Structuring Data to Compare Driver Behavior Across Driving Simulators

Abstract

Driving simulators have been used to train and assess drivers and develop vehicle technology in a safe and controlled environment. An increasing number of simulators are being used as research tools to study human performance in general, and the Federal Highway Administration (FHWA) has begun using driving simulators to evaluate the effect of different road designs on driving behavior and safety. The general goal has been to establish a sufficiently high level of fidelity to assure that the adopted training programs and performance assessments offer meaningful transfer to reality. Simulators have become more common, their applications more diverse, and simulator fidelity has increased markedly, but the ability to easily compare data between simulators has not progressed in a similar manner. This paper demonstrates an approach to structure data to facilitate comparison and aggregation across simulator platforms. This work is part of a bigger project to compare and contrast data from multiple simulators with the purpose to determine the level of fidelity simulator needed to make meaningful road design assessments that transfer to reality. A study was performed to quantify the degree to which four different driving simulators each configured to different levels of realism were able to accurately reproduce real world drivers’ behavior in different roundabouts and residential roads with infrastructural speed treatments. To facilitate comparison of many behavioral elements across the simulators and the real world, it is crucial to first transfer all the data to a standard data structure. This paper motivates and details such a data structure and demonstrates its utility by showing several informative comparisons between these data sets.