Computer Sciences Commons^™

Augmenting Deep Learning For Efficient Nextg Wireless Communication And Sensing Systems, Hem Kanta Regmi

Theses and Dissertations

Wireless networks have become an integral aspect of our daily lives. Over the years, earlier generations of wireless networks have enabled some innovative applications, such as wireless gaming, fast internet browsing, and home automation, which were previously unattainable. However, to support emerging technologies like autonomous driving, virtual reality, telemedicine, and intelligent manufacturing, which require high data throughput and low latency, there is a need for advanced wireless networks. Legacy networks like WiFi/LTE, which operate below 6 GHz, have limited bandwidth and are insufficient to fulfill the high data throughput demands of several applications. Millimeter-wave (mmWave) networks, operating between 30 GHz …

Event-Based Camera Simulation And Neural Network Processing For Autonomous Aerial Refueling, Stephanie C. Hanson

Theses and Dissertations

Event-based cameras excel in dynamic environments, and do not face challenges like washout and motion blur, like a frame-based camera. This work describes the process used to collect the first EBS data collect for use in AAR, and develops an event simulator to generate synthetic training data for evaluating CNN architectures on asynchronous data. The three models compared are a traditional CNN, a YOLO-based CNN, and an asynchronous sparse CNN. The YOLO-based model achieved the best accuracy, while the sparse CNN, despite being less optimized, maintained an average IoU of 0.9. These results highlight the potential of asynchronous approaches for …

Lethality And Survivability Of Autonomous Self-Sensing Uavs, Jeffrey T. Wilson

Theses and Dissertations

Unmanned Aerial Vehicles (UAVs) have seen increased usage over the past two decades during the Global War on Terrorism (GWOT), operating in low-risk environments against dispersed enemies with minimal counter-drone capabilities. However, as the U.S. military shifts focus to Multi-Domain Operations (MDO) and Large Scale Combat Operations (LSCO), UAVs face significantly higher risks, including frequent and successful attacks, as well as the exploitation of their technology. Battle damage assessment (BDA) is not new; however, autonomous self-assessment by UAVs represents a novel advancement. Currently, UAV BDA relies on manual inspection, requiring approximately eight hours per drone. By adopting self-sensing technology, UAVs …

Hyperparameter Tuning For Robust Autonomous Vehicle Vision, Nico D. De Ros

Theses and Dissertations

Classification “flickering,” where the classification of an object changes inconsistently between consecutive video frames, remains a persistent issue in modern object classification algorithms. This problem undermines the reliability of autonomous vision systems and poses significant risks in high-stakes applications such as autonomous vehicles. This thesis explores the use of response surface methodology, a statistical design of experiments technique, to optimize hyperparameters across three object classification pipelines. The first pipeline combines YOLOv8 with SORT to establish a benchmark. The second integrates a Bayesian back-end, while the third employs an exponential smoothing back-end. Hyperparameter tuning was conducted using a two-step process: an …

Evaluating Educational Benefits Of A Custom Cyber Game: ‘Hvac Attack!’, Jillian S. Valente

Theses and Dissertations

Cyber competition and conflict remain an enduring concern for the Department of Defense (DoD). Positive control of cyberspace is crucial across the vast diversity of military operations and supporting activities. Military members play an important role in cyber prevention, detection, and remediation, but most receive relatively little training outside of the annual Cyber Awareness Challenge. Particular career fields within the DoD may benefit from specialized training in cybersecurity, in particular the civil engineering (CE) community supporting critical infrastructure protection. Prior research has suggested that game-based learning (GBL) can be beneficial for teaching cyber concepts.

Evaluating Learning Outcomes In A Serious Game: A Practical And Model Checking Approach, Matthew D. Douglas

Theses and Dissertations

This research introduces a novel computational framework to evaluate and predict the educational impact of serious games during development. By using finite state machines (FSM) and model-checking techniques, this study evaluates two serious games. Traditional evaluation approaches, often reliant on resource-intensive human trials, lack scalability and fail to provide early insight into the alignment of game mechanics with learning objectives. This study addresses these challenges of traditional evaluation methods.

A Multi-Objective Reinforcement Learning Framework For Title Autonomous On-Orbit Inspections, Austin C. Reynolds

Theses and Dissertations

The rapidly evolving landscape of space operations necessitates dynamic and autonomous systems to address complex challenges such as Resident Space Object (RSO) inspections. This research explores the application of a Multi-Objective Reinforcement Learning (MORL) framework to rendezvous and proximity operations (RPO), enabling agents to balance conflicting objectives like time efficiency, fuel conservation, and information gain. Unlike traditional reinforcement learning, MORL allows dynamic reweighting of objectives without retraining, offering adaptability and efficiency in multi-objective environments. The study demonstrates MORL's capabilities through custom 2D and 3D simulations of Hill-Clohessy-Wiltshire (HCW) environments and comparing its performance to traditional RL in RPO scenarios. Tasks …

Graph Neural Network-Based Uav Coverage Planning For Robust And Efficient 3d Environments, Gal Tsfaty

Theses and Dissertations

This thesis addresses the challenge of generating optimized UAV waypoints for complete coverage of complex 3D environments, utilizing graph-based computational techniques. The proposed framework replaces computationally intensive steps—triangulation and three-coloring—within the Vantage Waypoint Set Generation Algorithm (VWSGA) pipeline with Graph Neural Networks (GNNs). By learning structural patterns, the GNN achieves scalable and robust triangulation and node classification, enabling enhanced coverage planning in irregular geometries. A novel penalty mechanism ensures alignment with graph structure during adjacency prediction. Experimental results demonstrate the effectiveness of GNNs in balancing accuracy, computational efficiency, and adaptability, advancing UAV coverage optimization.

Evaluating A Military Digital Badging System Prototype, Benjamin T. Pederson

Theses and Dissertations

The Department of Defense is committed to developing and maintaining a highly skilled workforce capable of defending the United States and associated interests abroad. Digital badging systems, a form of micro-credentialing, offer a way to record service member competencies. By providing decision-makers with granular data, this technology could augment the military’s development of a highly skilled workforce, especially in technical career fields including cyber operations. Mixed-method data from thirty-six participants suggest that establishing a digital badging program could increase deterrence and operational effectiveness.

Autonomous Vehicle Path Planning Under Uncertainty, Madison C. Gillan

Theses and Dissertations

Autonomous vehicles are increasingly being deployed for use in high-stakes and uncertain environments where safe and efficient navigation is critical. In these scenarios, traditional path planning approaches, which rely primarily on deterministic models and fixed assumptions, fall short due to the inherent uncertainty of dynamic threats, sensor inaccuracies, and incomplete information. This research addresses these challenges by developing a novel path-planning methodology that combines the Chance-Constrained Rapidly Exploring Random Tree* (CC-RRT*) algorithm with a probabilistic risk assessment heuristic. This method models uncertainty in sensor detection zones, obstacles in the environment, and the Autonomous Vehicle itself, which allows for uncertainty during …

Spatiotemporal Prediction Of Atmospheric Events Through Recurrent Deep Learning Model, Brian W. F. Popick

Theses and Dissertations

The main contributions of this research is to add to the growing library of literature on the use of deep learning algorithms for the spatiotemporal prediction of dangerous atmospheric and hydrologic phenomena. Specifically, we develop novel attention-based and non-attention-based recurrent neural network frameworks to produce short-range sequential forecasts for lightning and tornado occurrences. Additionally, we introduce methods that account for and include error in the model tuning process to generate more reliable models. Furthermore, we have created a lightweight spatiotemporal tornadic prediction dataset that we plan to make publicly available. The first component of this research develops three novel spatiotemporal …

Hypergame Models For Cyber Defense In A Purple Team Setting, Thomas N. Whitney

Theses and Dissertations

Hypergame theory and purple teaming are two fields that can support an increased cybersecurity posture. This research investigates a hypergame theory framework that incorporates fittingly into the purple team feedback loop. Additionally, this research integrates empirical data into the hypergames. The data and the hypergames are supported by the MITRE ATT&CK framework. This research also includes a review of available game theory and hypergame theory software, five different hypergame models, a comparison of the two hypergame formats, and an innovative analysis technique using a multi-stage hypergame to represent the cyber kill chain. One finding of this research is that a …

Uranium Particle Classification Using Statistical Machine Learning And Deep Neural Networks For Nuclear Forensics, Lee C. Lambert

Theses and Dissertations

The classification of uranium particles from scanning electron microscopy (SEM) imagery is critical to nuclear forensics, but has traditionally relied solely on skilled analysts whose classification accuracy and procedures may vary widely. Existing morphology lexicology [1] provides standardization guidelines to aid analysts but cannot fully address analyst variability. Using a dataset of 1,906 SEM images across 13 unevenly distributed particle classes and 73 magnification levels, final accuracy between statistical and deep learning methods were compared to find the best classification techniques. Ultimately, the deep learning model achieved an impressive 82% accuracy (80% balanced accuracy) on a withheld test set. This …

Analyzing And Comparing Refinement Indicators For Rbf-Fd Adaptive Algorithms, Anders R. Johnson

Theses and Dissertations

Recent progress has been made in the development of collocation-based iterative algorithms that approximate solutions to PDEs. These algorithms rely on the ability to identify regions within a domain where a finer discretization is required. Such iterative algorithms are beneficial particularly when solution functions have highly localized behavior. This thesis proposes an indicator for node refinement that is constructed by approximating the forward error. This proposed indicator also helps to establish confidence in the accuracy of a given solution estimate. The proposed error estimator is theoretically examined and compared with contemporary refinement indicators. It is shown that an iterative algorithm, …

Emergency Response Digital Twin: Integrating Augmented Reality And Live Position Data With Simulation-Aided Decision-Making In Real-Time, Joseph Fuentes

Theses and Dissertations

With the growing use of simulation across industries, the digital twin remains an underexplored research area, particularly in emergency management and response. Its real-time updating capability is often overlooked due to the misconception that "digital twin" is merely a complex term for simulation. This paper highlights its distinctiveness through an evasion exercise involving two independent entities in a collocated environment. Using a highly integrated virtual environment (HIVE) and internet of things (IoT) devices, we link the physical system with an analytical simulation, demonstrating the impact of lag times in high-pressure scenarios. The computational model leverages agent-based modeling (ABM) and discrete-event …

Performing Requirements Specification And Analysis Through Open Generative Pre-Trained Transformers, Harvey J. Hurst

Theses and Dissertations

Every acquisition program begins with a requirement, and for those programs to succeed, robust requirements engineering (RE) must be implemented. RE encompasses eliciting, analyzing, specifying, and validating requirements—a critical process throughout a program's lifecycle. Despite its importance, RE faces challenges such as scope creep, ambiguity, redundancy, and inadequate automation support, often exacerbated by reliance on historical data. To address these issues, this thesis leverages advancements in Generative Technology, particularly large language models (LLMs) such as Generative Pre-Trained Transformers (GPTs). This research developed two GPT-based tools: the Single Requirement Analysis Tool and the Set of Requirements Analysis Tool. These tools were …

Early-Phase Cost Estimation For Department Of Defense Construction Projects Via Artificial Neural Networks: Implications And Policy Recommendations For Artificial Intelligence Integration, Kobe T. Thompson

Theses and Dissertations

The Department of Defense lost over 500 million dollars between 2016 and 2024, partially due to poor early cost estimates resulting in cost overruns. practice for cost estimation relied on parametric techniques that incorporate historical data, subject matter experts in cost estimating, and predictive software applications. The main motivation for this study was to assess the viability of artificial neural networks as a means of providing a more accurate cost estimate in the early design phases of a construction project. The dataset initially contained approximately 48,000 data points from a database of various Air Force projects, including maintenance, repair, minor …

Cloud One Migration Duration And Its Drivers, Grayson T. Hall

Theses and Dissertations

As modern warfare evolves with rapid technological advancements, cloud computing plays a critical role in managing the vast amounts of data required for real-time decision making, as well as enabling seamless organizational access to mission-critical programs and information from around the globe. Recognizing its importance, the Department of Defense (DoD) identified cloud computing as essential for maintaining the military’s technological edge. However, despite cloud computing’s strategic significance, the DoD faces challenges in successfully implementing department-wide cloud computing. In contrast, the Air Force’s cloud computing environment, Cloud One, is fully operational and has already integrated over 145 systems into its platform. …

Heterogeneous Boid Swarm Performance Under Environmental And Neighbor Communication Link Variability, Jonathan C. Oxborrow

Theses and Dissertations

Artificial swarms are of growing interest in numerous fields and use cases. As their utilization increases drones and robots with different capabilities will be required to coordinate for task completion thus creating heterogeneous swarms. Swarm individuals generally communicate with all neighbors inside their sensor range generating a significant amount of message traffic. Previous research of a heterogeneous group in a non-physical environment has shown that restricting communication to only one neighbor of each different capability maintained performance. This work applies that finding to a heterogeneous boid swarm with the addition of varied environmental conditions. The swarm is comprised of three …

Improving Smartphone Gnss Jammer Localization With Cloud-Based Environmental Occlusion Modeling, Glenn H. Jones

Theses and Dissertations

The advancement of Global Navigation Satellite System (GNSS) technology in modern smartphones has made these devices pervasive in both civilian and military applications. Although smartphone GNSS chipsets are more susceptible to jamming and spoofing than military grade hardware, smartphone networks offer an underutilized opportunity to detect and mitigate threats to position, navigation, and timing (PNT) services essential to the Department of Defense (DoD) and civilian first responders. Traditional methods for geolocating ground-based jamming sources using smartphone GNSS often fail in environments with dense vegetation or significant occlusions, resulting in substantial localization errors.

A Machine Learning/Deep Learning Investigation On Remote Manufacturing Machine State Classification, Ajeet S. Parmar

Theses and Dissertations

Determining the extent of manufacturing capabilities with respect to adversarial or hostile nations is a topic of significant importance to the Department of Defense. Manufacturing capabilities can serve as indications of a nation's industrial power and its economy of force in warfare. Remotely detecting machine operations via electromagnetic sensors may be possible via Deep Learning (DL) and Machine Learning (ML) algorithms. To predict machine states, sensor data is collected externally from a machine shop on a college campus to monitor the operating states of lathes and mills in individual and concurrent operation. Furthermore, several sensors are placed in various positions, …

Jamming-Tolerant Low-Rate Wireless Personal Area Network For Detection Sensor Networks, Michael A. Eddy

Theses and Dissertations

This research evaluates the impact of electronic warfare, particularly jamming, on an audio-based drone detection wireless sensor network (WSN) using Monte Carlo simulations. A six-node IEEE 802.15.4 network, with five edge nodes and a central sink, is tested against jamming probabilities ranging from 0-100% in 5% increments across 30 iterations per configuration. Results show that packet delivery ratio (PDR) degrades linearly at approximately 20% per jammed node, while detection performance often exceeds PDR. Even at 80% jamming, detection success rates remain above 57%, highlighting resilience despite network degradation. The study reveals that jamming effectiveness depends on node placement relative to …

Machine Learning Techniques To Predict Solar Particle Events And Radiation Of Aircrew, Haley Traub

Theses and Dissertations

Solar Particle Events (SPEs) are high-energy phenomena from the Sun that pose risks to technology, human health, and Air Force operations. Accurate prediction of SPEs exceeding 100 MeV is crucial for mitigating these risks. This thesis explores using Bayesian statistical models to predict such events, integrating prior knowledge from solar physics with the ability to update predictions based on new data. The research uses a dataset spanning three solar cycles (21–23) and incorporates attributes like flare fluence, peak flux, latitude, longitude, and class. Four Bayesian models (PyMC, Bnlearn, and two Dredge models) were compared to machine learning models. The Bayesian …

A Robust Framework For Graph Construction In Vision Graph Neural Networks, Ismael Elsharkawi

Theses and Dissertations

In Computer Vision, the method of representing an image has a profound effect on the performance of a model. Traditionally speaking, an image is treated as a grid of pixels and can be processed via Convolution Neural Net- works (CNN). An image can also be treated as a sequence of patches. Vision Transformers and MLP-Mixers (Multi-Layer Perceptron Mixers) are two types of models that process an image as a sequence. A more generic representation than grids and sequences would be graphs. That is why Vision Graph Neural Network (ViG) construct a graph for an image and process the image as …

Securing A Virtual Reality Classroom Using Unity, Okemute Samuel Idonor

Theses and Dissertations

This study examines how security features might be incorporated into a Unity created virtual reality (VR) classroom, with a particular emphasis on important concerns including user privacy, data security, and illegal access. The increasing usage of virtual reality (VR) technology in educational settings has made it crucial to ensure the security of these immersive systems, especially when it comes to safeguarding sensitive user data. Designing and implementing a secure VR classroom prototype that could successfully protect against common security concerns, such as unauthorized access and data breaches, was the main goal of this project. Multi-Factor Authentication (MFA), Role-Based Access Control …

Using Satellite Image Segmentation To Detect Trails, Jeremy Reynolds

Theses and Dissertations

This masters thesis proposes an innovative approach to satellite image segmentation by focusing on the detection and mapping of walking, hiking, and biking trails. The motivation behind this project comes from the underexplored area in segmentation techniques for trail identification and offers potential benefits for urban planning, environmental monitoring, and public health. The problem statement addresses the need for a model that can differentiate between various trail types and other natural or man-made elements. The project aims for efficiency and scalability in processing satellite imagery across different compute hardware. The work details several stages: researching existing segmentation techniques, specifically road …

Neural Congruency Contrastive Learning Framework Validation Using Artificially Created Eeg Data For Dyslexia Research, Jacqueline M. Torres

Theses and Dissertations

Analysis of electroencephalogram (EEG) recordings in children with dyslexia has been commonly used to explore the neural mechanisms underlying reading disorders. Yet, challenges such as low signal-to-noise ratio (SNR), high inter-subject and inter-trial variability, and the inherently multivariate nature of EEG signals hinder the isolation of neural components elicited during reading diagnostic tests. To mitigate these challenges, the neural-congruency analysis framework was recently proposed, leveraging traditional machine learning optimization methods to incorporate domain knowledge about the congruency of neural responses across participants (i.e., consistent neural responses among proficient readers). However, the application of deep learning techniques, specifically contrastive learning, remains …

Leveraging Distributed Semantics From Deep Learning Architectures For Literature-Based Discovery, Clint A. Cuffy

Theses and Dissertations

Literature-based discovery (LBD) is a scientific process that introduces methods to automatically identify novel insights between non-interacting sets of literature. To date, numerous statistical and machine learning-based methods have been applied in the biomedical domain to find treatments for diseases such as Raynaud's disease, Parkinson's disease, and Multiple Sclerosis. However, the lack of standardized practices and creation of bespoke methodologies produces a scenario where the adoption of LBD remains challenging in real-world systems. Our work addresses these concerns through the improvement of five critical areas: 1) error propagation within LBD's a priori dependent tasks, 2) exploring the integration of modern …

Neural Network Algorithm And Analysis For Multi-Label Ecg Data Classification, Akhil Raghava Kalal

Theses and Dissertations

Electrocardiogram (ECG) analysis is a fundamental diagnostic tool in cardiology, providing critical insights into cardiac function that directly impact patient care decisions and treatment outcomes. As healthcare systems face increasing demands, automated ECG interpretation using artificial intelligence offers promising solutions to improve diagnostic accuracy, reduce physicians workload, and enhance early detection of life threatening conditions.

This thesis compares two advanced deep learning architectures, CNN-GRU and Wide and Deep Transformer, for multi-label classification of 12-lead ECG data. Using data from the PhysioNet/Computing in Cardiology Challenge 2020, I evaluated both architectures across 27 different cardiac abnormalities. Results demonstrated that CNN-GRU architecture consistently …

The Impact Of Loss Function Topology On Gradient Descent, Robert B. Skudnig Jr.

Theses and Dissertations

Gradient descent is a popular optimization method that utilizes a model’s prediction error to iteratively improve its parameters for a given task. The functions that measure this error can be defined to align with the user’s goals and sometimes satisfy metric or norm properties. It is common for these functions to measure over Rn, but any differentiable space allows for gradient descent to occur. There has been some research investigating the influence of topological spaces on optimization methods, but it is a limited field of study. This thesis further explores this phenomenon by applying a transformation prediction model to multiple …