Bridging the Gap Between AI and Reality • Rhodes, Greece
Time: Tuesday, 4.11
Room: Room C
Authors: Falk Howar
Abstract: Verifying AI-based autonomous systems remains largely infeasible: neural networks are too complex to explore and lack formal specifications. Testing therefore remains the primary means of quality assurance. While it cannot prove correctness, it can provide confidence if test coverage correlates with defect detection. We approach this question from a requirement-based perspective, focusing on scenario coverage that captures the diversity and difficulty of environmental conditions under which a system operates. We present methods for defining and eliciting challenging conditions, constructing and modifying scenarios, and decomposing coverage over scenario features.