Position paper: a vision for the dynamic safety assurance of ML-enabled autonomous driving systems

Ensuring the progress of autonomous driving technology can save lives, prevent injuries, and enable reductions in traffic volume, accidents, and environmental damage caused by vehicles. Developing industry-wide safety standards and making sure producers of autonomous driving systems (ADSs) comply with them is crucial to foster consumer acceptance. Producers of ADSs can rely on assurance cases to demonstrate to regulatory authorities how they have complied with such standards. Assurance cases are mainly used in safety-critical domains (e.g., automotive, railways, avionics) to deal with high-risk concerns and show to stakeholders that such systems are safe according to domain-specific criteria. Most assurance cases are static i.e., only suitable before the deployment of a system. Dynamic Assurance Cases (DACs) have recently been introduced to provide assurance throughout the lifecycle of a system. However, from our perspective, existing standardized SACs (Static Assurance Cases) notations do not sufficiently support the representation of DACs. This hinders the standardization and adoption of DACs. In this position paper, we propose a novel approach aiming at extending existing standardized SAC notations to dynamically design DACs.