Classification and Integration of Software Component Models for Robotics

The Component Based Software Engineering (CBSE) allows the realization of modular and flexible systems composed by reusable software components. In the robotics field, the CBSE principles struggle to become spread used due to the extreme variability in functionality, applications and involved hardware that characterize the domain. This variability directly impacts on the component implementations limiting their reuse capability. The component model concept plays a fundamental role since it defines the rules that supervise the components definitions and their interactions. These rules strongly influences the capability of components to cope with the variability of the robotics domain. The high number of proposed component models, both in literature and in the industrial field, claims for a model for their comparison and classification in order to highlight commonalities and differences among different approaches. In this work we propose a feature-based classification model that exploits a clear separation of concerns when analyzing different component models. The four identified concerns refer to the Communication, the Computation, the Configuration and the Coordination aspects. This approach allows for a better understanding of the constructs provided by different models in order to manage the variability of the system. From the classification, it results clear that robotic domain specific models offer constructs that are well suited for addressing the specific issues of robotics although they lack in providing a sufficient level of flexibility. On the other hand, general purpose models offer a higher level of flexibility without offering relevant robotic specific features. We propose the integration of component models as a key to overcome this limitation. In particular, the integration between the Service Component Architecture (SCA) and the Robot Operating System (ROS) allows the developers to fully exploit the benefits of both approaches while mitigating their deficiencies. We present the control software of the BART robot as a proof of the benefits of this approach.