Risk-Driven Testing and Certification in Simulated Environments

Abstract

Simulated environments are essential for testing smart software-intensive systems that rely on AI components in domains like manufacturing, automotive or aerospace. However, it is essential to guide testing by requirements which enable the coverage of risks. This approach has also high potentials for virtual certification of system components, which has huge potential to safe cost in safety-critical system design. In this talk, we present an approach to risk-driven automated testing of collaborative AI systems in the manufacturing domain and reflect on potentials and limitations of risk-driven testing and certification in safety-critical domains.