Taxonomy of Safe Uncertainty Handling in the Context of Dynamic Risk Management

Autonomous systems bear great potential to provide a multitude of benefits for society. In this context, the term utility describes the system's usefulness. Due to uncertainty related to the operating conditions as well as the utilised technologies, providing high utility and simultaneously ensuring safety is often contradictory. Due to imprecise risk assessments, systems might operate more conservatively even though the situation permits more permissive control approaches. Dynamic Risk Management (DRM) addresses this issue through the continuous monitoring of the system's operating context in order to optimise utility while simultaneously ensuring safety. This work advances the state-of-the-art by classifying the spectrum of DRM schemes in the dimensions of utility and assurability. The ability of a system to provide precise and trustworthy uncertainty estimates is thereby identified as major factor to improve utility.