Fitness-for-Purpose (F4P) Analytics is a novel methodology for assessing the reliability of AI-based perception in automated driving systems. It addresses the SOTIF challenge of understanding the relations between triggering conditions and functional insufficiencies. The methodology models the system’s information flows and evaluates how uncertainties—such as sensor limitations, component degradation, and environmental conditions—affect the performance of critical AI-driven functions. Using causal reasoning, the approach quantifies the likelihood that the system can operate safely and effectively within its intended domain. This presentation introduces the F4P framework and demonstrates its application in automated driving verification & validation and risk assessment.

• Learn about the novel Fitness-for-Purpose methodology for assessing the reliability of AI-based perception in automated driving systems.
• Learn how to support SOTIF analysis by linking triggering conditions to functional insufficiencies via structured modeling.
• Learn how to model information flows to identify where and how uncertainty impacts AI-based system performance.
• Learn how to use causal reasoning to quantify the likelihood of safe operation under varying conditions.
• Learn how environmental and technical factors are integrated into performance assessment using probabilistic models.