Case Study Analysis of STPA as Basis for Dynamic Safety Assurance of Autonomous Systems

Although in the last decade we have seen a rapid increase in technological and academic developments in the domain of autonomous systems, ensuring the safety of these systems remains extremely challenging. Previously, the concept of dynamic safety cases (DSCs), ConSerts and runtime monitoring frameworks have been presented as an engineering solution for through-life safety assurance. However, these techniques will (initially) be only as good as the claims or rules they hold. Therefore, in this paper we provide a critical view on the use of STPA as a safety and risk analysis technique underlying the design of autonomous systems to aid with the system design and runtime safety assurance. Based on two case studies, we conclude that STPA is a promising technique towards safer autonomous systems. Although hazard generation and modularisation are not easy, we argue that STPA provides a different view on safety which aligns much better with an autonomous system view.