Multi-source enriched driving behavior models A driving simulator case study in Germany

According to the World Health Organization, an estimated 1.2 million people die each year around the globe because of road crashes (WHO, 2024). Despite noticeable progress in improving road safety, reducing road injuries and fatalities remains a challenge. Globally, efforts have focused on improving the understanding of driving behavior, so that it can be improved in the long-term on the one hand, but also incorporated in driver assistance systems and automated vehicles on the other hand. Yet, there is still a gap in the development of comprehensive models that consider simultaneously the relationship between driver, vehicle, and environment factors and their impact on selected driving behavior parameters. This paper aims to fill this existing gap by developing enriched multi-source driving behavior models, that encompass those different parameters. This study assessed data from a German driving simulator experiment in which 60 drivers participated, with tailgating and pedestrian-vehicle interactions designed as risk factors. Accordingly, panel regression models have been developed to assess driving behavior by considering various dependent variables such as time-to-collision, time-headway, maximum acceleration, brake pedal displacement and gas pedal displacement. Independent variables included data from heterogeneous sources including driving simulator data, eye movement data, and socio-demographics and attitudinal data. Obtained results highlighted the significance of driver, vehicle, and environment-related parameters on driving behavior, paving the way for future research, beyond the realm of driving simulators.