Defining Pass-/Fail-Criteria for Particular Tests of Automated Driving Functions

Abstract

The approval effort for automated vehicles is a key challenge to achieve their introduction on public roads. Current approval methods in the automotive industry would lead to an unfeasible introduction of automated vehicles. While new methods such as the scenario-based approach are developed to reduce the approval effort for an automated vehicle regarding the whole system, another strategy is to separate the system into several subsystems which are approved independently. A method to decompose the system into functional layers and reduce the parameter space has been developed in the project PEGASUS, while within the research project UNICARagil the system is divided into modules to raise the controllability of the parameter space of each module. This paper shows the challenges of these proposals and presents a method to face these challenges by defining particular pass-/fail-criteria. Starting with safety goals, causal factors are derived on subsystem level using proven failure and hazard analysis methods. Safety requirements based on these causal factors are defined and finally used for the definition of pass-/fail-criteria. An example of this method using functional layers and modules is given and an exemplary use case for defining thresholds to make these criteria testable is provided.