Improving the Evaluation Process for Active Safety Functions

Sammanfattning: The general aim of the present thesis was to improve key steps in the procedure for functional, formative evaluation of Advanced Driver Assistance Systems (ADAS). Five unresolved theoretical and empirical issues were identified and addressed. The first identified issue was the lack of a general conceptual framework for ADAS evaluation that can help formulate functional specifications and generate testable hypotheses on ADAS influence in critical driving scenarios. In response, a conceptual framework called Situational Control was developed. The second issue concerned the current ways in which crash data is used to specify ADAS evaluation scenarios. An improved methodology for linking a set of in-depth investigated case studies to a general crash type was developed and successfully tested. The third issue concerned the extent to which data from in-depth investigations of fatal crashes can be used to specify ADAS evaluation scenarios. Some countries have fully representative in-depth investigated datasets for this crash type, but their relevance for ADAS evaluation has not been investigated. An empirical study of causation information in fatal intersection crashes was performed. However, the information collected in these investigations was found to be limited in ways which made them less useful for defining ADAS evaluation scenarios. The fourth issue was whether sufficiently critical driving events that result in realistic driver responses can be created and repeated in driving simulator based ADAS evaluation. A study was performed in which two groups of drivers, one with and one without FCW, were exposed to repeated critical lead vehicle braking events. Results indicate that while creating a single surprise event is possible, interaction effects that compromise result generalizability occur when the critical event is repeated. The fifth issue concerned principles for how to assess the combined influence multiple ADAS when present in the same vehicle. A study of an FOT evaluated ADAS bundle consisting of FCW and ACC was carried out to empirically test whether existing conceptual models for calculating the combined effect of multiple safety functions were applicable. The results indicate that existing models were too simplistic to account for the complex modifications of driver behavior found in the data.