Online Program

Comparing Random and Spurious Baselines as Control Events in Naturalistic Driving Studies (306522)

Keywords: safety, research, driving safety, risk, baselines

Naturalistic driving studies generate an exceptional data set- driver and roadway video and vehicle sensor data is collected continuously from study participants driving in real time. These studies have been instrumental in increasing driver and roadway-user safety, by identifying unsafe driving behaviors and conditions. Risk analyses have compared safety-critical events, such as crashes and near-crashes, to “random baselines” of normative driving random samples.

Technology vendors have developed in-vehicle onboard safety monitoring systems (OBMS), which collect data from cameras and sensors to provide feedback to professional drivers. Possible safety events are identified using sensor data algorithms and reviewed in video to determine if the event is a true safety-related event or a non-safety event (spurious baseline). OBMS have collected hundreds of millions of events from a larger sample of drivers. However, spurious baselines may introduce bias in assessing driver engagement in tasks. The presentation will describe findings of a study comparing random and spurious baselines from 6,379 vehicles and the impact on future safety research.