Lifting ROS to Model-Driven Development: Lessons Learned from a bottom-up approach

Abstract

The benefits of using model-based technologies are well proven in different sectors. However, in robotics, where the dominant framework is one whose nature is to create the code by hand, MDD approaches are struggling to gain a foothold. Our approach is distinctly different from traditional MDD solutions, we have created models and tools with a bottom-up perspective, i.e. analyzing “ everyday” code and encapsulating this existing code in models. In this paper we present the lessons learned from (1) designing the models to encapsulate ROS hand-written code, (2) creating tools to ease the production and debugging of complex systems using the models, and (3) applying them to real use cases. Through this work we have identified in which situations it is worth using model-based technologies and in which situations the traditional approach is more meaningful. Our aim is to give a broad view of our conclusions and inspire the design of new model-based tools for the assistance of roboticists.