Computer-Assisted Engineering for Robotics and Autonomous Systems

Motivation

An autonomous system is a system which performs certain intended tasks based on its current state and sensing, without human supervision. There are several fields in which autonomous systems play an increasing role like, e.g., aerospace, railway signaling or automotive engineering.

In the absence of external control, it is highly important to make sure that autonomous systems are functionally safe. The more critical system safety is, the more important is to introduce standardized certifications to assure their correct functioning. However, though au- tonomous systems are increasingly involved in our everyday life, both exact formalizations of safe functionality (standards, what we want to be confident in), and methods to achieve confi- dence (methodologies, how we get confident in the properties we want to assure) are still scarce and not well integrated.

To move towards the initiation of such standards and the development of such technologies, in this seminar we want to discuss at an abstract level (1) what are general safety requirements for autonomous systems, (2) what are potential key technologies that could be employed to assure that those safety requirements hold, and (3) what are the obstacles on the way to the application of those technologies.

Robotics is a typical area for the development of Autonomous Systems . As models are a basic requirement for the formal analysis of systems, Model-driven Software Engineering plays an important role to enable the application of Formal Methods. With this seminar we would like to bring together these communities to exchange knowledge, opinions and needs, to develop recommendations, and to initiate road maps to increase the safety of autonomous systems.

In the above communities, a plethora of interesting activities and innovative results can be observed, however, without close connections between the different areas. For example in Robotics, Model-driven Software Engineering methods are still seldomly used, and even when they are applied, the approaches are not general, i.e., not directly transferable to other ap- plication domains. Also Model-driven Software Engineering and Formal Methods are not well connected, e.g., novel results in hardware/software verification are not directly transferred to Model-driven Software Engineering technologies. As Formal Methods need formal models of the systems of interest, Model-driven Software Engineering could build a bridge between Robotics and Formal Methods, being an enabling condition for the application of Formal Methods to Autonomous Systems.