A Novel Approach for the Safety Validation of Emergency Intervention Functions Using Extreme Value Estimation

Abstract

As part of the safety validation of advanced driver assistance systems (ADAS) and automated driving (AD) functions, it is necessary to demonstrate that the frequency at which the system exhibits hazardous behavior (HB) in the field is below an acceptable threshold. This is typically tested by observation of the system behavior in a field operational test (FOT). For situations in which the system under test (SUT) actively intervenes in the dynamic driving behavior of the vehicle, it is assessed whether the SUT exhibits HB. Since the accepted threshold values are generally small, the amount of data required for this strategy is usually very large. This publication proposes an approach to reduce the amount of data required for the evaluation of emergency intervention systems with a state machine based intervention logic by including the time periods between intervention events in the validation process. For this purpose, a proximity measure that indicates how close the system is to an intervention at each point in time during the test drive is proposed. The application of this proximity measure and the definition of a corresponding threshold value makes it possible to expand the set of observable intervention events by events in which the system is close to an intervention. Thus, a subsequent assessment of these additional events regarding HB enables the data basis to be expanded to include events in which the system is close to exhibiting HB. This additional information is intended to be leveraged in the application of an extreme value estimator for deriving an estimate of the frequency at which the system is expected to exhibit HB on longer test distances. This publication focuses primarily on deriving and demonstrating the described proximity measure and provides an outlook on further steps required to validate the proposed approach.