Doctoral Dissertations

Orcid ID

Date of Award


Degree Type


Degree Name

Doctor of Philosophy


Civil Engineering

Major Professor

Asad Jan Khattak

Committee Members

Hamparsum Bozdogan, Candace Brakewood, Shuai Li


Human factors including driver behavior and performance are highly relevant to the safe systems approach as these factors tend to be predominant causes of road crash occurrence. Transportation safety can be enhanced by applying the safe systems approach to harness new forms of large-scale naturalistic driving study data. The naturalistic driving study data, collected through the 2nd Strategic Highway Research Program, include real-world microscopic details about pre-crash driving behaviors, performance, instantaneous vehicular movements (speed, acceleration, and deceleration), and distracted driving (along with its duration) which were collected via recording equipment in the instrumented vehicles. One of the unique aspects of the naturalistic driving data is that it includes data about baselines (non-event driving) and near-crashes where the earlier ones could be used as a surrogate to the exposure to computing the risk of a crash or near-crash. To examine the role of human factors and improve crash investigations, this dissertation first develops a systematic taxonomy of driving errors and violations using a subsample of naturalistic driving data. Next, after developing a taxonomy of naturalistic driving errors and violations, and deriving measures of driving volatility, this dissertation investigates the role of driving errors, violations, duration of distraction, and driving volatility in safety-critical events (crashes and near-crashes) in diverse roadway environments. To unveil the aforementioned key relationships, more robust path analytic approaches including the ones via joint estimation are used to account for potential correlation between the unobserved factors in the two models. From the prediction standpoint, this dissertation applies machine learning and artificial intelligence methods to understand the significance of human factors (driving errors, violations, duration of distraction, and instability in driving) compared to roadway environmental factors in predicting safety-critical events. Finally, the dissertation looks at the effects of errors, violations, and impairment by both pedestrians and drivers on pedestrian injuries using comprehensive data of North Carolina Pedestrian-vehicle crashes coded by professional data experts. One of the unique methodological aspects of this dissertation is to account for the unobserved heterogeneity in pedestrian injuries and in the effects of observed variables (e.g., light conditions and impairment) on pedestrian injuries due to some unobserved factors where the later ones could be used to classify pedestrians into latent (unobserved) classes using finite mixture models. Overall, this dissertation contributes by providing firm knowledge and foundation about the role of human factors in road crashes which could help in selecting proactive countermeasures thus enhancing the overall safety. Based on the findings, this dissertation discusses the practical implications and discusses potential countermeasures including traditional roadway (built-environment) changes, policy interventions, vehicle technologies (e.g., adaptive cruise control, automatic braking system, dilemma zone mitigation systems, and collision warning systems) which have the potential to proactively reduce driving errors, violations, and driving instability thus mitigating the chance of safety-critical events. From the automated vehicles standpoint, this research discusses the potential procedure which can be used to predict driving errors, violations, instability in driving, and especially safety-critical events in real-time based on human factors in diverse roadway environments.

Available for download on Sunday, December 15, 2024

Files over 3MB may be slow to open. For best results, right-click and select "save as..."