Doctoral Dissertations

Permanent URI for this collection

Browse

Recent Submissions

Now showing 1 - 20 of 17948
  • Using Automated Gaze-Based Feedback to Enhance Motor Skill Acquisition in Laparoscopic Surgery Training
    Deng, Shiyu (Virginia Tech, 2025-10-20)
    Laparoscopic surgery is now considered the standard of care for many abdominal procedures, offering benefits such as reduced postoperative pain, faster recovery, and shorter hospital stays. However, this type of surgery requires a distinct set of psychomotor skills that are more technically challenging than those required for open surgery, due to factors such as the separation of visual and operative fields, limited depth perception, restricted instrument movement, and motion inversion caused by the fulcrum effect. Thus, ensuring that surgical trainees acquire these skills effectively and efficiently is critical for patient safety and optimal surgical outcomes. The first stage of laparoscopic training currently relies on self-directed practice, which can be inefficient because trainees may struggle to identify and adopt strategies that effectively improve performance. To address this challenge, this research leverages eye tracking and computational techniques to examine the role of visual attention in laparoscopic skill acquisition and to evaluate personalized, gaze-based feedback for accelerating laparoscopic skill acquisition. The first study provided a foundational understanding of which eye metrics are most sensitive for early laparoscopic skill evaluation and how they reflect cognitive processes underlying performance differences. Results showed that scene-dependent eye metrics, developed relative to specific areas of interest in the visual scene, were more sensitive and more strongly correlated with performance and motion measures than scene- independent metrics, highlighting their value for assessing early skill development. Building on these findings, the second study employed scene-dependent eye metrics to track changes in visual attention as novices developed technical proficiency and to compare gaze behaviors between fast and slow learners. The results showed that better performance and faster learning were linked to more efficient gaze behaviors, laying the groundwork for the final study. The final study employed a custom software application to investigate timely gaze-based feedback on skill acquisition. Unlike prior gaze-training approaches that relied on human instructors for continuous verbal feedback, this application analyzed eye gaze data, computed eye metrics for specific gaze behaviors, compared them with benchmark values, and provided actionable feedback. Results demonstrated that the system could improve laparoscopic skill acquisition, particularly during early training stages. Taken together, these studies demonstrate the feasibility and promise of using automated eye-tracking feedback to enhance laparoscopic skill evaluation and training. By integrating advanced eye tracking with AI-driven video and data processing, this work represents a critical first step toward developing scalable, cost-effective virtual coaching systems capable of personalizing formative feedback, accelerating psychomotor skill acquisition, and optimizing surgical training outcomes.
  • Harnessing Artificial Intelligence to Guide Exoskeleton Adoption in Construction
    Okunola, Akinwale Samuel (Virginia Tech, 2025-10-20)
    The construction industry is a major contributor to the Gross Domestic Product in the United States, yet it continues to experience persistent health and safety challenges from work-related musculoskeletal disorders. These disorders, arising from physically demanding and repetitive tasks, result in muscle fatigue, diminished work capacity, increased lost workdays, long-term disability, and contribute to the shortage of skilled labor. The back is one of the most affected body regions. While efforts have been made to mitigate these disorders through administrative, engineering, and training measures, as well as emerging technologies such as computer vision and wearable sensors, exoskeletons have emerged as a human-centered solution that could extend the longevity and sustainability of the construction workforce. Specifically, active back-support exoskeletons have been identified as a potential solution to the prevalence of back-related musculoskeletal disorders. However, limited evidence exists on the criteria for appropriate selection, adoption and integration of active back-support exoskeleton technology in construction practice. Guided by the Technology-Organization-Environment framework, this research investigates the potential of an Artificial Intelligence-enabled decision support system to assist stakeholders in selecting suitable active back-support exoskeletons for construction tasks. The research followed a multi-stage design. First, facilitators and barriers to adoption were identified using the Delphi technique and semi-structured interviews with construction stakeholders. Next, a laboratory study assessed the physical, physiological, and psychological impacts of active back-support exoskeleton use simulated construction tasks, drawing on biofeedback sensors and subjective evaluations. Building on these insights, a data-driven analytical decision-support system was developed by integrating large language models with digital twin technologies. This system was subsequently evaluated for usability, organizational fit, and environmental compatibility. This research lies at the intersection of construction ergonomics, wearable robotics, and intelligent decision systems. It contributes to an emerging interdisciplinary field by translating biomechanical evidence and Artificial Intelligence-driven analytics into practical decision-support tools aimed at reducing injury risk and enhancing workforce sustainability. Accordingly, this research introduces a human-centered Artificial Intelligence framework for exoskeleton selection in construction, strengthening stakeholders' capacity to make informed adoption decisions.
  • Modal Analysis of Axisymmetric Structures using Zernike  Polynomials and Machine Learning
    Parthasarathy, Sudharsan (Virginia Tech, 2025-10-20)
    The rise of electric mobility has amplified the need for advanced vibration analysis to control noise in both electric cars and aircraft. In vehicles, tire-induced vibrations have become a significant contributor to cabin noise, making an understanding of tire mode shapes crucial for effective vibration mitigation. Likewise, lightweight stiffened panels in electric aircraft demand careful vibration control to ensure passenger comfort. Addressing these challenges calls for innovative approaches not only to interpret complex vibration patterns but also to streamline the analysis process. In the first part, ML-based classification frameworks are developed for categorizing tire mode shapes, aiming to automate the traditionally manual and labor-intensive process. Leveraging Zernike Annular Moment Descriptors (ZAMD) as feature maps, supervised learning models, such as decision trees, random forests, and XGBoost, achieve a high classification accuracy, thus eliminating the need for manual intervention. Furthermore, convolutional neural networks (CNNs), trained on physics-informed modal displacement data from finite element analyses, are employed to classify tire mode shapes for both unloaded and loaded cases. The CNN-based approach, enhanced with transfer learning techniques, also achieves high classification accuracy, validating its effectiveness across different tire conditions. The second part of the thesis focuses on advancing vibration analysis methods for stiffened circular plates. The Ritz method, utilizing Zernike and Legendre polynomials as trial functions, is implemented to enable free-vibration analysis without the meshing constraints typically associated with traditional finite element methods. This approach allows arbitrary stiffener placement while maintaining computational efficiency and accuracy, particularly for higher-order modes. To address the limitations of the Ritz method in optimization studies, where large, fully populated matrices pose computational challenges, a graph neural network (GNN) model is proposed. The GNN, designed with edge-aware message passing, predicts the first natural frequency and corresponding Ritz constants for varying stiffener configurations, achieving low mean absolute errors on the test dataset. By integrating classical mathematical methods with modern machine learning techniques, this work presents a comprehensive framework for analyzing and interpreting free-vibration behavior in complex structural systems.
  • Towards Logical Reasoning and Learning in Open and Dynamic Environments
    Alotaibi, Fatimah Dhaifallah (Virginia Tech, 2025-10-20)
    We live in an increasingly open and dynamic world where knowledge is constantly evolving, and Artificial Intelligence (AI) systems must adapt to newly added information. A crucial aspect of AI systems is performing robust logical reasoning that makes reliable inferences, generates hypotheses, and extracts meaningful insights from vast and complex data. Logical reasoning is fundamental in high-impact applications such as medical diagnosis, autonomous driving, and scientific discovery. However, traditional AI models, designed under static assumptions, struggle with reasoning in open and dynamic environments. They fail to generalize beyond their training data, leading to unreliable conclusions when encountering novel entities, relationships, unforeseen scenarios, or incomplete knowledge. This limitation poses a significant barrier to these high-impact applications. For example, in biomedical research, knowledge graphs often contain thousands of entities and relationships representing gene-disease-compound interactions. While many of these relationships are well-established, new ones emerge as scientific knowledge evolves. This highlights the need for robust logical reasoning in open environments to handle distributional shifts, complex logical inferences, and knowledge discovery. As a result, a critical research question arises: "How can AI systems effectively reason in dynamic environments where knowledge is constantly evolving and uncertainty is inherent?" To tackle this challenge, this thesis introduces a comprehensive framework for enhancing logical reasoning in open and dynamic environments. The framework is structured around three core components (Aim1), Out-of-Distribution (OOD) Logical Reasoning, which develops techniques to enable AI models to generalize beyond their training data, especially in knowledge graphs where new queries and entities frequently arise. This involves characterizing uncertainty and distributional shifts, thereby identifying novel, incomplete, and uncertain data for logical reasoning tasks, ultimately enhancing the robustness of logical reasoning models. (Aim2) Graph-Augmented Logical Reasoning integrates symbolic logic with graph-based representations to enhance LLM's reasoning accuracy, interpretability, and robustness, addressing hallucinations and ambiguous inference. (Aim3) Applications: This aim highlights the real-world applications of open and dynamic logical reasoning, including Otrouha—an automated system for knowledge discovery in Arabic Electronic Theses and Dissertations (ETDs)—and a multi-agent framework for dynamic category discovery. It also includes a multi-agent framework for scientific hypothesis generation that integrates semantic processing and symbolic reasoning to operate over structured metadata, enabling the discovery of emerging research directions in evolving domains without requiring access to full text. By addressing the pressing challenges of evolving data and the inherent uncertainty in AI in these three aims, our framework unites complementary approaches that collectively drive robust logical reasoning. The Characterization of uncertainty and OOD in (Aim1) provides insights for(Aim2) to manage complexity, ambiguity, and unseen logical tasks. Moreover, (Aim2) utilizes symbolic logic, graph-based methods, and prompt engineering to demonstrate that these strategies can handle challenging and unseen complex logical reasoning. Building on this knowledge from (Aim1) and (Aim2), the proposed novel approaches can be deployed in real-world applications in (Aim3), yielding tangible impact. Ultimately, the synergy among these three aims forms a unified framework that robustly advances logical reasoning in open and dynamic environments, enabling AI to adapt, generalize, and support real-world knowledge discovery and decision-making.
  • Neural Dynamics of Mental Imagery, Visual Perception, and Rapid Eye Movement Sleep
    Vess, Gavin Alexander (Virginia Tech, 2025-10-17)
    In order to better understand the functional role of rapid eye movement (REM) sleep, we sought to gain a deeper understanding of the differences between neural activity during REM sleep and during the awake state. We set out to investigate the temporal directionality of communication and coupling in neural activity between different areas of the brain during REM sleep and awake activities. One theory suggests that dreaming, which occurs predominantly during REM sleep, may be a mechanism for humans to incorporate information learned during the day, reflecting the memory consolidation function of sleep. Both mental imagery and dreaming are internally generated percepts while sensory processing is more externally generated, though similar neural regions are utilized. The difference in information flow between REM sleep, mental imagery, and stimulus perception may help us understand which regions of the brain and neural processes are key for the functional role REM sleep may serve. We conducted three studies comparing the oscillatory and topographical characteristics of REM sleep, visual perception, and mental imagery in an effort to help illuminate how REM sleep processes memories. Participants with no history of neurological disorders provided electroencephalography (EEG) data, while other participants provided intracranial data with electrodes surgically implanted as part of their epilepsy treatment plan. Both sets of human participants were monitored during visual stimulus processing, imagery, and REM sleep. The visual stimuli involved clock angles. The imagery task involved imagining two clock times and comparing their angles after an auditory stimulus. Brain activity during sleep was recorded during an overnight stay. Additionally, subjects performed a video imagery task where visual and auditory perception and imagery were tested by watching a video, then being asked to imagine the visuals or audio of that video. Intracranial participants provided access to data from internal structures of the brain and localized results, such as low frequency activity observed in the hippocampus during REM sleep. Conventional EEG participants provided access to data giving a better distributed image of the entire brain. Results from 20 EEG participants showed clear differences in the spectral content of certain regions of the brain when comparing the average power and coherence across the three conditions (visual stimuli, imagery, and REM sleep). Consistent for both imagery task paradigms, more power was observed frontally and centrally in the delta and theta frequency bands in REM sleep compared to perception and imagery, while both visual perception and imagery had higher power than REM sleep in most channels apart from the central midline channels in the beta and gamma frequency ranges, and more coherence occipitally and parietally in the gamma frequency band compared to REM sleep. Beyond a better understanding of the neural dynamics underlying mental imagery, visual perception, and REM sleep, these results may help in the construction of better brain machine interface algorithms and provide insight into diseases associated with REM sleep problems, such as Parkinson's disease, narcolepsy, and depression. In addition, we used sleep as a window into neurological disorders. In particular, we utilized high-density EEG polysomnography in Parkinson's disease (PD) patients to help reveal not only atypical REM sleep, but also disrupted Non-REM (NREM) sleep architecture, including reduced slow wave and spindle power and abnormal spindle-slow wave coupling. While this study was preliminary, findings suggest that these sleep abnormalities may underlie the motor memory deficits observed in PD. Collectively, this work highlights the importance of REM and NREM sleep in memory consolidation across visual perception, imagery, and motor learning tasks. These findings could potentially lead to insights into diseases involving REM sleep abnormalities.
  • Leveraging Open-Source and Crowdsourced Data to Evaluate Spatial Justice in Cultural Planning 
    Abdelgawad, Norhan Tareq Ahmed (Virginia Tech, 2025-10-16)
    With the rise of data-driven planning and decision-making, fueled by the abundance of digital information, data are increasingly being positioned as a "common language" for interrogating the built environment. These claims perceive data as neutral representations of reality, overlooking the social, political, and institutional contexts that shape them. However, digital data can vary widely in quality, completeness, and classification standards, raising concerns regarding their accuracy, applicability, and effectiveness in examining just planning outcomes. Focusing on cultural planning as a domain and Los Angeles as a case study, I examined the utility of data for evaluating the fair distribution of cultural resources. This research addressed the following question: What can open-source and crowdsourced data reveal about the equitable allocation of cultural resources? To address this question, I developed a socio-ecological framework that synthesized Henri Lefebvre's (1991) Spatial Triad Theory and Anthony Giddens' (1986) Structuration Theory for evaluating justice in planning contexts. This framework identified three different dimensions and three types of justice. (1) The institutional dimension, conceptualized as the conceived, is centered around understanding the institutional decisions and strategies that shape cultural planning and have procedural justice implications. (2) The environmental dimension, which is interpreted as the perceived focuses on the physical manifestation of resources represented as data points and can have distributional justice implications. (3) The experiential dimension, conceptualized as the lived, which addresses the meaning and significance of these resources to individuals, which can impact participatory justice. This framework provided the opportunity to bridge theory and practice by operationalizing meta-theories to create a diagnostic adaptive tool with actionable steps for examining just planning processes and outcomes. Guided by this framework, I conducted two studies to: (1) compare digital data obtained from institutional and crowdsourced sources in terms of quality and 'fitness of use' in justice research; and (2) to compare community perceptions of cultural resources with digital data representations. The first study focused on the perceived dimension, viewing data as representations of both institutional and crowdsourced physical cultural infrastructure. In this study, I examined and compared data quality and representations from institutional city and county sources with crowdsourced OpenStreetMap. The findings revealed differences across the datasets, with significant discrepancies in spatial patterns, cultural asset classification, and descriptive detail. Recognizing the trade-offs involved in selecting a dataset for justice research in cultural planning, this analysis highlighted the need for the integration and critical examination of institutional and community-sourced data with insights from the community. The second study focused on the lived dimension of the social ecological framework emphasizing the lived experiences of the community. Taking a critical Geographic Information Science (GIS) perspective, I leveraged cultural mapping as a tool for critical data inquiry and integrated Kevin Lynch's (1960) notion of imageability as an analytical lens to identify perceived discrepancies between community insights and digital datasets. The study provided a systematic approach for the multi-level examination of cultural resources, highlighting the conceptual fuzziness in the classification of cultural resources, spatial discrepancies between community perceptions and commissioned artworks, and overlooked dimensions such as accessibility and engagement, which are crucial for cultural dataset development. Integrating theory and methods from Sociology, Urban Planning, Data Science, Geography, and Environmental Psychology, this dissertation bridged theory and practice by developing and applying a diagnostic framework to examine the utility of different types of data in justice-oriented research in cultural planning. In doing so, this dissertation made theoretical and methodological contributions, spanning institutional, environmental, and individual levels of analysis.
  • Sharing the Burden of Reactivity: Synthesis, Characterization, and Reactivity of Phosphino-Alkoxide First-Row Transition Metal Complexes
    Williams, Matthew Jacob (Virginia Tech, 2025-10-13)
    Bond activation is both central in many reaction mechanisms and essential to industrial production of chemicals. Many industries, such as pharmaceutical and petrochemical production, use precious metals to perform critical synthetic transformations that rely on bond activation. Due to the scarce nature of these precious metals, alternative approaches to bond activation have been sought. One such alternative approach is the polarization of covalent bonds of the substrate with earth-abundant catalysts. Through careful control of the ligand environment and metal selection, specific substrates can be targeted with precision. This dissertation describes the synthesis of complexes featuring earth-abundant 3d transition metals in monometallic complexes, bimetallic complexes, and frustrated Lewis pairs. Specifically, a phosphino-alkoxide ligand was developed to promote high-spin complex formation, produce unconventional geometries, and enable controlled dimerization. The complexes produced with this ligand were screened for bond activation capabilities, revealing activity for alkyne cyclotrimerization with certain monometallic complexes, both individually and in pairs. Overall, this work demonstrates the promise of earth-abundant 3d transition metal complexes as alternatives to precious metal catalysts for bond activation.
  • Examining the Connection Between Teacher Practice, Teacher Beliefs About Student Engagement, and the Classroom Environment and the Activation of Achievement Emotions in Middle School Science Students - An Exploratory Case Study
    Terwilliger, Sarah Katherine (Virginia Tech, 2025-10-13)
    Examining the Connection Between Teacher Practice, Teacher Beliefs About Student Engagement, and the Classroom Environment and the Activation of Achievement Emotions in Middle School Science Students - An Exploratory Case Study Sarah K. Terwilliger Abstract Positive student engagement is linked to increased graduation rates, higher-order thinking, and improved academic achievement in the immediate and longer term. Researchers have dedicated considerable time and resources to understanding the factors that affect student engagement in the classroom. Understanding the specific teacher practices, teacher beliefs about student engagement, and classroom environments that increase student engagement offers educational leaders, professors in teacher preparation programs, and policymakers the ability to ensure that they are providing guidance, support, instruction, and evaluation that enhances practices and behaviors that increase student engagement. The study of achievement emotions is one area of concentration within the body of research regarding student engagement. Previous studies have established the relationship between the activation of positive achievement emotions and increased student engagement and student achievement. This study seeks to add to the body of research on achievement emotions by using an exploratory multi-case study methodology to develop an understanding of how teacher practices, teacher beliefs about student engagement, and the classroom environment influence the activation of achievement emotions within the learning environment -- specifically enjoyment, hope, anger, pride, boredom, anxiety, shame, and hopelessness. I aimed to provide preliminary data from this case study that can be used in future studies to test the relationship between specific teaching practices, teacher beliefs and perceptions of student engagement, classroom environments, and the activation of achievement emotions. The study was conducted at a middle school and included three cases, all of which were teachers of seventh grade life science. One teacher was a provisionally licensed career switcher, one was a mid-career teacher who had taught special education before moving to science three years earlier, and the third was an experienced teacher who has taught the same subject in the same school for his entire career. Data collected included three interviews for each case, one classroom observation, and a student survey using an adapted version of the Achievement Emotions Questionnaire. The study found that teachers who believe the responsibility for student engagement rests primarily with the teacher and who prioritized engagement as a primary factor in their lesson design and delivery may more strongly activate positive achievement emotions and deactivate negative achievement emotions. The study also found that the same connection may exist between the activation of achievement emotions and classrooms that are semi-structured, prioritize individual student needs, create a respectful environment, and provide a balance between independence and support for student progress. The findings from this study provide a basis for future research, using other methodologies, to explore and test more specific aspects of the teacher practices, beliefs about student engagement, and classroom environments and their connection to the activation of student achievement emotions. Such research would help inform teacher training, teacher evaluation and professional learning, and teacher practice through the lens of activating positive achievement emotions and deactivating negative ones to thereby increase student engagement and achievement.  
  • Synchronous Condensers and Grid-Forming Control for the Integration of Inverter-Based Resources in Weak Grids
    Fouladi, Ehsan (Virginia Tech, 2025-10-13)
    Power systems with a higher share of inverter-based resources (IBR) exhibit reduced system strength and inertia, which are otherwise provided by synchronous generators (SG). This is because IBRs do not have the rotational inertia of SGs, and they do not contribute to short-circuit fault currents as much as SGs do. The lack of strength and inertia increases the risk of instability during contingencies, e.g., short-circuit faults. This dissertation investigates and proposes methodologies to enhance the stability of IBRs by leveraging synchronous condensers (SC) and grid-forming control. First, it presents an optimization model to find the optimal location and size of SCs, aiming to minimize the total cost of SCs and maintain the strength (measured by short-circuit ratio [SCR]) above a desired value at the point of connection of all IBRs. Second, it develops a robust exciter controller for SCs to maintain terminal voltage stability under large disturbances in weak grids. Third, it proposes a method to identify the most effective subset of IBRs to operate in grid-forming mode, accounting for the dynamic interactions between SGs and IBRs, to improve voltage and frequency stability. These contributions collectively support the reliable and cost-effective integration of IBRs into future power systems.
  • Network-Level Structural Condition Data for Sustainable Pavement Asset Management
    Murekye, Angello (Virginia Tech, 2025-10-10)
    Pavement management systems (PMS) are critical tools that agencies use to optimize maintenance and rehabilitation (MandR) decisions. Traditionally, PMS frameworks rely primarily on surface condition indicators such as cracking, rutting, and roughness to trigger treatments and allocate resources. While surface condition measures provide valuable insight into the functional performance of pavements, they do not directly capture the structural capacity that influences long-term performance. As a result, agencies risk selecting treatments that are either insufficient for addressing underlying structural deficiencies or unnecessarily conservative, leading to inefficient allocation of limited budgets. This gap is particularly relevant at the network level, where decisions must balance cost-effectiveness, system performance, and sustainability. The integration of structural condition into PMS has historically been limited by challenges in data collection technologies. The falling weight deflectometer (FWD), while effective for point testing, is unsuitable for network-level applications due to its slow operation and traffic disruption. The traffic speed deflectometer (TSD), a continuous deflection measuring device, overcomes these challenges by collecting structural data at highway speeds, making it practical for large-scale use. This dissertation presents a comprehensive approach for incorporating TSD-based structural condition information into pavement management systems. Four interrelated studies were conducted to evaluate the feasibility, effectiveness, and implications of using TSD-derived structural metrics in network-level practices. Together, these studies cover treatment selection, network needs assessment, life cycle cost implications, and environmental evaluations, providing a holistic assessment of the role of structural data in sustainable pavement management. The first study focused on the feasibility of incorporating TSD-based structural condition into treatment selection processes. A pilot framework was developed to integrate TSD-derived metrics—specifically, the effective structural number (SNeff) and remaining structural service life (RSTL)—with surface condition information in decision trees. This framework was tested on a case study along Route 29 in Virginia. Results showed that including structural data led to significant changes in treatment selection outcomes. In particular, many sections that would have been recommended for costly structural rehabilitation under surface-only assessments were instead identified as candidates for less intensive surface treatments. This adjustment reduced treatment costs by up to 33.2% in a single maintenance cycle, illustrating the potential efficiency gains from better aligning treatments with true structural needs. The second study expanded the treatment selection framework to a larger portion of Virginia's network, encompassing more than 4,250 lane-miles of interstate and primary roads. This broader application demonstrated that structural condition information has measurable impacts on network-level needs assessment and prioritization. Findings revealed that interstate pavements were generally in better structural condition than primary roads, and the integration of TSD-derived indicators such as RSTL provided a more refined understanding of where structural interventions were truly needed. Compared to surface-only approaches, incorporating structural condition enabled agencies to better distinguish between sections requiring surface-level interventions and those needing structural rehabilitation, thereby improving the prioritization of treatments under constrained budgets. The third study examined the long-term economic implications of integrating structural data into PMS by conducting life-cycle cost analyses. Using 30-year planning horizons, the study compared functional-only strategies with those that incorporated structural condition metrics. Results showed that incorporating structural information reduced life-cycle costs by up to 11.2%. These savings were largely attributable to improved treatment timing and more efficient use of resources, including reductions in unnecessary heavy rehabilitation where structural capacity remained sufficient. The findings underscore that the benefits of structural integration extend beyond short-term efficiency to include long-term economic sustainability, particularly at the strategic planning level. The fourth study addressed the environmental dimension of pavement management by enhancing roughness prediction models with structural condition data. Traditional roughness models rely solely on pavement age, limiting their ability to capture the influence of structural condition on roughness progression. In this study, structural metrics such as the surface curvature index (SCI300) and the modified structural index (MSI) were incorporated as explanatory variables in roughness deterioration models. Comparative analyses across 1,513 km (940 mi) of Virginia's primary roads showed that these enhanced models provided improved predictive accuracy. Importantly, their application in environmental life-cycle assessment (LCA) showed that structurally weak pavements deteriorate more rapidly and result in higher use-stage greenhouse gas emissions than stronger pavements. For example, relative to age-only models, 10-year use-stage emissions for a structurally strong pavement decreased by up to 43.1 tons of CO2-equivalent per lane-mile. These findings highlight the significance of structural condition in quantifying environmental impacts and the role of TSD-enhanced models in supporting sustainability-oriented pavement management. Collectively, the results of the four studies show that incorporating TSD-based structural condition data into PMS strengthens decision-making across tactical, strategic, and sustainability dimensions. At the tactical level, structural data improve treatment selection and reduce costs by aligning interventions with true pavement needs. At the strategic level, structural integration lowers long-term life-cycle costs and improves the prioritization of investments. From an environmental perspective, enhanced prediction models enable more improved evaluation of use-stage emissions and could support greenhouse gas reduction initiatives. Overall, results of this research suggest that the integration of structural condition into PMS enhances the technical, economic, and environmental sustainability of pavement management systems.
  • Dynamic Characterization of Wide Bandgap Devices for Power Electronic System Integration
    Gill, Lee (Virginia Tech, 2025-10-10)
    Wide Bandgap (WBG) devices offer significant performance advantages for next-generation power electronic systems. However, the dynamic stresses induced by application-specific conditions present limited physical and behavioral understanding, posing critical concerns for reliability and performance degradation. Therefore, this dissertation presents novel characterization methods, measurement techniques, and analytical frameworks to investigate dynamic switching stresses in WBG devices and the integration strategies to optimize power electronic system performance. The key state-of-the-art challenges and research gaps are identified related to dynamic stress effects throughout the system integration process of the WBG devices. An application-oriented converter stress characterization methodology is developed to evaluate the impact of dynamic operating stresses. Advanced measurement circuits and stress-induced performance evaluation techniques are designed to enable a deeper understanding of device behavior and facilitate maintenance or screening procedures. Lastly, multi-objective design optimization and lifetime performance evaluation of system-level WBG device integration are applied to demonstrate how dynamic stress insights can be leveraged to optimize overall system performance and develop novel operational lifetime prediction methodologies. This dissertation provides both a framework with a methodological and practical foundation for integrating WBG devices into high-performance, application-oriented power electronic systems by addressing key industry and research needs for improved characterization, measurement, reliability, and lifetime modeling.
  • Post-secondary Students' Travel Behavior through the Lens of Urban and Rural Contexts
    Meghna, Nishat Naila (Virginia Tech, 2025-10-10)
    This four-manuscript research investigates the travel behavior of post-secondary students across urban and rural contexts through four interrelated studies. The research addresses gaps in understanding how activity choices, departure time decisions, and active transportation behaviors are shaped by contextual, demographic, and policy factors. Two studies utilize the StudentMoveTO dataset, a detailed activity-travel diary survey from the Greater Toronto and Hamilton Area (GTHA), to model multi-destination trip-based activity type choices and sequential departure time decisions. These models capture interdependencies across multi-destination trips, enabling a more realistic representation of student travel patterns in dense urban environments. The other two studies draw on a custom-designed revealed–stated preference (RP–SP) survey administered to post-secondary students in rural Virginia. This survey incorporates both actual travel behavior and hypothetical choice experiments to assess rural students' mode preferences and the mental health impacts of active transportation under varying infrastructure and service conditions. The research adopted advanced econometric and machine learning approaches to better understand post-secondary students travel behavior. The urban-focused studies employ the dynamic discrete choice models and deep learning architectures Gated Recurrent Units (GRU), Long Short-Term Memory (LSTM) networks, and Transformers to capture sequential decision-making and nonlinear dependencies in activity type and departure time choices. The rural-focused mode choice analysis estimates both RP–SP multinomial logit (MNL) and RP–SP mixed logit models, enabling the combination of actual revealed preference data with stated preference scenarios while also capturing unobserved taste heterogeneity across individuals. The MNL model provides a baseline understanding of average mode choice behavior, whereas the mixed logit model relaxes the independence of irrelevant alternatives (IIA) assumption and accounts for random variations in preferences influenced by rural context, demographics, and travel conditions. The mental health and active transportation study uses principal component analysis (PCA) for dimensionality reduction, followed by Random Forest and other interpretable machine learning methods to identify the most influential factors. This combined methodological framework leverages both behavioral realism and predictive accuracy, bridging traditional econometric analysis with modern data-driven approaches. The findings reveal that student travel decisions are strongly influenced by institutional schedules, socio-demographic characteristics, and built environment features, with notable differences between urban and rural contexts. Sequential modeling shows that earlier departure times for initial trips significantly constrain subsequent activity timing, while rural analyses highlight that infrastructure quality and service availability directly affect both mode choice and perceived mental health benefits of active travel. These insights provide valuable evidence for transportation planners and policymakers seeking to design targeted, context-sensitive strategies that enhance mobility options, support student well-being, and promote sustainable transportation in both urban and rural communities.
  • Enhancing Autonomous Robots in the AEC Industry using Foundation Models
    Naderi Aghbash, Hossein (Virginia Tech, 2025-10-09)
  • Toward AI-Mediated Immersive Sensemaking with Gaze-Aware Semantic Interaction
    Tahmid, Ibrahim Asadullah (Virginia Tech, 2025-10-09)
    Motivation. Analysts who work with large text corpora must forage for evidence, con- nect disparate facts, and synthesize explanations, which imposes a heavy cognitive load. Immersive Analytics offers improving the experience with spatial memory and embodied interaction that can reduce this burden, but does not save the analyst from exhaustively browsing the corpus to find what matters. However, modern head-worn displays include eye tracking, creating an opportunity to infer an analyst's perceived interest implicitly and to provide timely, intelligible, attention-aware assistance that can essentially help in offloading some of the cognitive work. Problem. How can we model an analyst's interest from their gaze so that an AI assistant guides foraging and supports synthesis while preserving analysts' agency over the layout? Specifically, we need methods that (a) predict perceived relevance at document and term levels during multi-document investigations, and (b) expose those predictions through visual cues that help users make sense of complex evidence. Approach. We introduced a gaze-derived interest model that combines fixation duration and dwell count, adjusted for high-frequency terms, to compute GazeScore for documents and words. In parallel, we studied analysts' acceptance of different automation levels to ground design principles for a gaze-aware assistant. We operationalized the model in EyeST, an immersive analytic tool that presents two levels of visual cues to externalize the analyst's interest. Global cues provide interpretable, low-overhead signals by ranking and color-encoded evidence. Local cues reveal relationships between documents to promote discovery without clutter, while being grounded in the analyst's interest. We conducted a feasibility study that compared GazeScore to the analyst's perceived relevance. In parallel, we assessed analysts' acceptance of automated systems with a clustering task offering three levels of automation. The findings from the two studies enabled us to develop gaze-aware semantic interactions for immersive sensemaking, followed by two studies: one examining its effects on foraging, while the other tested the effects of adaptive annotation during synthesis. Results. GazeScore separated relevant from irrelevant content at the word level from the outset, enabling a real-time document relevance predictor with high precision. The clustering study showed that analysts favored assistance that preserves analyst's control over the layout and provides clear rationales. Subsequent studies revealed that global cues increased the efficiency of the analyst by helping them spend more time on relevant information while avoiding noise. Local cues encouraged individual exploration and surfaced overlooked but useful documents. Both gaze-derived cues guided analysts to essential clues vital to the sensemaking task, reduced perceived physical demand, and reduced the need for explicit externalization of the analyst's interest. Implications. The findings point toward a design pattern for AI-mediated immersive sense- making: invest in bootstrapping evidence, emphasize global high-level signals early, reveal local relationships on demand, and pair every suggestion with a clear rationale to preserve trust and agency. More broadly, this work extends semantic interaction into implicit chan- nels by showing how gaze can externalize evolving interest in real time. While our focus was on predicting perceived relevance, the approach opens pathways to incorporate other implicit signals to capture a richer picture of analysts' cognitive states. Together, these insights pave the way for more adaptive, trustworthy, and human-centered gaze-aware systems that deepen human–AI collaboration in immersive analytics.
  • Optimization of Quinone-Based, Extracellular Electron Transfer Mediators for Bioelectronic Applications
    Blackburn, Benjamin Thomas (Virginia Tech, 2025-10-09)
    To survive in anaerobic conditions, some bacteria have evolved to respire onto non-oxygen terminal electron acceptors (TEAs). A subset of these organisms, also known as metal reducing (MR) bacteria, can produce a usable electric current. Production of electric current makes MR bacteria of interest for the development of bioelectronic devices like biosensors and microbial fuel cells. The overall process of these bacteria passing electrons to a TEA is referred to as extracellular electron transfer (EET). One mechanism of EET is mediated by diffusible, redox-active small molecules, or electron shutting compounds (ESCs). Quinones, a class of cyclic, conjugated, di-keto compounds often serve as ESCs. Lactiplantibacillus plantarum, a commensal lactic acid bacterium, can utilize quinoline-mediated EET through a type II NADH dehydrogenase protein, Ndh2. In nature, quinone-based shuttling compounds are often decorated with different substituents (e.g. hydroxyls, halogens, amines) that alter their chemical properties. Recent studies have shown that quinone-based mediators of different chemical properties facilitate EET at different rates. In this project, a library of 40 unique quinone-based mediators was assembled from commercial, natural, and semi-synthetic sources. The library was screened in a high-throughput iron(III) oxide nanoparticle reduction assay, and top performers were evaluated in a bioelectronic system (BES). It was found through screening that Ndh2-dependent EET in L. plantarum activity strongly correlates to a mediators' lipophilicity [LogD (pH 7.4)] and predicted free energy of binding ∆Gcomp. A mediator promoting very stable current in a BES for 5 days, 3-Amine-menadione, was discovered. Additionally, a library of 1,4-naphthoquinones, substituted with various biogenic amines was assembled. This library was designed around probing at interactions in the L. plantarum Ndh2 active site that promotes EET. Screening for Ndh2-mediated EET in L. plantarum in the nanoparticle reduction assay showed that 1,4-naphthoquinones substituted with amines able to promote pi-stacking interactions with Tyr-403 were top performers. The compound with the highest EET rate was the only primary amine-substituted mediator, 3-Amine-menadione. Ndh2-dependent EET is strongly correlated to LogD (pH 7.4). A predictive EET metric, or composite score, was created based on in vivo data, molecular modeling, and ADME properties of all mediators. Lastly, highly active EET amine-substituted 1,4-naphthoquionones are proposed to be covalently bound to polymer networks. The redox-active polymers are to be embedded on an electrode for the purposes of enhancing EET in BES. Mediators with two functional groups, epoxides and free thiols, will be synthesized for the addition of chitosan and methacrylate/methyl acrylate polymers, respectively. The redox-active polymers will be screened for Ndh2-mediated EET L. plantarum in nanoparticle reduction assays and BES. Highly active quinone-based mediators will be used to make bioelectronic communication faster and more efficient.    
  • Pushed into STEM: Investigating the Racialization and Cultural Influences on East Asian Americans' Decisions to Major in STEM
    Zang, Ye (Virginia Tech, 2025-10-09)
    Asians are portrayed as model minorities, with science, technology, engineering, and mathematics (STEM) education proficiency and a natural aptitude for STEM occupations. While the Model Minority Myth portrays Asian Americans as naturally suited for technical and academic success, it ignores the diversity of their goals and can dehumanize individuals by channeling them into narrowly defined career paths. This study aimed to highlight the social justice implications of these dynamics, examining how societal stereotypes and cultural norms can limit personal agency, overshadow individual talents, and ultimately contribute to systemic inequalities by racializing Asians in STEM. This study used social cognitive career theory and Model Minority Myth as its framework and employed a qualitative approach through semi-structured interviews to investigate East Asian Americans' career choices, focusing on how the intersection of cultural values with social and racial expectations shapes their career choices in STEM fields. This study argues that East Asian Americans' decisions to pursue STEM careers were influenced by a complex interplay of cultural values, familial expectations, and racialized perceptions. Although the Model Minority Myth is widely regarded as an external societal stereotype, this study discovered that it is actively reinforced within East Asian American communities themselves through family pressures, peer comparison, and communal definitions of success, which complicates individuals' career choices and identity formation. The findings from this study contribute to a broader understanding of the social and cultural pressures faced by Asian Americans, shedding light on the social justice mission embedded in fostering career diversity and freedom from racialized expectations. The study also advances social cognitive career theory by including the Model Minority Myth as a critical contextual variable, increasing the framework's ability to account for racialized stereotypes, cultural obligations, and the internalization of societal expectations, all of which have a unique impact on the career development of East Asian Americans.
  • Improvements to Enhance The Security and Reliability of Crowdsourced Spectrum Access Systems and Cognitive Radio Networks
    Tolley, Joseph D. (Virginia Tech, 2025-09-23)
    This dissertation addresses key challenges in dynamic spectrum sharing within Cognitive Radio Networks (CRNs) and Spectrum Access Systems (SASs), focusing on the U.S. 3.5 GHz Citizens Broadband Radio Service (CBRS) band. A structured survey of existing regulatory frameworks, coordination methods, and interference mitigation strategies provides context for the research contributions that follow. To enhance trust in user-reported data, the Whisper Key Location Verification method is introduced. It validates the physical location of crowdsourced nodes by combining radio and internet-based checks, filtering out falsified reports and improving Primary User (PU) protection. The Enhanced Heartbeat Protocol (EHP) enhances SAS–Secondary User (SU) communication through asynchronous messaging and an expanded message format, enabling faster spectrum reassignments and supporting mobile scenarios, such as UAV networks. To detect Spectrum Sensing Data Falsification (SSDF) attacks, a real-time framework using lightweight similarity metrics identifies duplicated or manipulated sensing data, increasing system resilience. Finally, the Radio Frequency Obstructed Observation Area Identification (RF-OOAI) method distinguishes environmental signal loss from intentional misreporting, preserving the reputation of honest users. These contributions collectively improve the accuracy, efficiency, and robustness of shared spectrum systems, advancing the design and reliability of CRNs and SASs in complex real-world settings.
  • Design and Evaluation of Network Algorithms and Deep Learning Models in Systems Biology and Biomedicine
    Tasnina, Nure (Virginia Tech, 2025-10-07)
    We present novel solutions to four interrelated problems in network biology and computational biomedicine. The first two address methodological challenges in network biology, while the latter two focus on applications of network-based computational modeling in therapeutics. (i) We introduce a provenance tracing framework for random walk-based network diffusion algorithms. By quantifying the contribution of individual paths to a node's diffusion score, the framework captures the influence of global network topology and identifies critical mediators of information flow. This approach enhances the interpretability of diffusion-based predictions, a key requirement for their reliable application in biomedical contexts. (ii) Recognizing that the utility of network diffusion and other network-based algorithms depends on the quality of the underlying networks, we present ICoN, an unsupervised coattention-based graph neural network model for integrating heterogeneous protein-protein interaction networks. ICoN learns joint embeddings across multiple networks and captures complementary biological evidence. ICoN surpassed individual networks across three downstream tasks: gene module detection, gene coannotation prediction, and protein function prediction. Compared to existing unsupervised network integration models, ICoN exhibits superior performance across the majority of downstream tasks and shows enhanced robustness against noise. This work introduces a promising approach for effectively integrating diverse protein-protein association networks, aiming to achieve a biologically meaningful representation of proteins. (iii) Shifting our focus towards therapeutic applications of computational models, we systematically evaluate deep learning models for drug synergy prediction, an important challenge in combination therapy design. Using the SynVerse framework, we assess 16 models across diverse feature types and deep learning based architectures and demonstrate that models often exploit dataset-specific shortcuts rather than biologically meaningful drug or cell line features. Robust evaluation, as enabled by SynVerse, may improve the likelihood that model predictions hold up when tested in experimental and clinical settings. (iv) Finally, SynVerse's findings motivate the need to build a database of synergistic drug combinations annotated with their underlying synergy mechanisms so that one may build mechanism-aware predictive models. Hence, in the final project, we construct a structured database of synergy mechanisms from biomedical literature using a large language modelbased pipeline. This approach integrates literature retrieval, information extraction, biomedical entity recognition, and knowledge graph construction to extract both free-form and structured representations of mechanisms. The resulting resource, encompassing around 3,000 drug combinations, lays the foundation for developing mechanism-aware predictive models for drug synergy.
  • Towards Explainability and Domain Knowledge-inspired Design of Online Real-Time Learning Techniques in NextG Wireless Systems
    Jere, Shashank Harish (Virginia Tech, 2025-10-06)
    The air interface of Next-generation (NextG) cellular and wireless networks are expected to incorporate artificial intelligence (AI) and machine learning (ML) in order to meet increasingly stringent performance requirements. Multiple-input multiple-output (MIMO) technology, including massive MIMO and subsequent variants, has played a central role across successive cellular generations, accompanied by continuous design and complexity evolution. AI/ML-based techniques can play a promising role in meeting these stringent performance demands especially with MIMO in NextG systems. However, designing AI/ML-driven approaches for the NextG air interface remains challenging due to the wide range of possible system configurations, the extremely dynamic nature of wireless channels, and the need for adaptability to real-time operational adjustments. Therefore, online and real-time AI/ML approaches can play a key enabling role in realizing this ambitious vision for NextG. To this end, this dissertation first introduces the theoretical underpinnings of online real-time learning architectures based on reservoir computing (RC). The effectiveness of RC in orthogonal frequency division multiplexing (OFDM) and MIMO-OFDM receive processing is established from the ground up with first principles, resulting in enhanced explainability and interpretability of RC-based architectures, thereby turning opaque ``black-box'' models into intuitive ``gray-box'' models. This solid foundation, founded on signal processing and information theory fundamentals, enables the systematic development of procedures to incorporate domain knowledge into the design of RC-based architectures, resulting in significantly improved performance, which is demonstrated in the context of OFDM and MIMO-OFDM receive processing, user beam tracking in massive MIMO systems and near real-time jamming detection and classification in NextG systems. This dissertation emphasizes the crucial role of explainability of AI/ML solutions deployed in NextG wireless systems, and the foundations laid in this dissertation provide a potential roadmap for developing explainable and domain knowledge-guided AI/ML-based techniques in NextG.