MROS: runtime adaptation for robot control architectures

Abstract

Known attempts to build autonomous robots rely on complex control architectures, usually implemented with the Robot Operating System (ROS). Runtime adaptation is needed in these systems, to cope with component failures and with contingencies arising from dynamic environments – otherwise these affect the reliability and quality of the mission execution. Existing proposals on how to build self-adaptive systems in robotics usually require a major re-design of the control architecture and rely on complex tools unfamiliar to the robotics community. Moreover, they are hard to reuse across applications. This paper presents MROS: a model-based framework for runtime adaptation of robot control architectures based on ROS. MROS uses a combination of domain-specific languages to model architectural variants and capture mission quality concerns, and an ontology-based implementation of the MAPE-K and meta-control visions for runtime adaptation. The experiment results obtained applying MROS in two realistic ROS-based robotic demonstrators show the benefits of our approach in terms of the quality of the mission execution, and MROS's extensibility and reusability across robotic applications.