Imagine that we are going to build an autonomous mobile robot. Here are some of the reasons why people choose ROS over other robotic platforms, such as Player, YARP, Orocos, MRPT, and so on:
- High-end capabilities: ROS comes with ready-to-use capabilities. For example, Simultaneous Localization and Mapping (SLAM) and Adaptive Monte Carlo Localization (AMCL) packages in ROS can be used for performing autonomous navigation in mobile robots, and the MoveIt package can be used for motion planning of robot manipulators. These capabilities can directly be used in our robot software without any hassle. These capabilities are its best form of implementation, so writing new code for existing capabilities is like reinventing the wheel. Also, these capabilities are highly configurable; we can fine-tune each capability using various parameters.
- Tons of tools: ROS is packed with tons of tools for debugging, visualizing, and performing a simulation. The tools, such as rqt_gui, RViz, and Gazebo, are some of the strong open source tools for debugging, visualization, and simulation. A software framework that has these many tools is very rare.
- Support for high-end sensors and actuators: ROS is packed with device drivers and interface packages of various sensors and actuators in robotics. The high-end sensors include Velodyne-LIDAR, Laser scanners, Kinect, and so on, and actuators such as DYNAMIXEL servos. We can interface these components to ROS without any hassle.
- Inter-platform operability: The ROS message-passing middleware allows communication between different nodes. These nodes can be programmed in any language that has ROS client libraries. We can write high-performance nodes in C++ or C and other nodes in Python or Java. This kind of flexibility is not available in other frameworks.
- Modularity: One of the issues that can occur in most of the standalone robotic applications is that if any of the threads of main code crash, the entire robot application can stop. In ROS, the situation is different; we are writing different nodes for each process, and if one node crashes, the system can still work. Also, ROS provides robust methods to resume operations even if any sensors or motors are dead.
- Concurrent resource handling: Handling a hardware resource via more than two processes is always a headache. Imagine we want to process an image from a camera for face detection and motion detection; we can either write the code as a single entity that can do both, or we can write a single-threaded code for concurrency. If we want to add more than two features in threads, the application behavior will get complex and will be difficult to debug. But in ROS, we can access the devices using ROS topics from the ROS drivers. Any number of ROS nodes can subscribe to the image message from the ROS camera driver, and each node can perform different functionalities. It can reduce the complexity in computation and also increase the debug ability of the entire system.
- Active community: When we choose a library or software framework, especially from an open source community, one of the main factors that needs to be checked before using it is its software support and developer community. There is no guarantee of support from an open source tool. Some tools provide good support and some tools don't. In ROS, the support community is active. There is a web portal to handle the support queries from users too (http://answers.ros.org). It seems that the ROS community has a steady growth in developers worldwide.
There are many reasons to choose ROS other than the preceding points.
Next, we can check the various reasons why people don't use ROS. Here are some of the existing reasons.
