Understanding Crash and Non-driving Related Task Engagement Patterns of At-risk Drivers: A Comparison of Non-driving Related Task Clusters and an Analysis of Naturalistic Driving

Jose Antonio Calvo IV

Advisor: William S Helton, PhD, Department of Psychology

Committee Members: Yi-Ching Lee, Tyler Shaw, Carryl Baldwin

Online Location,
December 01, 2021, 11:00 AM to 01:00 PM


There are nearly 35,000 fatal automobile crashes in the US each year. The two most vulnerable road user groups, based on fatality rates per mile driven, are younger drivers, age 16-24, and older drivers, age 65 and up. Understanding the similarities and differences in driving behaviors between these two groups can inform crash mitigation strategies. This dissertation, which included 3 studies, used the Second Strategic Highway Research Project (SHRP2) Naturalistic Driving Study (NDS) to analyze naturalistic driving behavior to better understand crash patterns and non-driving task engagement patterns for these age groups as well as middle-aged drivers. The SHRP2 NDS classified over 40 specific non-driving related (NDR) tasks, many of which differ in how they distract drivers. To better understand the patterns of engagement for these NDR tasks were categorized into larger groups or clusters and then how those groups impacted driver distraction, crash severity, and crash likelihood were analyzed .  In Study 1 two approaches to categorizing NDR tasks based on Multiple Resource Theory (MRT) (Wickens 1980) were compared. The first divided tasks based on perceptual modality and processing code (visual tasks only). The second approach divided tasks solely via perceptual modality. A significant main effect on distraction was found for both age and task type for both approaches. The interaction between age and NDR task type for both approaches was statistically insignificant, but the perceptual only approach trended towards a significant interaction . Study two involved categorizing NDR tasks using a data driven model. K-means cluster method was utilized to condense NDR tasks into groups based on average driver distraction and average engagement time for each of the 39 NDR tasks identified in the SHRP2 NDS. The cluster analysis resulted in 5 clusters. A significant main effect of age group and cluster on distraction was found. No significant interaction between age group and cluster was found, but the data were trending toward significant . Study 3 compared the NDR task grouping methods from study one and study two. Two logistic regressions were run analyzing the effect of age, maneuver judgement, gender, and NDR task engagement, on crash likelihood. The methods of NDR task grouping from the previous studies were used for the regressions.  Residuals for both were compared to find the best model for grouping types of NDR tasks.