Motor vehicle crashes are one of the leading causes of death in the United States. Traffic safety research targets at understanding the cause of crash, preventing the crash, and mitigating crash severity. This dissertation focuses on the driver-related traffic safety issues, in particular, on developing and implementing contemporary statistical modeling techniques on driving risk research on Naturalistic Driving Study data. The dissertation includes 5 chapters. In Chapter 1, I introduced the backgrounds of traffic safety research and naturalistic driving study. In Chapter 2, the state-of-practice statistical methods were implemented on individual driver risk assessment using NDS data. The study showed that critical-incident events and driver demographic characteristics can serve as good predictors for identifying risky drivers. In Chapter 3, I developed and evaluated a novel Bayesian random exposure method for Poisson regression models to account for situations where the exposure information needs to be estimated. Simulation studies and real data analysis on Cellphone Pilot Analysis study data showed that, random exposure models have significantly better model fitting performances and higher parameter coverage probabilities as compared to traditional fixed exposure models. The advantage is more apparent when the values of Poisson regression coefficients are large. In Chapter 4, I performed comprehensive simulation-based performance analyses to investigate the type-I error, power and coverage probabilities on summary effect size in classical meta-analysis models. The results shed some light for reference on the prospective and retrospective performance analysis in meta-analysis research. In Chapter 5, I implemented classical- and Bayesian-approach multi-group hierarchical models on 100-Car data. Simulation-based retrospective performance analyses were used to investigate the powers and parameter coverage probabilities among different hierarchical models. The results showed that under fixed-effects model context, complex secondary tasks are associated with higher driving risk. / Ph. D.

Fixed-object crashes have long time been considered as major roadway safety concerns. While previous relevant studies tended to address such crashes in the context of roadway departures, and heavily relied on police-reported accidents data, this study integrated the SHRP2 NDS and RID data for analyses, which fully depicted the prior to, during, and after crash scenarios. A total of 1,639 crash, near-crash events, and 1,050 baseline events were acquired. Three analysis methods: logistic regression, support vector machine (SVM) and artificial neural network (ANN) were employed for two responses: crash occurrence and severity level. Logistic regression analyses identified 16 and 10 significant variables with significance levels of 0.1, relevant to driver, roadway, environment, etc. for two responses respectively. The logistic regression analyses led to a series of findings regarding the effects of explanatory variables on fixed-object event occurrence and associated severity level. SVM classifiers and ANN models were also constructed to predict these two responses. Sensitivity analyses were performed for SVM classifiers to infer the contributing effects of input variables. All three methods obtained satisfactory prediction performance, that was around 88% for fixed-object event occurrence and 75% for event severity level, which indicated the effectiveness of NDS event data on depicting crash scenarios and roadway safety analyses. / Master of Science / Fixed-object crashes happen when a single vehicle strikes a roadway feature such as a curb or a median, or runs off the road and hits a roadside feature such as a tree or utility pole. They have long time been considered as major highway safety concerns due to their high frequency, fatality rate, and associated property cost. Previous studies relevant to fixed-object crashes tended to address such crashes in the contexture of roadway departures, and heavily relied on police-reported accident data. However, many fixed-object crashes involved objects in roadway such as traffic control devices, roadway debris, etc. The police-reported accident data were found to be weak in depicting scenarios prior to, during crashes. Also, many minor crashes were often kept unreported. The Second Strategic Highway Research Program (SHRP2) Naturalistic Driving Study (NDS) is the largest NDS project launched across the country till now, aimed to study driver behavior or, performance-related safety problems under real-world scenarios. The data acquisition systems (DASs) equipped on participated vehicles collect vehicle kinematics, roadway, traffic, environment, and driver behavior data continuously, which enable researchers to address such crash scenarios closely. This study integrated SHRP2 NDS and roadway information database (RID) data to conduct a comprehensive analysis of fixed-object crashes. A total of 1,639 crash, near-crash events relevant to fixed objects and animals, and 1,050 baseline events were used. Three analysis methods: logistic regression, support vector machine (SVM) and artificial neural network (ANN) were employed for two responses: crash occurrence and severity level. The logistic regression analyses identified 16 and 10 variables with significance levels of 0.1 for fixed-object event occurrence and severity level models respectively. The influence of explanatory variables was discussed in detail. SVM classifiers and ANN models were also constructed to predict the fixed-object crash occurrence and severity level. Sensitivity analyses were performed for SVM classifiers to infer the contributing effects of input variables. All three methods achieved satisfactory prediction accuracies of around 88% for crash occurrence prediction and 75% for crash severity level prediction, which suggested the effectiveness of NDS event data on depicting crash scenarios and roadway safety analyses.

Thesis (MEng)--Stellenbosch University, 2014. / ENGLISH ABSTRACT: Driving is a complex task that requires both the physical ability to drive a vehicle and the cognitive ability to do so safely. The ability to correctly integrate and apply information from the driving environment is essential for safe driving. In South Africa approximately 33 people per 100 000 population are killed annually in road accidents. Recent mortality data from South Africa has indicated that the age group 15 to 19 years old are the age group most likely to be involved in fatal vehicle crashes. Novice driver behaviour has been confirmed as problematic across the globe and extensive research into novice driver behaviour has been conducted to understand and ultimately to curb novice driver deaths. Very little is known about South African novice drivers. This lack of knowledge makes it difficult to plan for training, education or preparing young South African novice drivers for the challenges they are likely to face on the road. This study is a first stepping stone to understand this problem. This study utilises naturalistic driving studies as a method to explore differences between novice and experienced driver behaviour at a few preselected location types. Since 2005 naturalistic driving studies (NDS) have been employed extensively in the rest of the world and this study is South Africa’s first small attempt to employ this methodology and apply it to specifically novice driver behaviour. This thesis therefore not only explores novice driver behaviour in the context of South Africa, but also provides an overview of how the ND methodology can be developed for use in South Africa. The document provides an overview of both novice driver behaviour and naturalistic driving study methodologies from abroad. Where available, reference to South African research and reports are made. The literature review considers demographic, developmental and personality factors that could potentially (and have internationally been proven to) influence novice driver behaviour in the context of society, family and physical environments. Popular theories that have been applied to novice driver behaviour are reviewed. These theories include the Theory of Planned Behaviour, Social Learning Theory and the Theory of Intent. On the methodology side, the technology, its application as well as challenges and successes of the ND methodology are reviewed. The research process is described in terms of the participants and their risk attitudes to road traffic safety prior and after the study. The research process also details the specifications of the technology used, the data collected and the associated processes to make the data manageable. The research process took a number of unexpected turns which included the development of a coding scheme for the image material. Initially it was thought that this coding scheme should be predefined. However once the coding process commenced it was clear that in-vivo coding was necessary for inclusion of all elements of the environment and the behaviour. These elements differed from video to video and participant to participant. Grounded theory was introduced in an attempt to explain the novice behaviour. Although the data analysed was not extensive enough to substantiate the use of grounded theory it is considered useful in operationalizing this coding scheme in future. In addition to learning how to work with the data collection systems and how to integrate different types of quantitative and qualitative data in different formats, it also became clear that a strategy for managing large databases should be considered. This was an unexpected spin-off and is currently being investigated. The findings of the study showed that certain behaviours (such as the left scanning of a driving environment) were neglected not only by novice drivers but also by experienced drivers. Further investigations could include research into understanding this phenomenon. The preselected site types included stop streets, traffic lights, traffic circles and intersections. Traffic lights and intersections in particular have in recent years been highlighted as hazardous locations in Pretoria, where the study took place. Differences in behaviours were highlighted for intersections but not for traffic lights, stop streets or traffic circles. However the difference in the proportion of time that novice and experienced drivers took to scan their environments around these preselected hazardous locations differed significantly. Experienced drivers were much more thorough than their novice counterparts. This study was aimed at investigating the differences between novice and experienced drivers and aimed to develop recommendations that could potentially have implications for changing the driver training and education milieu in SA. However, the sample size (both participants and material selected for analysis) was too small to make meaningful recommendations towards change in this industry. It did however show clear differences between novice and experienced drivers, even in South Africa, and that this research needs to be expanded. The potential of this research for South Africa is enormous and could quite possibly, in future, change the way in which South Africans drive.

Introduction: Multiple studies have shown the symptoms of Parkinson's disease (PD) can impair driving performance. Studies have also found elevated crash rates in drivers with PD, however, none have controlled for exposure or amount of driving. Although a few studies have suggested that drivers with PD may self-regulate (e.g., by reducing exposure or avoiding challenging situations), findings were based on self-report data. Studies with healthy older drivers have shown that objective driving data is more accurate than self-estimates. Purposes: The primary objectives of this study were to examine whether drivers with PD restrict their driving (exposure and patterns) relative to an age-matched control group and explore possible reasons for such restrictions: trip purposes, perceptions of driving comfort and abilities, as well as depression, disease severity and symptoms associated with PD. Methods: A convenience sample of 27 drivers with PD (mean 71.6±6.6, range 57 to 82, 78% men) and 20 age-matched control drivers from the same region (70.6±7.9, range 57 to 84, 80% men) were assessed between October 2009 and August 2010. Driving data was collected for two weeks using two electronic devices (one with GPS) installed in each person‟s vehicle. Participants completed trip logs, questionnaires on background and usual driving habits, and measures of cognitive functioning, depression, quality of life, daytime sleepiness, driving comfort and abilities. Contrast sensitivity and brake response time were also assessed. Severity of PD was assessed using the Unified Parkinson‟s Disease Rating Scale (UPDRS) motor scores. An interview was conducted at the end of the second assessment to examine influence of the devices, driving problems and any departures from usual patterns over the monitoring period. Results: Of the 128 PD patients screened for possible study participation, 35% had already stopped driving. Former drivers were older, more likely to be women and had poorer UPDRS motor scores. Only 48% of those who were eligible for the study agreed to participate. Compared to controls, the PD group had significantly slower brake response times, higher depression and quality of life scores, less comfort driving at night and poorer perceptions of their driving abilities. The PD group also had significantly lower cognitive functioning scores than controls, and a significantly greater proportion (74% versus 45%) were classified as having mild cognitive impairment. Compared to vehicle recordings, both groups mis-estimated the amount they drove over two weeks (measurement error was 94 km for the PD group and 210 km for the controls). The PD group drove significantly less overall (days, trips, distance and duration), at night, on week-ends and in bad weather and for different purposes. Four of the PD drivers had minor accidents over the two weeks, while one lost his license. Conclusions: Self-estimates of exposure were inaccurate warranting the continued use of objective driving data. Overall, the findings suggest that drivers with PD appear to restrict their driving exposure and patterns relative to controls. The PD group were more likely to combine several activities into one trip, possibly due to fatigue. Moreover, they were more likely than controls to drive for medical appointments and less likely to drive for leisure activities and make out of town trips. The findings need to be replicated with larger samples and longer monitoring periods to examine changes in self-regulatory practices associated with disease progression and symptomatology. Other researchers are also likely to have similar difficulty in recruiting drivers with PD as this group may quit driving at an earlier age and those who are still driving are fearful of being reported to licensing authorities. Future studies also need to screen for cognitive impairment which often goes undetected, particularly in otherwise healthy drivers.

Introduction: Multiple studies have shown the symptoms of Parkinson's disease (PD) can impair driving performance. Studies have also found elevated crash rates in drivers with PD, however, none have controlled for exposure or amount of driving. Although a few studies have suggested that drivers with PD may self-regulate (e.g., by reducing exposure or avoiding challenging situations), findings were based on self-report data. Studies with healthy older drivers have shown that objective driving data is more accurate than self-estimates. Purposes: The primary objectives of this study were to examine whether drivers with PD restrict their driving (exposure and patterns) relative to an age-matched control group and explore possible reasons for such restrictions: trip purposes, perceptions of driving comfort and abilities, as well as depression, disease severity and symptoms associated with PD. Methods: A convenience sample of 27 drivers with PD (mean 71.6±6.6, range 57 to 82, 78% men) and 20 age-matched control drivers from the same region (70.6±7.9, range 57 to 84, 80% men) were assessed between October 2009 and August 2010. Driving data was collected for two weeks using two electronic devices (one with GPS) installed in each person‟s vehicle. Participants completed trip logs, questionnaires on background and usual driving habits, and measures of cognitive functioning, depression, quality of life, daytime sleepiness, driving comfort and abilities. Contrast sensitivity and brake response time were also assessed. Severity of PD was assessed using the Unified Parkinson‟s Disease Rating Scale (UPDRS) motor scores. An interview was conducted at the end of the second assessment to examine influence of the devices, driving problems and any departures from usual patterns over the monitoring period. Results: Of the 128 PD patients screened for possible study participation, 35% had already stopped driving. Former drivers were older, more likely to be women and had poorer UPDRS motor scores. Only 48% of those who were eligible for the study agreed to participate. Compared to controls, the PD group had significantly slower brake response times, higher depression and quality of life scores, less comfort driving at night and poorer perceptions of their driving abilities. The PD group also had significantly lower cognitive functioning scores than controls, and a significantly greater proportion (74% versus 45%) were classified as having mild cognitive impairment. Compared to vehicle recordings, both groups mis-estimated the amount they drove over two weeks (measurement error was 94 km for the PD group and 210 km for the controls). The PD group drove significantly less overall (days, trips, distance and duration), at night, on week-ends and in bad weather and for different purposes. Four of the PD drivers had minor accidents over the two weeks, while one lost his license. Conclusions: Self-estimates of exposure were inaccurate warranting the continued use of objective driving data. Overall, the findings suggest that drivers with PD appear to restrict their driving exposure and patterns relative to controls. The PD group were more likely to combine several activities into one trip, possibly due to fatigue. Moreover, they were more likely than controls to drive for medical appointments and less likely to drive for leisure activities and make out of town trips. The findings need to be replicated with larger samples and longer monitoring periods to examine changes in self-regulatory practices associated with disease progression and symptomatology. Other researchers are also likely to have similar difficulty in recruiting drivers with PD as this group may quit driving at an earlier age and those who are still driving are fearful of being reported to licensing authorities. Future studies also need to screen for cognitive impairment which often goes undetected, particularly in otherwise healthy drivers.

Collision avoidance systems (CASs) are being developed and fielded to reduce the number and severity of rear-end crashes. Kinematic algorithms within CASs evaluate sensor input and apply assumptions describing human-response timing and deceleration to determine when an alert should be presented. This dissertation presents an analytic assessment of dynamic function and performance CASs and associated driver performance for preventing automotive rear-end crashes. A method for using naturalistic data in the evaluation of CAS algorithms is described and applied to three algorithms. Time-series parametric data collected during 13 rear-end crashes and 70 near-crashes are input into models of collision avoidance algorithms to determine when the alerts would have occurred. Algorithm performance is measured by estimating how much of the driving population would be able to respond in the time available between when an alert would occur and when braking was needed. A sensitivity analysis was performed to consider the effect of alternative inputs into the assessment method. The algorithms were found to warn in sufficient time to permit 50–70% of the population to avoid collision in similar scenarios. However, the accuracy of this estimate was limited because the tested algorithms were found to alert too frequently to be feasible. The response of the assessment method was most sensitive to differences in assumed response-time distributions and assumed driver braking levels. Low-speed crashes were not addressed by two of the algorithms. Analysis of the events revealed that the necessary avoidance deceleration based on kinematics was generally less than 2 s in duration. At the time of driver response, the time remaining to avoid collision using a 0.5g average deceleration ranged from â 1.1 s to 2.1 s. In 10 of 13 crashes, no driver response deceleration was present. Mean deceleration for the 70 near-crashes was 0.37g and maximum was 0.72g. A set of the events was developed to measure driver response time. The mean driver response time was 0.7 s to begin braking and 1.1 s to reach maximum deceleration. Implications for collision countermeasures are considered, response-time results are compared to previous distributions and future work is discussed. / Ph. D.

Intersection crashes are among the most frequent and lethal crash modes in the United States. Intersection Advanced Driver Assistance Systems (I-ADAS) are an emerging active safety technology which aims to help drivers safely navigate through intersections. One primary function of I-ADAS is to detect oncoming vehicles and in the event of an imminent collision can (a) alert the driver and/or (b) autonomously evade the crash. Another function of I-ADAS may be to detect and prevent imminent traffic signal violations (i.e. running a red light or stop sign) earlier in the intersection approach, while the driver still has time to yield for the traffic control device. This dissertation evaluated the capacity of I-ADAS to prevent U.S. intersection crashes and mitigate associated injuries. I-ADAS was estimated to have the potential to prevent up to 64% of crashes and 79% of vehicles with a seriously injured driver. However, I-ADAS effectiveness was found to be highly dependent on driver behavior, system design, and intersection/roadway characteristics. To generate this result, several studies were performed. First, driver behavior at intersections was examined, including typical, non-crash intersection approach and traversal patterns, the acceleration patterns of drivers prior to real-world crashes, and the frequency, timing, and magnitude of any crash avoidance actions. Second, two large simulation case sets of intersection crashes were generated from U.S. national crash databases. Third, the developed simulation case sets were used to examine I-ADAS performance in real-world crash scenarios. This included examining the capacity of a stop sign violation detection algorithm, investigating the sensor detection needs of I-ADAS technology, and quantifying the proportion of crashes and seriously injuries that are potentially preventable by this crash avoidance technology. / Ph. D.

Abgelenktes Fahren gehört zu den Hauptursachen von Verkehrsunfällen und kann auf visuelle, manuelle oder kognitive Ablenkungsquellen zurückgeführt werden. Jede dieser Ablenkungsquellen wurde bereits mit negativen Effekten auf die Fahrerleistung in Zusammenhang gebracht. Obschon ein weitgehender Konsens über negative Auswirkungen von visueller/visuell-manueller Ablenkung besteht, sind die Wirkungen kognitiver Ablenkung auf Fahrfehler und Unfälle noch immer umstritten. Viele experimentelle Studien haben negative Auswirkungen kognitiver Ablenkung auf die Fahrerleistung berichtet. Demgegenüber stehen jedoch die Ergebnisse der Mehrzahl vorliegender „naturalistic driving studies“, die kein erhöhtes Unfallrisiko oder sogar protektive Effekte in diesem Zusammenhang fanden. Die aktuelle Entwicklung hin zu Mensch-Fahrzeug-Schnittstellen, die die Bedienung diverser Anwendungen mittels Sprachsteuerung ermöglichen, führt zu einem Anstieg von kognitiver Beanspruchung beim Fahren. Es ist daher von entscheidender Bedeutung, die Auswirkungen kognitiver Ablenkung auf die Fahrerleistung zu erfassen, um den Verantwortungsträgern in der Gesellschaft, den Regierungen und der Industrie eine Risikoabschätzung dieser Funktionen zu ermöglichen und die Sicherheit von Mensch-Fahrzeug-Schnittstellen zu erhöhen. Das Hauptziel dieser Dissertation bestand darin, die Effekte von kognitiver Ablenkung auf die Fahrerleistung zu untersuchen. Verschiedene Arten kognitiver Ablenkung, die sich beim Fahren unter realen Bedingungen häufig auf die Fahrer auswirken, wurden in dieser Arbeit kodiert und analysiert: kognitiv ablenkende Nebenaufgaben (z.B. telefonieren, singen), Fahreremotionen (z.B. Freude, Wut/Frustration, Traurigkeit) und Kombinationen von Fahreremotionen und Nebenaufgaben (z.B. Streit mit dem Beifahrer oder am Telefon). Bei der Untersuchung von Effekten kognitiver Ablenkung auf das Fahren sind Umwelt-, Situations- und Personenfaktoren zu berücksichtigen, da sie Mediator- und Moderatorvariablen bei der Erfassung des relativen Risikos von Ablenkung beim Fahren im Straßenverkehr darstellen. Daher folgte diese Dissertation dem ganzheitlichen Ansatz, so viele relevante Variablen wie möglich zu betrachten, die mit der Ausführung kognitiv ablenkender Tätigkeiten interagieren. Zu diesem Zweck wurden Daten der derzeit umfangreichsten „naturalistic driving study“ (the second Strategic Highway Research Program, SHRP 2) kodiert und analysiert, um möglichst viele Situationen, in denen eine kognitive Beanspruchung die Fahrerleistung potenziell beeinflusste, umfassend zu bewerten. Gleichzeitig wurde eine große Zahl von Mediator- und Moderatorvariablen betrachtet, die beim Fahren im realen Straßenverkehr auftreten (z.B. Einfluss von Kreuzungen, Wetter, etc.). Dieser Ansatz sollte das Verständnis und die externe Validität der Ergebnisse erhöhen und stellt einen wichtigen Schritt hin zu einem vollständigen Modell jener Variablen dar, die entweder zu unangemessen Verhaltensweisen und Unfällen beitragen oder sie reduzieren. Im Rahmen der Dissertation wurden vier Studien durchgeführt, die auf der Grundlage von zwei SHRP 2 Datensätzen die Zusammenhänge zwischen kognitiven und anderen Ablenkungsquellen, Umwelt-, Situations- und Personenfaktoren und Fahrerleistung untersuchten. Weiterhin wurden Kausalfaktoren in 315 vom Fahrer verursachten Unfällen und Beinaheunfällen, die mit Fahrerablenkung, Fahrerbeeinträchtigung oder keinem dieser Faktoren assoziiert waren, analysiert. Die erste Studie untersuchte die Auswirkungen von Wut beim Fahren und Streit mit dem Beifahrer oder jemandem am Telefon auf die Fahrerleistung. Wut beim Fahren ging mit einer Häufung aggressiver Verhaltensweisen einher, jedoch nicht mit einer Erhöhung von Fahrfehlern. Streitgespräche mit dem Beifahrer oder einer Person am Telefon (das heißt, wenn mutmaßlich das höchste Maß an kognitiver Ablenkung vorlag), schienen darüber hinaus mit keiner Form von unangemessenen Verhaltensweisen im Zusammenhang zu stehen. Die zweite Studie untersuchte, wie sich kognitive, visuelle und manuelle Fahrerablenkung, emotionale Beeinträchtigung sowie Umwelt-, Situations- und Persönlichkeitsfaktoren auf die Fahrerleistung auswirken. Ein Zusammenhang zwischen kognitiver Ablenkung und einer Verschlechterung der Fahrerleistung konnte nicht festgestellt werden. Die dritte Studie replizierte und erweiterte Ergebnisse der zweiten Untersuchung auf der Grundlage eines größeren Datensatzes, bestehend aus Fahrsegmenten, die Unfällen, Beinaheunfällen und Baselines vorausgingen und weder emotionale noch andere Fahrerbeeinträchtigungen enthielten. In Übereinstimmung mit den Ergebnissen der ersten und zweiten Studie, wurde keine Assoziation zwischen kognitiver Ablenkung und einer verschlechterten Fahrerleistung festgestellt. Bei der vierten Studie handelte es sich um eine vergleichende Analyse von Risikofaktoren für Unfälle/ Beinaheunfälle, die mit verschiedenen Arten von Ablenkung, Beeinträchtigung oder keinem von beiden, assoziiert waren. Unfälle, denen eine kognitive Ablenkung vorausgegangen war, waren vor allem mit von Ablenkung unabhängigen Fahrfehlern verbunden - genau wie die Unfälle, denen keine beobachtbare Nebentätigkeit vorausgegangen war. Dieses Ergebnis lässt vermuten, dass in früheren „naturalistic driving studies“, das Unfallrisiko von kognitiv ablenkenden Nebentätigkeiten eventuell sogar überschätzt wurde. Zusammenfassend legen die Ergebnisse die Schlussfolgerung nahe, dass kognitive Ablenkung durch beobachtbare emotionale Beeinträchtigung, (überwiegend) kognitiv ablenkende Nebenaufgaben oder die Kombination dieser beiden Faktoren, nicht mit sichtbaren negativen Auswirkungen auf die Fahrerleistung im tatsächlichen Straßenverkehr assoziiert werden kann. Im Gegensatz dazu hatten ablenkende Tätigkeiten, die zu Blickabwendungen von der Straße führen, und solche, die mit einem besonders hohen Unfallrisiko assoziiert werden, die größte Wahrscheinlichkeit Fahrfehler und Unfälle zu verursachen. / Driver distractions are among the leading causes of motor vehicle accidents. Such distractions can stem from competing visual, manual, or cognitive resources, all of which have been associated with detrimental effects on driving performance. Although the negative impacts of visual/visual-manual distraction are widely agreed upon, the effects of cognitive load on driving errors and crash risk are still debated. On the one hand, numerous experimental studies have shown adverse effects of cognitive distraction on driving performance. In contrast, most existing naturalistic driving studies have either not revealed increased crash/near-crash risk due to cognitive distraction, or have even reported a safety benefit. The number of in-vehicle tasks placing cognitive load on the driver is increasing in recent years due to the development of auditory human–machine interfaces such as voice control for several functions. This has enhanced the need to assess how cognitive distraction affects driving performance. These results are necessary to provide society, government, and industry with valid risk estimates, which will affect decision making regarding how to enhance the safety of using in-vehicle human-machine interfaces while driving. Therefore, the main objective of this thesis was to investigate how cognitive distraction affects driving performance. Different types of cognitive distraction that commonly affect most drivers in naturalistic conditions were coded and analyzed in the present thesis, including: cognitively distracting secondary tasks (e.g., talking on the phone, singing), driver emotion (e.g., happiness, anger/frustration, sadness), and combinations of driver emotion and secondary task demand (e.g., arguing with a passenger or with someone on the phone). Environmental, situational, and individual factors cannot be ignored when investigating the effects of cognitive distraction on driving performance, as they are mediating and moderating variables for estimating distraction relative risk in naturalistic driving. Therefore, a holistic approach guided this thesis towards incorporating as many important variables as possible that interact with the engagement in cognitively distracting activities. Data from the largest naturalistic driving study ever conducted (the second Strategic Highway Research Program, SHRP 2) were coded and analyzed to comprehensively assess many situations in which cognitive load potentially affected driving performance. Further, the goal was to simultaneously consider many possible mediating and moderating variables existent in real-world traffic (such as intersection influences, weather, etc.). This approach should increase understanding and external validity of the results, as well as represent an important step towards building a complete model depicting variables that contribute to or mitigate aberrant driving behaviors and crash risk. Four different analyses focused on two SHRP 2 data subsets to assess the relationship between cognitive and other distraction sources, environmental, situational, and individual factors, as well as driving performance. In addition, contributing factors in 315 at-fault crash and near-crash events associated with driver distraction, driver impairment, or neither of the two were analyzed. The first study examined driving performance in relation to driving anger as well as arguing with a passenger or with someone on the phone. Results showed that driving anger was associated with more frequent aggressive driving behaviors without increasing driving error frequency. Furthermore, when a conflict arose with a passenger or with someone on the phone (i.e., when the level of cognitive distraction was expected to be highest), there did not appear to be a link to any type of aberrant driving behavior. The second study analyzed driving performance based on cognitive, visual, and manual driver distraction, emotional impairment, as well as environmental, situational, and individual factors. Cognitive distraction was not associated with any decline in driving performance. The purpose of the third analysis was to replicate and extend the second study’s effects based on a larger data sample of driving segments preceding crashes, near-crashes, and matched baselines, of drivers not exhibiting emotional or other impairment types. Corroborating the first and second study’s results, there was no association between cognitive distractions and impaired driving performance. Finally, the fourth study compared the risk factors of crashes/near-crashes associated with either different driver distraction types, impairment, or neither. Crashes preceded by cognitive distraction were mainly associated with driving errors unrelated to the secondary task demands, as were the crashes preceded by no observable secondary task. This finding suggests that previous studies analyzing naturalistic driving data may have even overestimated the crash risk of cognitively distracting secondary task engagement. In summary, this thesis provides compelling evidence that cognitive distraction, either through observable emotional impairment, (mainly) cognitively distracting secondary tasks, or the combination of both, has no apparent relation with poorer driving performance observable in real-world traffic. On the contrary, distracting activities requiring the driver’s gaze to move away from the forward roadway and those associated with a particularly high crash risk had the highest chances of causing driving errors and crashes.

This dissertation aims to develop statistical methodologies for estimating the effects of time-fixed and time-varying factors in recurrent events modeling context. The research is motivated by the traffic safety research question of evaluating the influence of crash on driving risk and driver behavior. The methodologies developed, however, are general and can be applied to other fields. Four alternative approaches based on various data settings are elaborated and applied to 100-Car Naturalistic Driving Study in the following Chapters. Chapter 1 provides a general introduction and background of each method, with a sketch of 100-Car Naturalistic Driving Study. In Chapter 2, I assessed the impact of crash on driving behavior by comparing the frequency of distraction events in per-defined windows. A count-based approach based on mixed-effect binomial regression models was used. In Chapter 3, I introduced intensity-based recurrent event models by treating number of Safety Critical Incidents and Near Crash over time as a counting process. Recurrent event models fit the natural generation scheme of the data in this study. Four semi-parametric models are explored: Andersen-Gill model, Andersen-Gill model with stratified baseline functions, frailty model, and frailty model with stratified baseline functions. I derived model estimation procedure and and conducted model comparison via simulation and application. The recurrent event models in Chapter 3 are all based on proportional assumption, where effects are constant. However, the change of effects over time is often of primary interest. In Chapter 4, I developed time-varying coefficient model using penalized B-spline function to approximate varying coefficients. Shared frailty terms was used to incorporate correlation within subjects. Inference and statistical test are also provided. Frailty representation was proposed to link time-varying coefficient model with regular frailty model. In Chapter 5, I further extended framework to accommodate multi-type recurrent events with time-varying coefficient. Two types of recurrent-event models were developed. These models incorporate correlation among intensity functions from different type of events by correlated frailty terms. Chapter 6 gives a general review on the contributions of this dissertation and discussion of future research directions. / Ph. D.

Erkenntnisse über das Fahrerleben, also die subjektive Sicht von Fahrern auf Ereignisse im Fahralltag, wurden bisher vor allem retrospektiv gewonnen und sind damit anfällig für mitunter weitreichende Verfälschungen. Diese Dissertation verfolgte den Ansatz, das subjektive Fahrerleben auf alltäglichen Fahrten so situationsnah wie möglich zu erfassen. Ziel der Untersuchung war es, eine breite Datenbasis von subjektiv beanspruchenden Situationen für jüngere und ältere Fahrer zu generieren, um auch unbewusst vorhandene Unterstützungsbedürfnisse zu erfassen und damit eine bedarfsgerechte Entwicklung von Fahrerassistenzsystemen zu unterstützen. Es wurden 40 jüngere (M = 32,35 Jahre; SD = 3,58) und 40 ältere (M = 66,05 Jahre; SD = 4,13) Fahrer, je 20 Männer und Frauen, in ihrem Fahralltag über einen Zeitraum von jeweils 10 Tagen untersucht. Die Probanden hielten für sie relevante Situationen in kurzen Sprachprotokollen während der Fahrt über die Aufnahmefunktion in einem Smartphone fest. Unterstützt wurde die Situationsbetrachtung durch eine Videoaufnahme der Fahrsituation, sowie im Smartphone erfasste Geschwindigkeits-, Beschleunigungsdaten und GPS. Im Versuchszeitraum wurden insgesamt 1074 für die Auswertung relevante Sprachprotokolle während der Fahrt aufgezeichnet. Es ließen sich dabei 301 verschiedene Auslöser für die Aufnahme eines Sprachprotokolls unterscheiden. Ausgehend von den Ergebnissen und den geäußerten Unterstützungswünschen der Probanden wurden fünf verschiedene Unterstützungsbedürfnisse identifiziert, aus denen sich Anforderungen für Fahrerassistenzsysteme und Mensch-Maschine-Schnittstellen ableiten lassen. / Insights into driver experience, i.e. the driver's subjective view on events in everyday driving, have so far been gained mainly retrospectively (e.g. through interviews or online surveys). From a methodic perspective this means that reports and judgment are provided somewhat later after the event and therefore can be biased. In recent years, research in the field of driver-vehicle interaction has increasingly been enriched by natural driving studies (NDS). Since this method captures driving behavior in the natural driving context, it provides very realistic insights into events that drivers experience on a daily basis. So far, however, the focus has been on an objective view of driving behavior and environmental conditions with the aim of capturing parameters that provide an indication of future safety-critical situations. The subjective view of the driver, e.g. which situations he assesses as demanding, was hardly considered. The idea of the dissertation thesis was to capture subjective driver experience and support wishes of drivers in a NDS-like study. The aim of the study was to generate a broad database of subjectively demanding situations for younger and elder drivers, in order to detect unconsciously existing support needs and thus to support the user-centered development of driver assistance systems. Participants were 40 younger (M = 32.35 years, SD = 3.58) and 40 elder (M = 66.05 years, SD = 4.13) drivers, 20 men and 20 women each. Over the period of 10 days they documented each journey by questionnaires before and after each ride. Furthermore a smartphone with a specially programmed application was used as a recording device. Through short speech protocols the participants commented on every relevant driving situation. Considered as relevant - beyond critical events - were all special incidents or situations that were notable for the drivers or in which they wished for support in whatever form. The application also recorded GPS, speed and acceleration data as well as a video of the driving situation during relevant situations. A personal interview took place at the end of each trial period. As a result, a total of 1074 speech protocols were recorded while driving. They showed a very high range of situations that were notable for the participants. A total of 301 different triggers for recording a speech protocol could be distinguished. In addition, the consideration of the verbalized reaction of the drivers in the situations was important. The results were examined with regard to possible differences in age and gender groups. On the basis of the results and the expressed support wishes of the participants, five different support needs were identified, from which requirements for future driver assistance systems and human-machine interfaces can be derived.