On Risk Formalization of On-Line Risk Assessment for Safe Decision Making in Robotics

In the near future service robots may fulfill demanding tasks in unstructured and complex human-like environments. A central challenge is to prevent such autonomous systems from provoking hazards by interaction with their environment, especially when it is partially unknown. Real-time risk information is seen as essential basis for safe decision making processes. Therefore, generalized and conservative safety principles are used to determine the interaction risks. The solution for formalizing safety principles and for quantifying necessary risk values is shown.