Middleware Interoperability for Robotics: A ROS–YARP Framework

Abstract

Middlewares are fundamental tools for progress in research and applications in robotics. They enable the integration of multiple heterogeneous sensing and actuation devices, as well as providing general purpose modules for key robotics functions (kinematics, navigation, planning). However, no existing middleware yet provides a complete set of functionalities for all robotics applications, and many robots may need to rely on more than one framework. This paper focuses on the interoperability between two of the most prevalent middleware in robotics: YARP and ROS. Interoperability between middlewares should ideally allow users to execute existing software without the necessity of: (i) changing the existing code, and (ii) writing hand-coded ``bridges'' for each use-case. We propose a framework enabling the communication between existing YARP modules and ROS nodes for robotics applications in an automated way. Our approach generates the ``bridging gap'' code from a configuration file, connecting YARP ports and ROS topics through code-generated YARP Bottles. % %The configuration file must describe: (i) the sender entities, (ii) the way to group and convert the information read from the sender, (iii) the structure of the output message and (iv) the receiving entity. Our choice for the many inputs to one output is the most common use-case in robotics applications, where examples include filtering, decision making and visualization. % We support YARP/ROS and ROS/YARP sender/receiver configurations, which are demonstrated in a humanoid on wheels robot that uses YARP for upper body motor control and visual perception, and ROS for mobile base control and navigation algorithms.