Multi-paradigm modeling for early analysis of ROS-based robotic applications using a library of AADL models

ROS, the Robot Operating System, is a middleware that eases the programming of robotic applications significantly, bringing standard communication and synchronization mechanisms to a wide variety of operating systems and computers or embedded computer boards. However, a robotic application is a complex set of many services, with many relations between them, and multiple choices have to be made regarding the software and the hardware architectures. To rely on a formal representation of those two, from which early analysis can be performed, is extremely beneficial since it allows detection of design errors early in the process. This is what we present in this paper. Our approach is based on the Architecture Analysis and Design Language (AADL) and on a set of AADL models to represent both the application and the robot, including its embedded computers and its many devices. Those models are gathered into an open source AADL library from which ROS developers can largely benefit.