An interaction model for the driver-vehicle-environment loop is established to clarify terms like subjective borderline situation or traffic conflict. In addition, the reasons for occurrence of borderline situations are outlined. In an experiment in real traffic three versions of Adaptive Cruise Control (ACC) were tested, differing mainly in the maximum deceleration which could be applied by the system. The goal of this experiment was to study borderline situations caused by a mismatch between the driver's internal model of the ACC and the real ACC. The drivers showed different frequencies of braking intervention depending on the ACC version. The decision making on intervention could be modeled using the predicted minimum approach time distance as a criterion. It can be shown that drivers are able to learn ACC system behaviour thus reducing driver workload caused by the intervention decision process, and this workload turned out to be highest with an ACC with medium deceleration capability. As a result, adequate display systems can be developed now to reduce borderline situations by supporting the driver's prediction capabilities. (A)
Abstract