University of California Institute of Transportation Studies

To improve the safety of automated vehicles, this study developed a mathematical model to indicate the degree to which certain traffic conditions, driver characteristics, and behaviors of an automated driving system (ADS) affect the probability of a driver safely taking over control in a level 3 ADS to avoid a potential crash. (A level 3 ADS requires the driver to intervene when the ADS cannot safely handle a situation.) Experiments in which 50 participants used a driving simulator informed the model development. The presence of a warning signal from the ADS telling the driver to intervene significantly reduced collision probability, improved driver reaction time, and increased the safety margin between the study vehicle and the vehicle in front of it. Traffic complexity—the number of nearby vehicles—increased the probability of hard-braking, tailgating, and steering variability, and it increased the drivers’ perceived workload and distrust of automation. The absence of a driver warning from the ADS interacted with higher traffic complexity, so that their combination greatly increased collision probability. These findings indicate that driver warnings provide measurablesafety benefits and will be used to refine the model and examine the effects of the design and timing of warnings.