xml: This launch file will not be discovered by the roslaunch command tab completion facility.

The roslaunch command has the following basic structure:

    roslaunch ROS_PACKAGE LAUNCH_NAME.launch

Basically, the roslaunch command searches a given ROS package (i.e. ROS_PACKAGE) for a .launch file (i.e. LAUNCH_NAME.launch.) One neat thing about roslaunch is that it implements [tab completion](https://en.wikipedia.org/wiki/Command-line_completion] whereby it will reduce overall typing by allowing you few characters of the package name and/or launch file name followed by a tab character to cause the roslaunch to fill in as many unambiguous characters as possible. When it comes to finding .launch files, roslaunch recursively visits all of the directories and sub-directories in a ROS package and identifies every file that ends in .launch. It does not matter what package sub-directory the .launch file is in, roslaunch will find it. It is really that simple.

A robot launch file repository will have many launch files. Many of these file are likely to only be used via the <include> tag in some other launch files. These launch files use the .launch.xml suffix, so that when you are using tab completion for roslaunch, they do not up as one of the possible completions. That is all that is going on here. To let roslaunch show a launch file via tab complete, use the .launch suffix; otherwise, use the .launch.xml suffix to keep roslaunch from showing the launch file via tab completion. It is that easy.

Launch File Parameterization

The goal of a robot launch repository is to provide high quality launch files that work across multiple robot platforms and configurations. It would be possible to build monolithic launch files that do not use any <include> directives. The reason for not doing that is because you would have lots of replicated text across multiple launch files. Fixing a problem in one launch file would have to be manually propagated to all the other launch files. This would be a maintenance nightmare.

The solution is to break the launch files into a number of smaller launch files and create the robot configuration via composition as described in the roslaunch Architecture.

To get additional reuse, the launch files need to be parameterized such that the same launch file can be used for multiple robots.

For example, the most common parameter is the robot base name. This is called the robot_base parameter and it expected to be given a robot base name (e.g. loki, magni, botvac, etc.) This argument is used to select between different parameter files (e.g. loki.yaml vs. magni.yaml, or loki.urdf vs. magni.urdf, etc.)

Launch File Documentation Structure

The front portion of a launch file is structured as follows:

    <launch>
      <!--Summary: One line summary of launch file -->
      <!--Overview: Multi-line overview 1
          Multi-line overview 2
          ...
          Multi-line overview N -->

      <!-- Required Arguments -->
      <arg name="input_name1" />
        <!--input_name1: Documentation about `input_name1` -->
      ...
      <arg name="input_nameN" />
        <!--input_nameN: Documentation about `input_nameN` -->

      <!-- Convenience Arguments -->
      <arg name="short_name1" value="short value 1" />
      ...
      <arg name="short_nameN" value="short value N" />

      <!-- Optional Arguments -->
      <arg name="option_name1" default="default value 1" />
        <!--optional_nameN: Documentation about `optional_nameN` -->
      ...
      <arg name="option_nameN" default="default value N" />
        <!--optional_nameN: Documentation about `optional_nameN` -->

      ...

      <!-- Rest of launch file goes here: -->

    </launch>

Each comment has the form:

    <!--NAME: Documentation text -->

where NAME is an alpha-numeric identifier with optional underscores. These comments are found and processed by the generate_launches.py program to extract documentation. The NAME field either specifies a launch file input argument, or the values Summary or Overview. The documentation text is written using markdown formatting rules (e.g. italics, bold, fixed_pitch, etc.)

Launch Sub-Directory Structure

Each launch sub-directory is organized as follows:

    LAUNCH_DIR_NAME/
        launch/   # Usually one `.launch` (or `.launch.xml`) file
        params/   # One or more `.yaml` (or other) parameter files
        rviz/     # One or more `.rviz` configuration files
        urdf/     # One or more `.urdf` configuration files

It is not clear why there is an additional level of sub-directory for each different data type, but the turtlebot launch directories have this structure, so it was decided to copy it. (There may be some subtle interaction with ROS bloom that is not yet fully understood. Alternatively, it could be monkey see, monkey do.)

Using git Branches to Experiment

There is not much to say here. If you want to tweak things to experiment, you can use git to get a copy of the files, create a branch and modify things to your hearts content. If you can make a the case that you configuration works better that what is currently in the ubiquity_launches git repository, please submit a pull request back to the master ubiquity_launches git repository.

Some Examples

Below are three examples:

• A single node launch file (i.e. n_*.)

• A multi-node launch file (i.e. m_*.)

• An executable script file. (i.e. bin/*.)

n_* Example

The launch file for n_robot_state_publisher can be found in .../n_robot_state_publisher/launch/n_robot_state_publisher.launch.xml. (There are uninteresting minor differences.) The file is shown below:

    <launch>
      <!--Summary: Launch the ROS `robot_state_publisher` node. -->
      <!--Overview: The launch file for this directory starts the ROS
          [`robot_state_publisher`](http://wiki.ros.org/robot_state_publisher)
          node.  This launch file selects the `.urdf` file based upon
          the `robot_base` argument.  The URDF files are stored in
          `...n_robot_state_publisher/urdf/{robot_base}.urf`. -->
    
      <!-- Required arguments: -->
      <arg name="robot_base" />
        <!--robot_base: Robot base name (e.g. "magni", "loki", etc." -->
    
      <!-- Convenience arguments: -->
      <arg name="root" value="$(find ubiquity_launches)" />
      <arg name="node"  value="robot_state_publisher" />
    
      <!-- Optional arguments: -->
      <arg name="node_name" default="n_$(arg node)" />
        <!--node_name: The name of the ROS node. -->
    
      <!-- Launch the robot state publisher:  -->
      <param name="robot_description"
       textfile="$(arg root)/n_$(arg node)/urdf/$(arg robot_base).urdf" />
      <node name="$(arg node_name)"
       pkg="$(arg node)" type="state_publisher" />
    </launch>

Note that we have a pretty standard order in which stuff is done:

1. Naturally, the <launch> tag is first.

2. The <!--Summary: ... --> one line summary is next.

3. The <!--Overview: ... --> multi-line summary is next.

4. The <!-- Required arguments: --> are next. In this case, there is one required argument called robot_base. robot_base is one of the more common arguments.

5. The <!-- Convenience arguments: --> are next. In this case, both root and rsp are defined. When root argument is fully expanded, it will be something like /home/name/catkin_ws/src/ubiquity_launches. The root argument is almost always present in an n_* launch file. The node argument is short hand for the package name. In this case, the package name is robot_state_publisher. Using $arg(node) takes 10 characters rather than 21 characters for robot_state_publisher.

6. The <!-- Optional arguments: --> comes next. In this example, node_name is defined. This defaults to n_robot_state_publisher. Note that the default="n_$(arg node)" uses one of the previously defined arguments. It should be mentioned that node_name is pretty much present for every n_* launch file. This lets people create multiple nodes using the same launch file, with each node having a different name.

7. The <param ...> is next. It specifies the parameter name and a fully qualified path to the .urdf file. On the Ubiquity standard Raspberry Pi system image with robot_base set to loki, $(arg root)/n_$(arg node)/urdf/$(arg robot_base).urdf expands to /home/ubuntu/catkin_ws/ubiquity_launches/n_robot_state_publisher/urdf/loki.urdf

8. The <node ...> is next. It actually launches the ROS node. The node name comes from $(arg node_name), which will default to n_robot_state_publisher. The package name comes from $(arg node) which expands to robot_state_publisher.

9. The </launch> comes last.

That should give you a pretty good understanding of how an n_* style launch file tends to be structured.

m_* Example

The launch file for m_joystick_teleop can be found in .../m_joystick_teleop/launch/n_joystick_teleop.launch. (There are some uninteresting minor differences.) The file is shown below:

    <launch>
      <!--Summary: Start joystick remote control nodes. -->
      <!--Overview: The launch file for this directory fires off the
          joystick nodes to support the wireless PS2/XBox game controller
          for  driving the robot around.  This launch file requires a
          `robot_base` argument to specify which robot base is being
          used (e.g `loki`, `magni`, etc.) -->
    
      <!-- Required arguments: -->
      <arg name="robot_base" />
        <!--robot_base: Base being used (e.g. "magni", "loki", etc.) -->
    
      <!-- Convenience arguments: -->
      <arg name="root" value="$(find ubiquity_launches)" />
      <arg name="bsm"  value="$(arg robot_base)_serial_master" />
      <arg name="rsp"  value="robot_state_publisher" />
      <arg name="joy"  value="joy" />
      <arg name="ttj"  value="teleop_twist_joy" />
    
      <!-- Fire off each node: -->
      <include
       file="$(arg root)/n_$(arg bsm)/launch/n_$(arg bsm).launch.xml" />
    
      <include
       file="$(arg root)/n_$(arg rsp)/launch/n_$(arg rsp).launch.xml">
        <arg name="robot_base" value="$(arg robot_base)" />
      </include>
    
      <include
       file="$(arg root)/n_$(arg joy)/launch/n_$(arg joy).launch.xml" />
    
      <include
       file="$(arg root)/n_$(arg ttj)/launch/n_$(arg ttj).launch.xml">
        <arg name="robot_base" value="$(arg robot_base)" />
      </include>
    </launch>

Note that we have a pretty standard order in which stuff is done:

1. The launch, <!--Summary: ...>, <--Overview: ...>, and <!-- Required arguments -->, <!-- Convenience arguments --> are basically the same as the n_* example. There are a few more convenience arguments. Lastly, there are no <!-- Optional arguments: ->

2. Next, we start each node. Each node that is launched has its own <include ...> tag. In this example, 4 nodes are nodes being launched, so there are for <include ... > tags. Two of the nodes require the robot_base argument, so they look as follows:

      <include
       file="$(arg root)/n_$(arg rsp)/launch/n_$(arg rsp).launch.xml">
        <arg name="robot_base" value="$(arg robot_base)" />
      </include>
   

   Notice how the file attribibute is specified. $(arg root) expands to $(find ubiquity_launces) which further expands to /home/ubuntu/catkin_ws/src/ubiquity_launches. The $(arg rsp) is a convenience argument that expands to robot_state_publisher. When fully expanded, the full path to the desired launch file is found. The <arg ...> specifies robot_base value to be $(arg robot_base) which was passed into this launch file. Thus, robot_base is being passed through. The launch files without required arguments do not need any <arg ...> tags.

bin/* Example

Here is an example:

    #!/usr/bin/env bash
    
    ##Summary: Cause Loki to collect local costmap.
    ##Overview:
    ##  This program is run on the robot and starts up a Loki platform
    ##  that starts up robot that is running the both the PS3/XBox
    ##  joystick nodes and the fiducial detection and slam nodes.
    ##  The file is focused on generating a local cost map for viewing
    ##  using the `loki_rviz_local_costmap` program.
    
    roslaunch ubiquity_launches m_fiducial_slam.launch.xml robot_base:=loki

Some Extra Thoughts

Right now all the files in the .../ubiquity_launches/bin directory have a platform base name embedded in them (e.g. loki_raspicam.) This is kind of bogus. It causes there to be a plethora of programs in the bin directory. In order to solve this problem, we need a way to identify the platform base name separate from the program name.

For the desktop/laptop, the patform name should just be fetch from the ROS parameters table. Something like /ur/base_name would be fine. Thus, loki_view_raspicam would simplify to ur_view_raspicam and would work for the stage, loki, magni and botvac bases. The launch file, would set the base via appropriate

For the robot, we should just load the /ur/base_name parameter from a known file. There a multitude of different places to store the base name and get it set. We can also have a /ur/variant parameter for different variants of the same basic base platform (e.g. botvac50, xv11, etc.) After the dust settles, the user can just run ur_raspicam and it will bring up the RaspiCam on the appropriate Loki, Magni, or BotVac platform.

Changelog for package ubiquity_launches

0.2.1 (2016-02-18)

• Restore motor_node which was deleted
• Contributors: Rohan Agrawal

0.2.0 (2016-02-18)

• Update Cmakelists
• Restored n_move_base which I accidentally removed.
• Added travis build status to README
• use bash || for testing both ssh commands
• Hack in automatic user detection in roslauncher
• Replace all the old loki_camera stuff with raspicam and raspicam_view
• Got raspicam_view to work with new machine tag args
• Got m_raspicam_raw to work with new machine infasrtuture
• Replaced loki_base and magni_base with unified robot_base script
• Update m_move_base to reflect move of cmd_vel_mux
• Add all required things to sim
• Change default machine for n_relay to robot
• Machine tag in n_relay
• Moved machine def to the top of m_sim_base
• Move turtlebot_model to m_sim_base
• Using robot_base instead of stage_ros directly
• Support sim in robot_base
• Use not shutdown loop instead of infinite loop
• launching cmd_vel_mux with include
• Made cmd_vel_mux launcher (uses nodelets)
• Passing machine_name from m_move_base
• Machine name argumentizeation for rsp
• removed unnessary machine definition
• Further adding of machine_name args
• Added viewer args to more stuff
• Argument-zied machine name with reasonable defaults
• Pass robot host/user to amcl and map_server instead of viewer
• n_amcl and n_map_server now use machine_host and user
• Passing viewer/robot on keyboard_nav and move_base_view
• Have m_move_base require all nessacary args
• Removed old refrences to keyboard_drive
• n_rviz run through ssh
• Launch keyboard_naviagte though ssh to viewer
• Merged branch master into master
• Using ssh to localhost to run teleop
• Use generic machine args for loki_base nodes
• Add teleop_twist_keyboard with any machine format
• Merge branch 'master' of https://github.com/UbiquityRobotics/ubiquity_launches
• Banged together bin/yaml2launch.py
• Using move_base_view as the starter
• Luanch -> Launch
• Move viz to move_base_view
• Launching cmd_vel_mux on remote
• Fixed Typo, moved back to whoami
• New line for clarity of nodelet stuff
• More renaming to keyboard_navigate
• Output screen on teleop to make sure that we can see the output
• fix missing end bracket
• Switch robot_name to correct robot_host
• Robot_base launch is not xml
• fixed bug of mb vs rb
• Remo.ved an. exce.ess. period. in. fil.na.me
• Using ubuntu as user instead of whoami
• Change keyboard_drive to launch.xml
• Added keyboard_drive to unified-ily teleop
• Removed buggy space
• Rename folder for keyboard drive as well
• rename keyboard_drive to keyboard_navigate
• We ha`d an acc`ess of ac`cent ma`rks, it has been rectified
• Fixed always on roscore documentation
• Update Cmakelists to not fail on build
• Make sure robot_platform is being sent to loki_base correctly
• Using robot_base to auto switch between loki and magni
• Consolidate loki and magni base into robot base
• Get rid of loki_serial_master
• More tweaking of m_move_base.launch.xml
• More conversion of move_base.
• Starting to de-turtlebot m_move_base.
• Fixed raspicam_view
• Convert robot_base to robot_platform
• More launch file work.
• Fixed some typos
• Added recipe to configure Ubiquity Robotics development environment.
• Merge branch 'master' of https://github.com/UbiquityRobotics/ubiquity_launches
• Remove old ros_launcher.py program
• Added n_sleep_forever to bring up roscore using robot_upstart package.
• More unified launch file development
• More work on unified launch files.
• Renamved platform_probe.py to platform_probe.
• Merge branch 'master' of https://github.com/UbiquityRobotics/ubiquity_launches
• Changed platform to robot_base
• Some more work on unified launch files.
• Updated test.launch and urtest.sh
• Fixed a typo.
• Tweaked up bin/test.launch and bin/urtest.sh to work with one another.
• Merge branch 'master' of https://github.com/UbiquityRobotics/ubiquity_launches
• Added platform_probe.py and ros_launche.pry to ubiquity_launches/bin
• Fixed up robot.py
• Added platform_probe.py and roscore_start.sh to ubiquity_launche/bin
• Contributors: Rohan Agrawal, Wayne C. Gramlich, Wayne Gramlich
• Added travis build status to README
• use bash || for testing both ssh commands
• Hack in automatic user detection in roslauncher
• Replace all the old loki_camera stuff with raspicam and raspicam_view
• Got raspicam_view to work with new machine tag args
• Got m_raspicam_raw to work with new machine infasrtuture
• Replaced loki_base and magni_base with unified robot_base script
• Update m_move_base to reflect move of cmd_vel_mux
• Add all required things to sim
• Change default machine for n_relay to robot
• Machine tag in n_relay
• Moved machine def to the top of m_sim_base
• Move turtlebot_model to m_sim_base
• Using robot_base instead of stage_ros directly
• Support sim in robot_base
• Use not shutdown loop instead of infinite loop
• launching cmd_vel_mux with include
• Made cmd_vel_mux launcher (uses nodelets)
• Passing machine_name from m_move_base
• Machine name argumentizeation for rsp
• removed unnessary machine definition
• Further adding of machine_name args
• Added viewer args to more stuff
• Argument-zied machine name with reasonable defaults
• Pass robot host/user to amcl and map_server instead of viewer
• n_amcl and n_map_server now use machine_host and user
• Passing viewer/robot on keyboard_nav and move_base_view
• Have m_move_base require all nessacary args
• Removed old refrences to keyboard_drive
• n_rviz run through ssh
• Launch keyboard_naviagte though ssh to viewer
• Merged branch master into master
• Using ssh to localhost to run teleop
• Use generic machine args for loki_base nodes
• Add teleop_twist_keyboard with any machine format
• Merge branch 'master' of https://github.com/UbiquityRobotics/ubiquity_launches
• Banged together bin/yaml2launch.py
• Using move_base_view as the starter
• Luanch -> Launch
• Move viz to move_base_view
• Launching cmd_vel_mux on remote
• Fixed Typo, moved back to whoami
• New line for clarity of nodelet stuff
• More renaming to keyboard_navigate
• Output screen on teleop to make sure that we can see the output
• fix missing end bracket
• Switch robot_name to correct robot_host
• Robot_base launch is not xml
• fixed bug of mb vs rb
• Remo.ved an. exce.ess. period. in. fil.na.me
• Using ubuntu as user instead of whoami
• Change keyboard_drive to launch.xml
• Added keyboard_drive to unified-ily teleop
• Removed buggy space
• Rename folder for keyboard drive as well
• rename keyboard_drive to keyboard_navigate
• We ha`d an acc`ess of ac`cent ma`rks, it has been rectified
• Fixed always on roscore documentation
• Update Cmakelists to not fail on build
• Make sure robot_platform is being sent to loki_base correctly
• Using robot_base to auto switch between loki and magni
• Consolidate loki and magni base into robot base
• Get rid of loki_serial_master
• More tweaking of m_move_base.launch.xml
• More conversion of move_base.
• Starting to de-turtlebot m_move_base.
• Fixed raspicam_view
• Convert robot_base to robot_platform
• More launch file work.
• Fixed some typos
• Added recipe to configure Ubiquity Robotics development environment.
• Merge branch 'master' of https://github.com/UbiquityRobotics/ubiquity_launches
• Remove old ros_launcher.py program
• Added n_sleep_forever to bring up roscore using robot_upstart package.
• More unified launch file development
• More work on unified launch files.
• Renamved platform_probe.py to platform_probe.
• Merge branch 'master' of https://github.com/UbiquityRobotics/ubiquity_launches
• Changed platform to robot_base
• Some more work on unified launch files.
• Updated test.launch and urtest.sh
• Fixed a typo.
• Tweaked up bin/test.launch and bin/urtest.sh to work with one another.
• Merge branch 'master' of https://github.com/UbiquityRobotics/ubiquity_launches
• Added platform_probe.py and ros_launche.pry to ubiquity_launches/bin
• Fixed up robot.py
• Added platform_probe.py and roscore_start.sh to ubiquity_launche/bin
• Contributors: Rohan Agrawal, Wayne C. Gramlich, Wayne Gramlich

0.1.1 (2016-01-24)

• Added robot.py and test.launch .
• Merge branch 'master' of github.com:UbiquityRobotics/ubiquity_launches
• modifications to make loki_base work
• Updated README.md
• Updated README.md
• Added install rule for loki_base exec
• Add install rule for loki_base We need to find a better way to manage the install rules
• Added a loki_base exec file
• Added basic loki_base launch file
• Added Travis Button
• Dummy Commit for travis
• Added travis file added travis file to make sure cmake is valid
• Added install rules Added install rules for bin and all the launches to make sure files are avalible in non-devel environments, such as when installing from debs.
• Remove footprint_layer, which has been removed from ROS
• Correct default fidicual size
• Fix bugs with slam
• Fix bugs with camera
• More documentation.
• Updated README.md
• Sweep through and updated program comments, fix typos, etc.
• Added multiple base testing.
• Search for unused files.
• Fixed launch files for Magni base.
• Added some magni bring up launch files. Not debugged though.
• Wrapped up initial file scraper.
• Wrote generate_launches_md.py file to scrape all the launch directories for documenation.
• Reworked the raspicam launch files to generate an uncompress image stream.
• Added more arguments.
• Update loki.yaml.
• Increased maximum speed for Loki to .25M/sec.
• Added raspicam stuff.
• Merge branch 'master' of https://github.com/UbiquityRobotics/ubiquity_launches
• Add loki_raspicam and loki_raspicam90
• Fixed broken link again.
• Renamed ubiquity-misc to ubiquity_main.
• Added documentation for loki_camera and loki_view_camera.
• Merge branch 'master' of https://github.com/UbiquityRobotics/ubiquity_launches
• Added loki_camera to bin.
• Merge branch 'master' of https://github.com/UbiquityRobotics/ubiquity_launches
• Added missing file n_map_server.launch.xml
• Fixed n_map_server
• Fixed an incorrect path to global map.
• Updated README.md. Remove n_ros_arduino_bridge and rviz_loki_sonar directories.
• Finished cleaing up loki_robot except for documentation.
• Fixed up the loki_rvix executables.
• Removed loki_rviz_local_costmap
• Converted to rvzi_local_costmap.
• Switched over to new robot_base argument in .launch files.
• More work on rviz .launch files.
• Started cleaning out old stuff. Added loki_rviz_local_costmap.
• Create loki_local_costmap .
• Continued work on local costmap launch file.
• More clean up of local cost map.
• More reworking of the local cost map .launch files.
• Switched over to m_joystick_teleop .launch file.
• Started using more <arg ...>'s.
• Converted remaining .launch files to use ubiquity_launches as the package name.
• Added n_bus_server node.
• Added rviz_loki_sonar .
• Updated README.md
• Started switching over to launch repository architecture.
• Documented proposed new structure for robot lauch repositories.
• Added xloki_joy
• Merge branch 'master' of https://github.com/UbiquityRobotics/loki_robot
• Tweaked some command for local_costmap.
• Set up the rviz_local_costmap.launch file and updated README.md .
• Moved fiducial_slam launch file to loki_robot repository.
• Moved fiducial_detect launch file to loki_robot repository.
• Moved camara launch file and calibration file to loki_robot repository.
• Moved joystick over to loki_robot repository.
• Moved map server to run out of loki_robot repository.
• Moved move_base parameters over to xlocal_local_costmap/params
• Got wayne.yaml into the repository.
• Added camera_pose to loki.urdf file.
• More work on local_costmap.launch
• More hacking on local_costmap.launch.
• Switched over to loki_robot robot state publisher
• Added new local_costmap.launch file. This one still has hooks into the robot-configurtions repository (which is scheduled to go away.)
• Got odometry to work with bus_server.py.
• Merge branch 'master' of https://github.com/UbiquityRobotics/loki_robot Merge in new PID parameters for Loki
• Added exprimental/wayne
• Change PID parameters for new firmware changes
• Added experimental sub-directory to loki_robot repository.
• Added rviz_sonar.launch
• Merge branch 'master' of https://github.com/UbiquityRobotics/loki_robot
• Added sonar.launch .
• Added rviz_description.launch
• Missed description.launch
• Got loki_description sub-directory to work.
• Renamed ros_arduino_bridge.launch to bringup.launch.
• Rearranched files into task based sub-directories.
• Added some more launch files.
• Added package.xml
• Initial launch and configuration files.
• Initial commit
• Contributors: Mark Johnston, Rohan Agrawal, Tony Pratkanis, Wayne C. Gramlich, Wayne Gramlich, waynegramlich