Model-Driven Safety of Autonomous Vehicles

We make the case that since model-based development of complex software-intensive systems has proven to be so effective, a model-based paradigm that encompasses assurance of the system makes excellent sense and will result in more rigorous, less ad hoc approaches to the development and maintenance of assurance cases. This will become especially clear in the manufacturing of autonomous motor vehicles. Adequate demonstration of the safety of autonomous vehicles is a huge challenge. Doing it once for a single vehicle is difficult. Doing it for multiple vehicles in a product family and coping with incremental changes in design from one model version to the next without redoing the complete safety analysis is even more difficult. We show that a comprehensive, rigorous model-driven approach to development and assurance holds the promise of more efficient and more effective assurance in general and also provides a mechanism for incremental assurance. We also briefly compare that with one of the current staples for documenting assurance cases – Goal Structuring Notation.