This artikel presents the experimental plan and results of driver trials conducted with a semi-dynamic driving simulator and an equipped research vehicle in real traffic conditions. 12 trainees, 12 novice (driving license less than 1.5 years; 25,000 totally run km by average) and 12 experienced drivers (driving license more than 12 years; 320,000 totally run km and 28583 km only the last year by average) participated in the trials, conducting the same driving sessions in the simulator and on-the-road, consisting of a free driving and a following vehicle scenario in the context of highways, rural and urban roads. Their maximum speed, difference of their average speed from each speed limit and their average lateral deviation in driving simulator and on-the-road were compared. A novel research framework was established for the results analysis dealing with the existence of numerical proximity between the driving simulator and the on-the-road measurements (in the sense of absolute validity), the degree of this proximity and the possibility and the type of the correction that could be applied leading to an acceptable absolute validity. Derived results have shown that several degrees of numerical proximity apply for different combinations of driver behaviour metrics, driver cohorts, driving tasks and road contexts, whereas the type of the driver behaviour metric seems to be the most determining factor for absolute validity existence. Tangible findings consist of transfer algorithms that serve as the basis for redefining warning and intervention thresholds of ADAS but also driver training and assessment schemes based on driving simulators. (Author/publisher)
Abstract