Overview of ROS

After you installed ROS, you will learn how to actually use ROS. Therefore, you will get an overview of what ROS does for you and how it works.

The Robot Operating System (ROS) is a meta-operating system that is installed on top of your actual operating system. In this case, you installed Ubuntu as your operating system.

To simplify things, you can see ROS as a tool-box including libraries and programs to develop and run robots. Some of the libraries are the OpenCV library for computer vision tasks and the tf library for frame transformations. Some of the tools that are included are Gazebo for robot simulations and RVIZ for visualization of your robot. In addition to that, there are software packages that have been created for specific tasks such as robot navigation or motion planning which you can simply install as they are openly available. For more information about ROS, you can have a look at the ROS Wiki.

ROS is released in different versions that go along with the current Ubuntu versions. The version discussed on this website is ROS Melodic that has been released for Ubuntu 18.04.

ROS Structure

Programs withing ROS are not simply started like other programs. Each ROS program is called a ROS node. Each node is usually built for one specific task such as navigation, reading sensor data or managing data and tasks. The nodes can communicate with each other to provide data to another node. The nodes can run with different frequency rates and they can even be executed asynchronously, which means one node can wait for the output of another node depending on the application. Having several nodes has the advantage that a failure that may crash a single node will not affect other nodes.

As an example, imagine the robotic arm will have an error and causes its node to crash, the navigation and the vision nodes will still be running. Furthermore, ROS can then restart the single node that crashed and send a warning to an operator and write the crash into a log file.

The implementation of having nodes run in parallel is also a very easy way to implement multi-threading which means the processor can run many tasks at the same time instead of one task after another.

Communication between Nodes

When several nodes want to exchange data, they need to communicate with each other. ROS has three main methods of implementing the communication between nodes: Topics, Services and Actions.

Topics

Topics are the most important type of communication in a ROS system. They provide a constant flow of data from one node to another. This can be compared with broadcast radio where there is a constant signal with audio information. The radio doesn’t wait for the radio station to send a signal and the station doesn’t care how many radios are listening.

In ROS, such a broadcaster is called a Publisher node which is publishing to a Topic. A node that is listening to that Topic is called a Subscriber which subscribes to this Topic. The Topic is the actual data-stream containing a specific type of information such as camera data or movement data.

In other words, Topics send data from a Publisher to a Subscriber in form of a data stream. This is important for applications that require constant data input. These data streams create a lot of data traffic within the ROS network which needs to be kept in mind when creating another Topic.

The most prominent Topic used in a robot is the /cmd_vel Topic that contains the velocity commands for the robot. These commands tell the robot where to go and at what speed.

For more information about Topics, you can have a look at the ROS wiki page.

Services

Services are less frequent than Topics but they can be very handy to reduce the overall data flow. Services do not provide a data stream, but instead, they work with asynchronous data requests similar to a web server hosting a website. When a person is calling a website, the computer sends a request to the server for receiving the web site content. After the web server has executed its task, it will send the requested data to the user and the communication is finished.

A Service is also implemented through setting up a Service Server which will wait for a call while the Service Client will call this Server. The Server will then perform its task once and send a signal back to the Client to tell if it finished successfully or failed. That’s it, there is no more communication between both sites. Also, the Client will wait for the response from the Server and will only resume its task when it received a response.

An example on when to use a Service would be a drone that is taking of, or even landing. When taking off, the drone simply needs a single signal to start the takeoff process and it doesn’t require constant input from the controller. When the drone is up in the air, it will send a quick response to the controller to say that it has succeeded.

For more information about Services, you can have a look at the ROS wiki page.

Actions

Actions can be seen as a middle ground between a Topic and a Service. Just like a Service there is an Action Server and a Client. Now the difference to a normal Service is that a Service Client is waiting for a response from the Server while an Action Client will continue with another task. On the Action Server, the task will be executed and the Action Server will send regular updates to the Client. These updates are less frequent than a Topic but they occur at a user determined frequency.

In other words, with Actions, the Client doesn’t need to wait, but similar to a Service it will get a response when the Server has finished its task.

A good example of using an Action would be the action of a mechanical gripper. While it is closing, it will provide you information whether it is closed or not. Also, Actions are ideal for tasks that do not need a constant data stream but yet take more time which makes it impossible to simply wait for the result. If your robot shall drive and perform another task, you cannot allow the robot to wait for the feedback of a Service, therefore, an action would be ideal in this case as the robot is still capable of observing its environment while waiting for the result of the Action.

For more information about Actions, you can have a look at the ROS wiki page.

Combination of Topics, Services and Actions

Usually, you take the tools that are best suited for the job. In the case of ROS, you have to choose whether to use a Topic, a Service or an Action. In a complex system, there are usually many different ways of communication used at the same time.

For example, then you want to control an autonomous drone, you will simply use a Service for a takeoff command while you will have a Topic that gives you the information of the position and height of the drone. Once you worked with some ROS packages, you will get a feeling of when to use which type of communication.

Software Management

The entire ROS system is managed by something called the roscore which observes all the active nodes and manage all the communication done through Topics, Services and Actions. The roscore is what creates the magic behind ROS as it manages all the tasks in the background.

The roscore is watching each node and its state. If a node crashed, it will keep track of this and depending on the configuration, it will run that node again to make sure the underlying task will be executed. As the roscore is the root of all the ROS behavior, this node needs to be the first node to be running.

Besides from watching the nodes, the roscore also manages the data traffic by looking which Topics, Services and Actions are currently available, which of these are currently called and which node is calling which message type at what time.

Programs in ROS

ROS officially supports two different programming languages: C++ and Python. C++ is more resource-efficient and the code runs faster as it is a rather low-level programming language and because it is compiled (this means, the source code is converted into machine code) before the execution. Python on the other hand is slower as it is a scripting language but this makes it faster to test the code as no compilation is necessary. If you want to learn more about how to program with Python, have a look at this programming guide!

Next to C++ and Python, many more programming languages are not officially supported by ROS, but there are libraries to bring the ROS functionality to the corresponding language such as Go or Rust.

Packages in ROS

ROS is build to be modular. This means that the software is split into several packages that can interact with each other. usually, there are packages for navigation, sensing the environment, mapping the environment, computer vision and the general control of the robot. Having several packages makes it easy to share the functionality of one package across several different robots. At the same time, when you create a new package, you only need to program the parts that are not already implemented in another package. To increase reusability, it is important to keep your code robot-agnostic which means the code should work for all situation and not only for this very specific robot. If the package works despite it doesn’t know on which robot it is running, it is more modular and can be shared with colleagues across the world or simply withing the robotics group that you are working with.