ROS Development
The Robot Operating System (ROS) is a flexible framework for writing robot software. It is a collection of tools, libraries, and conventions that aim to simplify the task of creating complex and robust robot behavior across a wide variety of robotic platforms
ROS will give us the possibility to write and run different processes (called nodes) that communicate with each other by sending and receiving messages on named buses (called topics) or by calling remote procedures (called services). Please read the ROS/Concepts Wiki page to get a more clear overview of the concepts related to ROS.
This section will describe some basic ROS functionality that can be accomplished with stock Leo Rover.
Prerequisites
Introspecting ROS network with command line tools
ROS comes with some command line tools that can help to introspect the current network of running nodes. Some of the available tools are as follows:
- rosnode - printing information about currently running nodes, killing them, testing connectivity,
- rostopic - listing and printing information about topics currently in use, printing published messages, publishing data to topics, finding a type of published messages
- rosservice - listing and printing information about available services, calling the service with provided arguments,
- rosmsg - displaying the fields of a specified ROS message type.
Let's try running some examples. Before that, connect to Leo Rover via SSH.
Start by reading currently running nodes:
rosnode list
You should see most of all the nodes described in the
first section of
this tutorial. Among them, the rosserial server node (called /serial_node in
this case) "bridges" communication with the CORE2 board, so any topics it
publishes or subscribes are created and used in the firmware.
Let's get more information about this node:
rosnode info /serial_node
You should see all the subscribed, published topics and services that the firmware provides. You can learn more about each topic in leo_firmware README page.
Among published topics, you should see the /battery topic. Let's read the
published values using rostopic tool:
rostopic echo /firmware/battery
Now, let's look at the /cmd_vel topic. This topic is used by the firmware to
receive drive commands. We can look at its type:
rostopic type /cmd_vel
You should get geometry_msgs/Twist. This is a standard message in ROS for
commanding velocity controlled ground robots. We can look up the message
description using rosmsg tool:
rosmsg show geometry_msgs/Twist
The description should look like this:
geometry_msgs/Vector3 linear
float64 x
float64 y
float64 z
geometry_msgs/Vector3 angular
float64 x
float64 y
float64 z
The linear field represents linear velocity (in meters per second) along x, y,
z axes. angular field represents angular velocity (in radians per second)
along the same axes.
You can read more about standard units of measure and coordinate conventions in REP103
For differential drive robots like Leo Rover, only linear.x and angular.z
values are used.
We can use rostopic tool to actually command the rover to move forward, by
sending messages to /cmd_vel topic:
rostopic pub -r 10 /cmd_vel geometry_msgs/Twist -- "linear: {x: 0.2}"
The rover should start moving forward with a velocity of 0.2 m/s. To stop message publishing, simply type Ctrl+C.
The -r 10 argument tells the rostopic tool to publish the message repeatedly 10 times per second instead of publishing only one message. This is necessary because the firmware implements a timeout that will stop the rover if it doesn't receive the next command after some time (half a second by default).
Using ROS client library to publish messages
ROS provides several client libraries that let you write ROS nodes in different languages. The most common ones are roscpp for C++ and rospy for Python.
Here is a simple Python node that commands the rover by publishing to /cmd_vel
topic:
#!/usr/bin/env python3
import rospy
from geometry_msgs.msg import Twist
# Initialize ROS node
rospy.init_node("test_drive")
# Create ROS publisher
cmd_pub = rospy.Publisher("cmd_vel", Twist, queue_size=1)
# Write a function that drives the Rover with specified
# linear and angular speed for 2 seconds
def drive(linear, angular):
# Initialize ROS message object
twist = Twist()
twist.linear.x = linear
twist.angular.z = angular
for _ in range(20): # repeat 20 times
cmd_pub.publish(twist) # publish message
rospy.sleep(0.1) # sleep for 100ms
# Now let's actually test driving the Rover
# linear speed is in m/s and angular speed in rad/s
drive(0.2, 0.0)
drive(0.0, 0.0)
drive(-0.2, 0.0)
drive(0.0, 0.0)
drive(0.0, 1.0)
drive(0.0, 0.0)
drive(0.0, -1.0)
drive(0.0, 0.0)
Copy this script to Raspberry Pi filesystem.
You can paste this to new file when using a terminal. Copy the script to clipboard, then type:
cat > test_drive.py
Type Ctrl+Shift+V when using Linux terminal or Shift+Ins when using Putty. Then type Ctrl+D to end the file.
Add execute permission to the file:
chmod +x test_drive.py
And execute it by typing:
./test_drive.py
The rover should drive forward and backward, then, turn in place in left and right directions.
Make sure you don't have a Web UI running at the moment as it may cause conflicts on /cmd_vel topic
Adding additional functionality to the rover
LeoOS provides an easy mechanism for adding new functionalities without building any of the base packages. The whole process of starting the ROS nodes at boot can be summarized by the following files:
- /etc/ros/robot.launch - a
launch filethat starts the robot's functionality. It includes the launch file from the leo_bringup package which starts the base functionality of the rover, but also allows to add additional nodes to be started or parameters to be set on the Parameter Server.
A launch file is an XML file that describes a set of nodes to be stared with specified parameters. It can be interpreted with roslaunch tool.
- /etc/ros/setup.bash - the environment setup file that sets all the
environment variables necessary for the successful start of the ROS nodes. It
sources the environment setup file from the target ROS distribution (by
default
/opt/ros/melodic/setup.bash) and sets additional environment variables used by ROS. - /etc/ros/urdf/robot.urdf.xacro - the URDF description (in
xacro format) that is uploaded to the Parameter
Server by the
robot.launchfile. It includes the robot's model from the leo_description package, but also allows to add additional links or joints to the model. - /usr/bin/leo-start - a script that starts the robot's functionality. In
short, it sources the
/etc/ros/setup.bashfile and launches the/etc/ros/robot.launchfile. - /usr/bin/leo-stop - a script that stops the currently running
leo-startprocess.
On top of that, the leo systemd service starts the leo-start script when the
computer boots.
Starting the functionality manually
To start the nodes manually, first, you need to stop the ones that are currently
running. You can do this either by using the leo-stop script:
leo-stop
or by stopping the leo service:
sudo systemctl stop leo
If you wish to disable the service from starting at boot, you can type:
sudo systemctl disable leo
To turn the service back on, just type:
sudo systemctl enable leo
Now, to start the nodes manually, type:
leo-start
Type Ctrl+C to stop the nodes and exit the script.
Adding additional nodes to the launch file
To add additional nodes to be started, you can modify the
/etc/ros/robot.launch file. Take a look at the
launch file XML specification (especially
the node and
param tags) for reference.
Here's an example that uses node and param tags:
<param name="name_of_the_global_parameter"
value="value_of_the_parameter"/>
<node name="name_of_the_node"
pkg="name_of_the_package"
type="name_of_the_executable">
<param name="name_of_the_private_parameter"
value="value_of_the_parameter"/>
</node>
Modify it to your needs, add it to the /etc/ros/robot.launch file and restart
the nodes.
If you want your additional functionality to be easily switchable, you can put
these lines, embedded into <launch> tag, into a separate file (e.g.
/etc/ros/function1.launch) and add these lines to the /etc/ros/robot.launch
file:
<include if="$(optenv USE_FUNCTION1 false)"
file="/etc/ros/function1.launch"/>
Then, add this line to the /etc/ros/setup.bash file:
export USE_FUNCTION1=true
Now, you can toggle the functionality simply by changing the USE_FUNCTION1
environment variable and restarting the nodes.
Expanding the URDF model
When integrating a sensor or other device to your rover, you might sometimes want to extend the robot's URDF model to:
- visualize the device attached to the rover in RViz
- make the robot aware of device's collision geometry
- provide additional reference frames (for example for the sensor readings)
You can create a separate URDF file for your attached device like this one:
<?xml version="1.0"?>
<robot>
<!-- a link representing visual and collision
properties of the sensor -->
<link name="sensor_base_link">
<visual>
<origin xyz="0 0 0.05"/>
<geometry>
<box size="0.05 0.05 0.1"/>
</geometry>
<material name="red">
<color rgba="1 0 0 0.7"/>
</material>
</visual>
<collision>
<origin xyz="0 0 0.05"/>
<geometry>
<box size="0.05 0.05 0.1"/>
</geometry>
</collision>
</link>
<!-- fixed joint that attaches
the sensor to the rover's body -->
<joint name="sensor_base_joint" type="fixed">
<origin xyz="0.08 0 0"/>
<parent link="base_link"/>
<child link="sensor_base_link"/>
</joint>
<!-- reference frame for sensor readings -->
<link name="sensor_frame"/>
<!-- fixed joint that sets the origin
of the reference frame -->
<joint name="sensor_joint" type="fixed">
<origin xyz="0 0 0.06"/>
<parent link="sensor_base_link"/>
<child link="sensor_frame"/>
</joint>
</robot>
And include it in the robot's main URDF file by adding:
<xacro:include filename="/etc/ros/urdf/sensor.urdf.xacro"/>
Now, when you restart the nodes, a new URDF model should be uploaded to the Parameter Server and you should be able to view the new model in RViz.
You can use base_link as a reference frame for other links in the model. The
exact position of the base_link origin is defined as the center of this
mounting hole on the upper plane of the mounting plate:
The distance can be easily measured in CAD programs or even using physical measuring tools.
For more examples, you can look at these tutorials:
Building additional ROS packages
ROS uses its own build system for building packages. To learn about it, read the catkin/conceptual_overview and catkin/workspaces ROS wiki pages. Here's a brief summary:
The packages are the main unit for organizing software in ROS. The current build system used to build ROS packages is catkin. Catkin packages can be built as a standalone project, but the system also provides the concept of workspaces.
When building a catkin workspace, the install targets are placed into an FHS compliant hierarchy inside the result space. A set of environment setup files allow extending your shell environment, so that you can find and use any resources that have been installed to that location.
The prebuilt ROS packages (installed from the repository) are placed into
/opt/ros/${ROS_DISTRO} directory. To use the environment setup file, just
type:
source /opt/ros/${ROS_DISTRO}/setup.bash
If you use LeoOS, this line is already added to ~/.bashrc file, so it will be
automatically executed when you log into the terminal session.
The catkin build system also supports an overlay mechanism, where one workspace can extend another result space. An
environment setup filefrom the result space of such workspace will extend your shell environment by packages from both workspaces.
The build system provides a catkin_make command for building workspaces, but we will use
catkincommand line tool from Python package catkin-tools as it delivers more user-friendly and robust environment for building catkin packages.
In this chapter, we will try to:
- create workspace that extends your ROS distribution
- add leo_robot to this workspace and build the packages
- modify the
/etc/ros/setup.bashfile to use our overlay
Let's start by creating an empty workspace inside home directory on Raspberry Pi:
mkdir -p ~/ros_ws/src
cd ~/ros_ws
catkin init
We want this workspace to extend the prebuilt packages that are already
installed on the system. It should be automatically done if you have already
sourced the /opt/ros/${ROS_DISTRO}/setup.bash file, but we can also explicitly
point out the space to extend:
catkin config --extend /opt/ros/${ROS_DISTRO}
We need to get the sources of the package to build. If the package is available
as a git repository (like in our case), you can use the git clone command:
cd src
git clone https://github.com/LeoRover/leo_robot.git
Some of the packages will require installing additional dependencies to build
and run them. As the leo_bringup package is already installed on the system,
this step is redundant. For any other package, you can use rosdep to
automatically install any dependencies:
cd ~/ros_ws
rosdep update
rosdep install --from-paths src -iy
Build the workspace:
catkin build
If everything works, a development space should be created inside the devel directory. Let's source the environment setup file inside it:
source ~/ros_ws/devel/setup.bash
Now, when you execute rospack list, you should see all of the packages
installed on your system, but rospack find leo_bringup should point you to the
directory on your newly created workspace.
The last step is to modify /etc/ros/setup.bash to use our overlay. Simply edit
this file (e.g. with nano) by removing or commenting out the first line and
adding:
# source /opt/ros/melodic/setup.bash
source /home/pi/ros_ws/devel/setup.bash
When you start the nodes with the leo-start script, /etc/ros/setup.bash will
use your overlay and the /etc/ros/robot.launch file should use the version of
leo_robot that you have built in your workspace.
Connecting another computer to ROS network
ROS is designed with distributed computing in mind. The nodes make no assumption about where in the network they run. Configuring your computer to be able to communicate with ROS network will let you run nodes that interfere with Leo Rover's hardware, as well as graphical tools (like rqt or rviz) directly on your host machine.
To install ROS on your computer, you can follow this tutorial:
In this section, we will assume you run Ubuntu 18.04 with ROS Melodic.
First, connect your computer to the same network your rover is connected to. It
can be either the Rover's Access Point (LeoRover-XXXX by default) or an
external router (if you followed
Connect to the Internet tutorial).
To properly communicate over the ROS network, your computer needs to be able to resolve the master.lan hostname. Open a terminal on your computer and type:
getent hosts master.lan
If you don't see any output, that means you cannot resolve the hostname.
If you are connected to Rover's Access Point, you should be able to resolve it,
but if there is an issue with DNS server on the rover or you are connected
through external router, add this line to the /etc/hosts file on your
computer:
10.0.0.1 master.lan
If you are connected through router, you need to change 10.0.0.1 to IP address
of the Rover on your local network.
If everything works, you should be able to ping the rover by it's hostname. To
check it, type: ping master.lan.
Now, to be connected in ROS network, you need to set some environment variables. Start by sourcing the result space you are using:
source /opt/ros/${ROS_DISTRO}/setup.bash
Specify the address of the master node:
export ROS_MASTER_URI=http://master.lan:11311
And your IP on the network:
export ROS_IP=X.X.X.X
Replace X.X.X.X with your IP address.
You can check your address by typing ip address. Search for your wireless
network interface and the inet keyword.
You will need these lines executed at every terminal session you want to use ROS
on. To do this automatically at the start of every session, you can add this
line to the ~/.bashrc file.
You should now be able to do all the things from the first section of this
tutorial on your computer.
Examples of ROS use
Apart from allowing communication between different processes on Raspberry Pi, ROS will give us the possibility to remotely control the rover on your computer, as well as run graphical tools to introspect and visualize the current state of the rover. A lot of these tools are available in distribution packages in the form of rqt and rviz plugins.
Below, are some examples possible to do on the stock Leo Rover: