Computer Engineering Commons^™

From Recommender To Actor: The Normative Boundary When Rag Tools Become Tool-Calling Agents, Md Tahmid Rashid

Faculty Publications - Information Technology

Retrieval-Augmented Generation (RAG) systems increasingly operate not only as tools for retrieving and synthesizing information, but also as agents that can invoke external functions, modify digital environments, and execute tasks across software systems. This development raises a specific normative problem: the point at which a model’s output ceases to be merely informational and becomes an executable intervention in the world. Building on existing work in Responsible AI, accountability, and human oversight, this paper argues that tool-calling architectures place particular pressure on these frameworks because they can fuse retrieval, reasoning, and action within a single operational pipeline. To clarify this transition, …

Personality Exceptions: Enhancing Character In Narrative Planners With Internal Memory, Elinor Rubin-Mcgregor

University of Kentucky Doctoral Dissertations

Stories and narratives are powerful tools for forging connections between people, as they induce emotions shared through their audience and help build empathy and understanding through sharing unique perspectives. Interactive stories can do all this and more, as they can be used to better educate and engage their audience. But in such stories, the characters must have consistent personality and human-like emotions in order for the audience to relate to them. Current implementations of personality in interactive narratives either portrays characters with static personality that feel less realistic, or else require a great deal of additional work from the author …

Non-Invasive Digital Restoration Of Damaged Photographic Film Negatives, Ankan Bhattacharyya

University of Kentucky Doctoral Dissertations

Physical restoration of damaged photographic film causes more damage. Also, existing non-invasive digital restoration of film negatives does not produce print-quality optical images. Instead, they produce X-ray projections, which are not the same as optical projections. This thesis addresses these problems and establishes a framework that can digitally restore old, damaged films without the need to open them physically. Virtual Unwrapping is an existing pipeline that has proven itself over the last couple of decades to work on unopenable papyrus scrolls, like the Herculaneum Scrolls. This thesis utilizes the concept of virtual unwrapping to restore damaged photographic film negatives. Due …

Detection And Management Of Attacks On Synchronized Networks, Michael T. Spearman

Honors Theses and Capstones

The Precision Time Protocol (IEEE 1588) provides sub-microsecond clock synchronization across packet-switched networks and has become foundational infrastructure in 5G fronthaul, industrial control systems, and financial exchanges. Despite its criticality, most deployed PTP networks operate without active security monitoring, and no standardized detection mechanism exists for the class of attacks that deliberately stay below conventional jitter thresholds. This thesis investigates whether hardware-level ptp4l offset logs alone are sufficient to reliably detect two such attacks, slowly wandering packet delay injection and rogue master spoofing, and whether detection can occur before severe synchronization failure.

A hardware-in-the-loop testbed was constructed using two hosts …

Formalizing Asymmetric Control-Telemetry Separation In Distributed Industrial Control Systems, Andrew Manison

College of Graduate Studies: Theses & Dissertations

Distributed industrial control systems often place control and telemetry traffic on the same communication substrate even though the two workloads impose different requirements. Control paths need bounded request-response latency and predictable acknowledgement semantics, whereas telemetry paths benefit from scalable publish-subscribe fanout and tolerance for consumer-side delay. This thesis argues that, for the tested class of mixed workloads on shared commodity infrastructure, these communication roles should be separated architecturally rather than forced through a single protocol. To evaluate that claim, the thesis formalizes an asymmetric control- telemetry pattern and instantiates it in the Asymtra framework using gRPC for synchronous control and …

Comprehensive Performance Evaluation Of Devops Infrastructure Under Dynamic Workloads, Abdulrazaq Mamud

College of Graduate Studies: Theses & Dissertations

This research aims to investigate performance optimization and reliability issues related to cloud-based computing environments through an analysis of three key infrastructure components: virtualized CPU resource management, distributed API rate limiting, and web server deployment architectures. This research combines machine learning and system experimentation as a way of exploring the impact of infrastructure-level behaviors on overall system performance and scalability. The first component of the research focuses on analyzing CPU Fragmentation in Virtualized Environments, where unbalanced workload allocation on Virtual CPU Cores causes increased tail latency, resulting in Service Level Agreement violations. Metrics are analyzed using the Random Forest classifier …

Minimizing Performance Overheads For Crash-Consistency In Disaggregated Persistent Memory, Khan Shaikhul Hadi

Graduate Studies Theses and Dissertations 2026

Compute express link (CXL) enables persistent memory disaggregation with memory pooling and hardware managed multi-host memory sharing capability, resulting in better resource utilization, increased scalability. Persistency-aware applications need to manage crash consistency across the system which results in significant performance overhead. This dissertation systematically investigates performance overhead to achieve crash consistency in disaggregated persistent memory and proposes solutions to enable persistency-aware application scaling for distributed system. First, we study persistent parallel programming to scale computation capability beyond single processor and determine the underlying hardware limitation to adopt lock-free data structure. We propose hardware support to design durable atomic instruction (DAI) …

Memory-Efficient Acceleration For Emerging Applications Via Hardware/Software Co-Design, Shilin Tian

Graduate Studies Theses and Dissertations 2026

Emerging artificial-intelligence and data-intensive scientific workloads increasingly face a memory wall: irregular access patterns and large intermediate data volumes make data movement, rather than arithmetic, the primary constraint on performance and energy efficiency. This dissertation develops a memory-centric hardware/software co-design methodology that jointly reshapes algorithms, architectures, and dataflows to retain frequently reused data on chip. The methodology is demonstrated through three accelerators and an RTL design tool. VITA replaces multi-head attention in vision-transformer-based 3D human mesh recovery with hardware-friendly average pooling and maps the resulting operators to a reconfigurable datapath, achieving 5.05-fold and 69.12-fold speedups over a state-of-the-art GPU and …

Designing Narrative-Based Ai Assistance For Sensemaking In Collaborative Environments: Case Studies In Education And Dementia Care, Dylan Edward Moore

Dartmouth College Ph.D Dissertations

This thesis addresses a gap in the human-computer interaction literature regarding the design, development, and evaluation of narrative-based AI assistance for collaborative, complex problem solving. I explore this design space through three case studies across the domains of education and dementia care. This work encompasses multi-year industry partnerships and longitudinal fieldwork, user-centered design, dataset curation, model training, and system evaluation.

Specifically, the first case study considers a story-based web platform for teaching AI literacy through peer-generated, personalized narrative scaffolding. Learners on the platform showed significant knowledge gains and other learning-related outcomes. To describe the novel design of this system, I …

Machine Learning Acceleration Of Micro-Credentialing, Knowledge Gap Remediation, And Micro-Recruiting Of Stem Assessments: From Learner Data To Hardware, Paul Amoruso

Graduate Studies Theses and Dissertations 2026

The growing demand for a skilled STEM workforce calls for adaptive, data-driven education that personalizes learning while aligning competencies with industry needs. This work introduces a multi-layered framework that integrates artificial intelligence (AI), data mining and hardware-aware optimization to support this goal. At its core, the AchieveUp system applies generative AI-assisted skill tagging to digitized assessments, enabling multi-semester tracking of student progress. Analysis across 14 semesters shows that underperforming students were found to benefit most from attending score clarification sessions, a relationship validated by an R² value of 0.83, indicating strong predictive validity between attendance and skill acquisition. To remediate …

Restoring Digital Trust: Decentralized Frameworks For Identity, Consent, And Media Provenance, Raghu N. Avula

Graduate Studies Theses and Dissertations 2026

Digital platforms are deeply embedded in modern society through centralized network architectures. As online interactions transition to the physical world handling sensitive data, user privacy and cryptographic security require enhancement beyond traditional limits. This dissertation presents a decentralized “Tri-Layer Trust Architecture” that leverages Verifiable Trust Registries, Cryptographic Binding, and Edge Verification to prioritize data sovereignty and local authentication. To demonstrate the architecture’s applicability across distinct operational environments, we implement three solutions: (1) Verifiable Presence for peer-to-peer ride-sharing via D-RideX, which encrypts biometric data on physical hardware (TEEs) and Soulbound Tokens (SBTs) to prevent impersonation without cloud database exposure; (2) Verifiable …

Deceptive Design And Disability: How Dark Patterns Affect Users With Visual Impairments, Konstantin Metz

Graduate Studies Theses and Dissertations 2026

Dark patterns are deceptive interface designs that steer users toward unintended actions, often at financial or privacy cost. This thesis examines how those that are vision impaired are at greater risk of these dark patterns given their visual acuity deficit. We conducted semi structured interviews with nine participants across common contexts including e-commerce checkout, subscription and cancellation flows, software download pages, and mobile permission prompts. The data show recurring exposure to dark patterns such as hidden or de-emphasized fees, pre-selected add-ons, deceptive advertising and fake download buttons along with unequal consent paths, repeated permission nagging, and cancellation friction. Participants reported …

Abig4l: A Semantic Reasoning Engine For Verilog Via Abstract Interpretation, Alicia Thoney

Graduate Studies Theses and Dissertations 2026

Modern Electronic Design Automation (EDA) flows treat structural and semantic reasoning as separate concerns. Synthesis structurally optimizes RTL, while dedicated verification teams validate functional behavior through testbenches and formal methods. This organizational divide means designers have no automatic, sound way to understand what their RTL does at the moment they are writing it. This thesis introduces AbIG4L (Abstract Interpretation of Generic 4-valued Logic), a static analysis framework that bridges this gap by delivering sound semantic reasoning automatically, directly within the hardware design loop.

The framework operates on a language-agnostic intermediate representation to compute sound over-approximations of hardware behavior across 0, …

Watch Your Step, Mind The Crowd: Socially Compliant Navigation In Unstructured Environments, Jack Vice

Graduate Studies Theses and Dissertations 2026

Autonomous mobile robots increasingly operate in complex real-world environments characterized by uneven terrain, dynamic obstacles, and human presence. This dissertation addresses the fundamental challenge of developing adaptive navigation strategies that enable robots to safely and effectively traverse unstructured outdoor environments including construction sites, agricultural fields, and disaster areas while maintaining social compliance around people. We advance this capability through five complementary contributions that collectively target dynamic, unstructured, human-occupied environments. First, we demonstrate evolutionary optimization of legged mobility for uneven terrain. Second, we propose using a spatiotemporal autoencoder to learn the temporal dynamics of moving people and objects. Third, we introduce …

Dashboard And Racing Telemetry, Cole Barach, Jacob Koshel, Ethan Zifzal, Matthew Sullivan

Williams Honors College, Honors Research Projects

The main goal of the project is to design and manufacture a combined dashboard and data logger for the vehicles produced by the Zips Racing design team. The dashboard will intuitively display real-time information to the driver and record all received information while driving. This information may be pulled off the device later for performing data analysis. This project will incorporate custom PCB design, surface mount soldering, embedded software development, and the CAN communication protocol.

Robotic Air Hockey Table, William Forcey, Andrew Piunno, Kaden Carpenter, Xander Zavatchen

Williams Honors College, Honors Research Projects

Air hockey, a popular arcade game, is traditionally designed for two players. This limits the game’s accessibility for individuals who wish to practice or enjoy it as a single player. To solve this problem, a robotic system was implemented to play air hockey against a human player. The speed and acceleration of the puck and mallet were measured from a game played between humans to inform the required movement capabilities of the robot. The robotic opponent implemented observes the location of the puck on the table using a camera and predicts where it will be in the future. A Cartesian …

Secure Peer-To-Peer Messaging Using Distributed Servers, Jack Stoller

Williams Honors College, Honors Research Projects

Traditional messaging systems that rely on centralized infrastructure often introduce three security vulnerabilities: single points of failure susceptible to availability attacks, metadata leakage, and man-in-the-middle attacks.

This project designs, implements, and validates a distributed peer-to-peer messaging system that addresses each of these concerns through architectural design. A consistent hash ring is used with virtual nodes to distribute user data across a mesh of connected nodes, eliminating centralized dependency on any single server. Message contents travel directly between browser clients over a WebRTC data channel secured with end-to-end DTLS encryption, bypassing the server infrastructure. Inter-node communication is protected by mutual TLS …

Automated Pill Dispenser, Ryan Oderkirk, Connor Beaven, Rachelle Labrie, Josue Panchana

Williams Honors College, Honors Research Projects

The project we propose is an automated system for dispensing dosages of medication throughout the day. It will be able to alert a user when their pills need to be taken and give them the correct dosages of up to four different medications. These dosages are configurable as well as the scheduled time they are to be taken. In addition, the pill dispenser will alert users when they are low on medications and need to refill the machine.

Nail Salon Appointment Management System., Alexander Tran

Williams Honors College, Honors Research Projects

I will create an appointment management system for my family's nail salon. This will contain a database and tables that keep track of data of the employees, customers, and the receptionist to make it easier to book appointments for the nail salon. This was originally made to create an appointment system because I have family members who don't understand technology well and wanted to make something simple and intuitive for anyone to use, since the appointment system that we currently use is extremely off the cuff, and this should remedy that.

Toward Neurosymbolic Reinforcement Learning Via Editable Specifications, Vedant Khandelwal, Hong Yung Yip, Amit Sheth

Publications

Reinforcement learning systems are commonly adapted to new settings by retraining or fine-tuning policies. This default is costly, difficult to audit, and poorly aligned with structured requirement changes such as revised safety rules, new operational constraints, or updated user preferences. We argue for an alternative abstraction: adaptation via edits to an external, human-readable specification that the agent consults at execution time. We propose conditioning decision-making on an editable knowledge graph encoding (i) rules capturing action applicability and high-level effects, (ii) hard constraints defining feasibility, and (iii) soft preferences shaping tradeoffs among feasible behaviors. Requirement changes become graph edits, not policy …

A Dataset-Centric Analysis Of Evaluation Protocols In Face Morph Attack Detection, Sheerko R. Hmasalah, Abdulbasit K. Al-Talabani, Aras T. Asaad

Polytechnic Journal

Face morph attack detection (MAD) is often evaluated using datasets collected from diverse sources, created for different purposes, and acquired under varying conditions. This heterogeneity raises questions about how evaluation protocols and dataset characteristics influence reported performance. This work adopts a dataset-centric perspective to analyze the effects of evaluation protocol and dataset composition, including visual quality and morph-to-genuine imbalance, on MAD performance. Experiments are conducted on eight datasets and four morphing techniques using a fixed pipeline evaluated across six backbones, including Vision Transformer (ViT) and five CNN architectures, and ten evaluation scenarios covering homogeneous and heterogeneous cross-morph generalization. Performance is …

Performance Analysis Of Sparse Neural Networks In Brain Abnormality Detection, Megan Danh

Honors Undergraduate Theses

Neuroimages have held the capability of revealing to medical professionals patterns for brain abnormalities since their development. However, more recently, these professionals and researchers are looking to use neural networks to identify these brain abnormalities through neuroimages for early detection that would allow more effective treatment. Neuroimage datasets, specifically functional magnetic resonance imaging (fMRI), are extremely large in size. This would result in their processing and training to be computationally expensive, even with smaller neural networks. Fortunately, recent pruning methods have recently emerged, where network weights and neurons are pruned to reduce computational cost without compromising too much accuracy. By …

Lidar-Based Point-Cloud Human Modeling: Pose Estimation, Body-Parts Segmentation, 6g Localization, And Applications, Omar Rinchi

Doctoral Dissertations

"The human body represents a rich source of physiological and behavioral information, where precise analysis of anatomy, body parts, 3D pose, and motion enables cross-disciplinary precision applications. This dissertation formulates this challenge as a human modeling problem, where proposed algorithms convert the human body into a three-dimensional digital twin capturing anatomical structure and pose. These representations are further analyzed using additional algorithms to enable precision applications. To realize this vision, LiDAR sensing is adopted due to its privacy-preserving nature, robustness to lighting conditions, color-blind sensing characteristics, and decreasing cost.

Human modeling using LiDAR is challenging due to sparse, irregular, and …

Deep Learning And Computer Vision-Based Modeling For A Real-Time Traffic Congestion Monitoring Framework, Neil Dustin Benedict A. Agner, Angelo B. Dela Cruz, Aster Benedict A. Mangabat, Raphael B. Alampay, Patricia Angela R. Abu

Ateneo Laboratory for Intelligent Visual Environments

As intelligent traffic systems evolve to manage complex urban mobility, conventional congestion estimation techniques, such as the time-windowed Volume-to-Capacity (V/C) ratio, fail to capture capture the real-time traffic situation. Because these methods rely on vehicles crossing a specific point, they often fail to register stopped cars, creating a ‘zero flow’ paradox where complete gridlock is incorrectly categorized as an empty, free-flowing road. To address these limitations, this study proposes an end-to-end computer vision framework utilizing the Real-Time DEtection TRansformer (RT-DETR) model to quantify vehicle volume and immediately classify a roadway’s Level of Service (LOS) through spatial density. In processing video …

Artificial Sense-Making Dataset, Jason A. Bengtson, John Sandstrom, Nathan Camp

NMSU Library: Datasets

No abstract provided.

Improving Efficiency In Noma Schemes Having Inter-User Interference Using Mechanism Design, Zory Marantz

Publications and Research

Modern wireless systems utilize non-orthogonal multiple access to increase their rate capacities; however, the efficiency of the individual utility defined in bits per Joule has yet to be considered. Multiple variations of non-orthogonal multiple access have the interference of the signal-to-interference-plus-noise ratio as a function of the received power from multiple other users due to code implementations that are non-orthogonal or non-ideal cancellation in successive-interference-cancellation methods. Game theoretic concepts are used to improve user bits-per-Joule performance. Previous solutions increment transmit power and are not based on closed form systematic methods. The mechanism design presented here led to a non-cooperative Nash …

Probing Proficiency-Related Neural Representations With Pca And Ica, Onila R. Narayana Mudalige Don

Graduate Student Theses, Dissertations, & Professional Papers

How proficiency-related information is represented in task fMRI depends not only on the data themselves, but on the representational lens used to summarize them. This thesis examines second-language (L2) proficiency as a problem of representational organization rather than as a simple classification exercise. Using task fMRI from adult language learners performing semantic animacy judgments in their native language (L1) and second language (L2), I derive shared low-dimensional network representations with principal component analysis (PCA) and independent component analysis (ICA), then evaluate those representations under matched leakage-controlled decoding pipelines.

Across analyses, the central comparison is between representational frameworks rather than between …

Cognitive Load Classification Using Functional Near-Infrared Spectroscopy, Pratham Shah

Theses and Dissertations (Comprehensive)

This thesis investigates the classification of cognitive load using functional near-infrared spectroscopy (fNIRS) signals recorded during an N-back working memory task. The study introduces a novel short-channel correction layer designed to suppress superficial physiological noise adaptively, addressing limitations of traditional General Linear Model (GLM) based regression. A single participant dataset comprising 69 validated sessions was analyzed using both conventional machine learning and deep learning approaches. Traditional classifiers: Linear Discriminant Analysis (LDA), Support Vector Machines (SVM), Random Forests, and Gradient Boosting were first evaluated using statistical features (mean, variance, peak, and slope). Among these, Gradient Boosting achieved the highest accuracy (55.6%), …

Lidar-Based Framework For Detecting Suspicious Human Activities, Ahd Aljumah, Charalampos Antoniadis, Hakim Ghazzai, Nawfal Guefrachi, Ahmad Alsharoa, Gianluca Setti

Electrical and Computer Engineering Faculty Research & Creative Works

This study explores the development of Human Activity Recognition (HAR) systems capable of identifying suspicious activities to enhance security in public spaces. We propose an innovative solution that integrates LiDAR sensors with deep learning technologies. Our method employs advanced models operating on LiDAR point cloud, PV-RCNN for human detection, and LidarGait++ for classifying activities into categories such as standing or walking (non-suspicious) and sneaking or fighting (suspicious). Due to the scarcity of suitable real-world datasets for training such systems, we utilize a 3D simulation tool, Blender, to create realistic environments and generate labeled point cloud data. This synthetic dataset allows …

Solving High-Dimensional Differential Equations Using Recurrent And Residual Neural Network Architectures, Hind Khaled Kolaib

Knowledge Engineering and Data Science

High-dimensional Partial Differential Equations (PDEs) form the foundation of complex process modeling in various scientific and engineering applications, including finance, physics, and optimal control. However, classical numerical methods are adversely affected by the curse of dimensionality, making them inapplicable for large-scale problems. Recently, however, deep learning-based approaches have provided a new toolbox for these high-dimensional PDEs, including methods such as the Deep Backward Stochastic Differential Equation (Deep BSDE) method. Our approach draws on a more sophisticated deep learning backbone, using neural networks (in our case, a Residual Neural Network and a Long Short-Term Memory network (LSTM) integrated into the Deep …