Using Naturalistic Driving Data to Develop Simulator Scenarios
A variety of techniques have been utilized to understand driver behavior as a means toward identifying the factors that contribute to crashes and driver safety. Such approaches include driving simulation and large-scale observational studies of naturalistic driving behavior (e.g., SHRP-2). Driving simulators offer the benefit of experimental control and the ability to draw cause-effect conclusions. However, the extent to which simulator results translate to on-road performance remains a question. Observational studies of naturalistic driving behavior, on the other hand, benefit from studying behavior in the context of real driving, and allow for estimation of crash risk. However, naturalistic studies provide only a single "snapshot" of a given driving event (e.g., a crash), which hinders the ability to draw conclusions about causality or to determine the impact of certain factors on crash risk. Thus, an important question is how naturalistic data can be used to inform the design of simulator scenarios to be used to investigate the role of driver and roadway factors in crashes.
The goal of the proposed study is to use the SHRP-2 naturalistic driving data set to develop simulator scenarios based on real-world roadways where real crashes occur. Then, using these simulator scenarios, we will experimentally investigate the impact of several factors (i.e., distraction, drowsiness) and driver characteristics (e.g., age) on driver behavior in these roadway locations. Finally, we will code videos of driver behavior during the simulation and compare this against previously completed SHRP-2 coding to establish the correspondence between naturalistic and simulator data.