In 2017, Shalev-Schwartz, Shammah and Shashua published an article on the preprint server arXiv entitled «On a Formal Model of Safe and Scalable Self-driving Cars». In this paper, a deterministic safety model called RSS is introduced and the article claims that if every road user adheres to the rules given by RSS, then no accidents can happen. The goal of this project was to inspect the article and to either validate the model or to find counterexamples to the claims made therein. Although the approach presented in the article seems promising, we were able to find shortcomings of the work by constructing scenarios in which RSS does not provide a guarantee for no accidents. Furthermore, we detected lines of thought along which the model could be refined.
Joint work with Corinne Hager Jörin and Florence Yerly.