Driving Behavior Analysis (DBA) plays a pivotal role in designing intelligent transportation systems, enhancing road safety, and advancing Autonomous Vehicles (AVs). Driver identification, as a key aspect of DBA, has the potential to provide unprecedented opportunities for enhanced security and driver profiling. However, the current solutions for driver identification suffer from demanding extensive data collection, limited scalability, and inadequate generalization. Furthermore, DBA is also essential for training AVs, addressing the main challenges they face: accurately perceiving their surroundings to make informed decisions and to navigate safely, and effectively handling unforeseen scenarios. In the first part of this thesis, we concentrate on behavior analysis for driver identification and verification and design two novel schemes aiming to reduce data dependency and enhance the generalization ability of existing approaches. First, we propose a novel driver identification model, called DriverRep, which reduces data dependency by presenting a fully unsupervised triplet loss training. DriverRep is the first model that extracts the latent representations associated with each driver, called driver embeddings, in an unsupervised manner. In addition, we develop a novel model to tackle driver verification and impostor detection tasks based on DBA and extracted driver embeddings. In the second part, we focus on behavior prediction for AVs and their surrounding agents. First, we tackle behavior prediction in dynamic and complex scenarios by introducing three novel prediction models for forecasting drivers intentions and behaviors at unsignalized intersections. We then address social reasoning by proposing a novel prediction model that analyzes agent interactions using graph neural networks, making the scene understanding process more informative for AVs. Our proposed prediction model, called STAG, explicitly activates social modeling with a directed graph representation while considering spatial and temporal inter-agent correlations. We further design a novel prediction system, namely CAPHA, which conditions the future behavior of agents on grid-based plans modeled as a Markov decision process and solves the prediction task via inverse reinforcement learning to produce scene compliant behaviors. Moreover, we introduce a novel goal-based prediction model, called GMP, which encodes interactions between agents and dynamic and static context information to estimate the distribution of target goals, efficiently considering the inherent uncertainty in agents behavior. Extensive quantitative and qualitative comparisons have been conducted between the developed solutions and related benchmark schemes using various scenarios and environments. The obtained results demonstrate the potential of these solutions for the understudy tasks of DBA and real-world applications.

No description available.

No description available.

No description available.

Efficient memory access is critical for computing systems, yet the CPU's management of data transfers can create bottlenecks. To counter this, most advanced high-throughput systems utilize Direct Memory Access (DMA) controllers, where peripherals (such as network interfaces and USB devices) can access memory independently of the CPU, improving transfer speeds. However, this bypass also introduces security vulnerabilities if the DMA controller is not configured correctly, as DMA devices may be used to overwrite critical data or leak information. This thesis proposes a method to represent complex DMA driver source code as an abstract mathematical model in the formal analysis tool HOL4 (where users can define models and prove properties about them with HOL4 and checking the correctness of the proofs). This model enables the formal verification of the DMA driver source code's critical properties like memory isolation, initial configurations, and many more. Additionally, the thesis introduces a methodology to convert the verified HOL4 models into executable C source code, thus obtaining a formally verified C source code. The synthesized code is evaluated against the original driver source code by emulating the DMA operation in software and using fuzzing techniques for any compile and runtime errors. This validates the approach, demonstrating that converting a C driver source code into a HOL4 model and then back into C source code after verification yields a formally verified C source code. This thesis applies this methodology to DMA controllers for four devices namely Intel 8237a, Intel IXGBE x550 Ethernet Controller, MPC 5200 SoC, and STM32 DMAC. / Master of Science / This thesis addresses the critical issue of ensuring secure memory access in computing systems, focusing on Direct Memory Access (DMA) controllers. DMA devices can bypass the CPU to access a range of memory directly, enhancing transfer speeds but introducing security vulnerabilities like overwriting or leaking critical data if not configured correctly. This thesis proposes a method to model complex DMA driver source codes such that they can be rigorously analyzed with computer assistance. This approach is significant as it provides a structured methodology for analyzing DMA driver source code, reducing the risk of errors and vulnerabilities. The thesis also proposes a method to convert the abstract representation into executable source code, thus improving the reliability and security of DMA operations in computing systems.

An ergonomic driver workspace and interface design is essential to ensure a healthier and comfortable driving experience in terms of driver perceptions, postures, and interface pressures. Developing more effective methods for driver-side interior design and evaluation, hence, requires thorough investigation of: 1) which perceptual responses are more relevant to ensuring ergonomic quality of a design, 2) the interrelationships among perceptual responses and objective measures, and 3) whether current assumptions regarding driver behaviors, and tools for specifying these behaviors, are valid for the design and evaluation. Existing studies, however, have rarely addressed these topics comprehensively, and often have been conducted with unsubstantiated assumptions. In contrast, this work sought to address these topics in a way that jointly considers characteristics of driver perceptions, behaviors, and objective measures to develop an improved design and evaluation methodology for driver workspace and interface, and that can also investigate the validity of implicit assumptions regarding perceptual relevance and drivers' behaviors. The first part of this work investigated drivers' perceptions in relation to driver workspace design and evaluation. Specifically, it examined the efficacy of several perceptual ratings, when used for evaluating automobile interface design. Results showed that comfort ratings were more effective at distinguishing among interface designs, in contrast to the current common practice of using discomfort ratings for designing and evaluating interface designs. Two distinct decision processes to relate local to global perceptions were also identified (i.e., global comfort as an average of local comforts, and global discomfort predominantly influenced by maximal local discomforts). These findings were observed consistently across age and cultural groups. In addition, this work provided empirical support for an earlier hypothetical comfort/discomfort model, which posited comfort and discomfort are complementary, yet independent entities. In order to facilitate the integration of driver perceptions and dynamic behaviors into driver workspace design and evaluation, the second part of this work clarified the relationships between perceptual ratings and various types of driver-seat interface pressure. Interface pressure was found to be more strongly related to overall and comfort ratings than to discomfort ratings, which is also in marked contrast with existing work that has focused on identifying association between discomfort and interface pressure. Specific pressure interface requirements for comfortable driver workspace design and evaluation were also provided. Lastly, this work specified more rigorous driving postures for digital human models (DHMs), based on actual drivers' perceptions, postural sensitivity, and static behavioral characteristics, to facilitate proactive design and evaluation that enables cost/time efficient vehicle development. Drivers' behavioral characteristics observed in this work were applied to the driver workspace design. First, postural sensitivity obtained by using a psychophysics concept has been applied to determination of core seat track ranges. Second, postural data have been used: 1) to review relevant industry standards on driver accommodation, 2) to investigate whether driving postures are bilaterally asymmetric, 3) to provide comfortable joint ranges, and lastly 4) to identify drivers' postural strategies for interacting with a vehicle. Overall, this work identified three important behavioral characteristics, specifically a bilateral imbalance in terms of interface pressure, bilaterally asymmetric joint posture, and postural strategies identified by cluster analysis. Such characteristics can be embedded in DHMs to describe more accurately actual driver behaviors inside a driver workspace, which is deemed to be a fundamental step to improved virtual ergonomic vehicle design and evaluation. In addition, the strategy-based classification method used in this work can be extended to simulate and predict more complex human motions. Practical and fundamental findings of this work will facilitate efficient and proactive design and evaluation of driver workspace and interface, and will help provide a healthier driving experience for a broader range of individuals. / Ph. D.

The purpose of this research was to explore traffic event severity relationships, evaluate the potentiality of a hazardous event, and develop a framework of observable event factors. Data was collected from three regions in Virginia, each assumed to exemplify a unique driving environment due to amount of traffic and infrastructure characteristics. In combination, a broad spectrum of site, traffic, and driver performance variables were accounted for. Observational techniques of surveillance, incident reporting, and inventorying were used to collect site, traffic, and driver data. This effort resulted in 368 observed traffic events that were evenly distributed among the three regions that represented metropolitan, mid-sized city, and town/rural driving environments. The 368 events were evaluated for severity and contributing variables where 1% of the events were non-injury crashes, 10% were serious, near-crashes, 24% were near-crashes, and the remaining 65% were serious errors with a hazard present. Exploratory analyses were performed to understand the general relationship between event severity levels. Binary logistic regression analyses (α = 0.05) were performed to further scope predictor variables to identify traffic event characteristics with respect to severity level, maneuver type, and conflict type. The results were that 69 of 162 observed predictor variables were valuable in characterizing traffic events based on severity. It was found that variables could be grouped to create event severity signatures for crashes, serious near-crashes, and near-crashes. Based on these signatures, it was found that there is a trend between severity levels that included a propensity for problems with straight path maneuvers, lateral and longitudinal vehicle control, and information density within the driving environment as contributing to driver error and hence crashes and near-crashes. There were also differences between the severity levels. These differences were evident in the degree of control the driver appeared to have of the vehicle, type of control regulating the driving environment, and type of road users present in the driving environment. Modifications to roadway evaluative techniques would increase awareness of additional variables that impact drivers to make more informed decisions for roadway enhancements. / Ph. D.

Researchers have postulated that the implementation of driving automation systems could reduce the prevalence of driver errors, or at least mitigate the severity of their consequences. While driving automation systems are becoming increasingly common on new vehicles, drivers seem to know very little about them. The following dissertation describes an investigation of driver behavior and behavioral adaptation while using driving automation systems in order to improve consumer education and training. This dissertation uses data collected from test track environments and two naturalistic driving studies, the Virginia Connected Corridor 50 (VCC50) Vehicle Naturalistic Driving Study and the NHTSA Level 2 Naturalistic Driving Study (L2 NDS), to investigate driver behavior with driving automation systems and make suggestions for modifications to current consumer education practices. Results from the test track study indicated that while training strategy elicited limited differences in knowledge and no difference in driver behaviors or attitudes, operator behaviors and attitudes were heavily influenced by time and experience with the driving automation. The naturalistic assessment of VCC50 data showed that drivers tended to activate systems more frequently in appropriate roadway environments. However, drivers spent more time looking away from the road while driving automation systems were active and drivers were more likely be observed browsing on their cell phones while using driving automation systems. The analysis of L2 NDS showed that drivers' time gap preferences changes as drivers gain experience using the driving automation systems. Additionally, driver eye glance behavior was significantly different with automation use and indicated the potential for an adaptive trend with increased exposure to the system for both glances away from the roadway and glances to the instrument panel. The penultimate chapter of this work presents training guidelines and recommendations for consumer education with driving automation systems based on this and other research that has been conducted on driver interaction with driving automation systems. The results of this research indicate that driver training should be a key focus in future efforts to ensure the continued safe use of driving automation systems as they continue to emerge in the vehicle fleet. / Doctor of Philosophy / While driving automation systems are becoming increasingly common on new vehicles, drivers seem to know very little about them. Previous studies have found that owners of vehicles equipped with advanced technologies have demonstrated misperceptions or lack of awareness about system limitations, which may impact driver comfort with and reliance on these systems. Partial driving automation systems are designed to assist drivers in some vehicle operation demands, they are not, however, designed to completely remove the driver from the driving task. The following dissertation describes an investigation of driver behavioral adaptation while using driving automation systems with the goal of improving consumer education and training.

With the development and use of Silicon-Carbide [Silicon-Carbide (SiC)] devices come a host of advantages, including higher switching frequency, improved thermal performance, and higher voltage rating. This higher switching frequency can reduce the size of the con- verter system, but is typically associated with higher dv/dt voltage slew rates that further increase electromagnetic interference (EMI) related phenomena. Conventional gate-drivers are very limited in the way that they can control this high dv/dt, and this leads to the use of active gate-drivers. This thesis will explore the use of an active voltage-controlled gate-driver for SiC devices, utilizing transiently a voltage closer to the Miller plateau than the nominal turn-on and turn-off voltage to introduce control over the switching transient. Various ap- plied voltages, and voltage sequences will be evaluated to determine their effectiveness for controlling dv/dt and their impact on switching loss. Through this work, a better under- standing of the advantages and drawbacks of an active gate-driver can be found. The main result from this work is the effective reduction in the dv/dt generated by MOSFET devices, which was attained at a lower switching loss penalty compared to conventional resistive gate-drivers operating at similar dv/dt rates. Simulation and experimental results obtained with a prototype active gate-driver circuitry were used for this evaluation. / Master of Science / Within power electronic systems such as an inverter used to connect solar panels to the grid, are electrically controlled switches. These switches traditionally have been made of Silicon (Si) which imposed limitations on how fast they could transition from off to on, and vice versa, they also could only switch a relatively small number of times per second. However, a new generation of devices made from a silicon carbide material are being increasingly adopted, some key advantages of these new devices include much higher number of times to switch per second, and faster transitions from off-on and on-off. The trade-off that comes with this faster operation is an increase in the electromagnetic noise generated by these switches, among other issues. This work looks to explore a more unique method of controlling the turn-on and turn-off of these new switches and evaluating its impact on the noise generated and the losses during switching.

Driver heterogeneity in travel behavior has repeatedly been cited in the literature as a limitation that needs to be addressed. In this work, driver heterogeneity is addressed from four different perspectives. First, driver heterogeneity is addressed by models of driver perceptions of travel conditions: travel distance, time, and speed. Second, it is addressed from the perspective of driver learning trends and models of driver-types. Driver type is not commonly used in the vernacular of transportation engineering. It is a term that was developed in this work to reflect driver aggressiveness in route switching behavior. It may be interpreted as analogous to the commonly known personality-types, but applied to driver behavior. Third, driver heterogeneity is addressed via latent class choice models. Last, personality traits were found significant in all estimated models. The first three adopted perspectives were modeled as functions of variables of driver demographics, personality traits, and choice situation characteristics. The work is based on three datasets: a driving simulator experiment, an in situ driving experiment in real-world conditions, and a naturalistic real-life driving experiment. In total, the results are based on three experiments, 109 drivers, 74 route choice situations, and 8,644 route choices. It is assuring that results from all three experiments were found to be highly consistent. Discrepancies between predictions of network-oriented traffic assignment models and observed route choice percentages were identified and incorporating variables of driver heterogeneity were found to improve route choice model performance. Variables from all three groups: driver demographics, personality traits, and choice situation characteristics, were found significant in all considered models for driver heterogeneity. However, it is extremely interesting that all five variables of driver personality traits were found to be, in general, as significant as, and frequently more significant than, variables of trip characteristics — such as travel time. Neuroticism, extraversion and conscientiousness were found to increase route switching behavior, and openness to experience and agreeable were found to decrease route switching behavior. In addition, as expected, travel time was found to be highly significant in the models that were developed. However, unexpectedly, travel speed was also found to be highly significant, and travel distance was not as significant as expected. Results of this work are highly promising for the future of understanding and modeling of heterogeneity of human travel behavior, as well as for identifying target markets and the future of intelligent transportation systems. / Ph. D.