From Models to Software Through Automatic Transformations: An AADL to ROS End-to-End Toolchain

摘要

Modelling complex systems is a common practice and de facto standard across most application domains in engineering. Although it would seem unreasonable - and quite impractical - to build a structure as complex as a bridge without a reference blueprint detailing how to arrange all of its building blocks, in Software Development, and, particularly in the context of Robotics, examples adhering to rigorous modelling routines are still relatively rare to find. Yet, models help understanding complex problems while pinpointing their potential solutions, through abstraction. Further, models aid communication, i.e., the unambiguous exchange of reasoning processes across the involved agents. The complexity of Robotic Software Systems suggests that a widespread application of modelling techniques, from the very initial implementation stages, would (i) ease the definition, engineering and debugging of the related sub-features significantly, and (ii) guide collaborative efforts towards a common standard. To this aim, we presented a toolchain conceived for parsing an input AADL model into a compilable code suite. Keeping the model building and the linkage of the robot application with the ROS environment in the developer's hands, this framework delegates all the remaining tasks to an automated code generator, producing a fully-functioning ROS packages (i.e., already configured and ready for compiling) as output. We first presented the discussed framework, highlighted its related advantages - when compared to the only other similar approach found in the literature -, and used it as an exemplary use case, to prompt broader discussions on the benefits of model-based software development in Robotics.