Increasing ROS reliability and safety through advanced introspection capabilities

Abstract

The Robot Operating System (ROS) brought together many different robotics research groups by providing a common software ecosystem. One important part of ROS is the publish-subscribe messaging infrastructure that facilitates building modular and reusable distributed robotics systems. However, the inherent complexity - flexibility, diversity and dynamics - of such a system opposes properties essential to robotics: reliability and safety. The paper details how to gain control of a ROS-based robotics system through the extension of ROS by advanced introspection capabilities. Once the introspection based on distributed gathering of metadata is in place, we are able to define a reference state for the whole ROS system. A monitoring node continuously compares the reference to the actual state. In case of a deviation the operator is alerted through a dashboard, in addition appropriate countermeasures can be initiated by the robotics system itself.