SwRI: Investigation of Techniques for Detecting Distracted Drivers Using SHRP2 Vehicle Data.

The motivation for creating the dataset was to enable SwRI to investigate the application of algorithms based on time series data to the prediction of a specific class of vehicle behavior. The work evaluated a set of measures for classifying streaming vehicle position data as to their ability to detect distracted driving. SwRI received the majority of time series data (excluding GPS position and radar) for a total of 8880 crash, near crash, and baseline events where “Locality” was equal to either “Interstate/Bypass/Divided Highway with no traffic signals” or “Open Country.” The event duration was one minute prior and three seconds after an event of interest. Data was sampled at 10 Hz. SwRI also received the SHRP2 Insite website Event table variables corresponding to the events identified for the time series data.