Merging onto the Freeway: Comparison of Target and Actual Performance to Develop ADAS

Abstract

The paper examines an approach for the development of Advanced Driver Assistance Systems (ADAS). The basic idea of this approach is to compare the actual driving behavior with a pre-defined target performance to detect deviations which are regarded as driving errors. This error detection can be used by ADAS to initiate countermeasures and thus avoid critical traffic situations. The experiment described in this paper examines this approach using the driving task to merge onto a freeway. The target behavior is defined from expert interviews of driving instructors. The actual behavior is measured with the DLR ViewCar® in test drives. The analyses show that the comparison of actual and target behavior results in different error classes corresponding to critical situations. From the description of the errors requirements for different ADAS functions can be derived to counteract these errors. Overall, at least for the driving task examined this approach is useful for detecting errors but not predicting them. However, this may be sufficient in many cases to avoid accidents.