Artificial Intelligence and Robotics Commons^™

Observations On Recurrent Loss In The Neural Network Model Of A Partial Differential Equation: The Advection–Diffusion Equation, Jonah A. Reeger

Faculty Publications

A growing body of literature has been leveraging techniques of machine learning (ML) to build novel approaches to approximating the solutions to partial differential equations. Noticeably absent from the literature is a systematic exploration of the stability of the solutions generated by these ML approaches. Here, a recurrent network is introduced that matches precisely the evaluation of a multi-step method paired with a collocation method for approximating spatial derivatives in the advection–diffusion equation. This allows for two things: (1) the use of traditional tools for analyzing the stability of a numerical method for solving PDEs and (2) bringing to bear …

Does Patient History Influence Capsular Contracture? An Exploratory Analysis With Machine Learning, Thomas M. Johnstone, Daniel Najafali, Jennifer K. Shaw, Justin M. Camacho, Chancellor Johnstone, Rahim S. Nazerali, Gordon K. Lee

Faculty Publications

Background: Capsular contracture (CC) is a frequent and distressing complication of breast augmentation and reconstruction. Although numerous patient-, surgical-, and implant-related risk factors have been proposed, reliable population-level predictors remain inconsistent across studies. This study evaluates whether administrative medical history, as encoded by ICD and CPT codes, contains sufficient predictive signal to identify patients at risk for CC using machine learning. Methods: Patients were queried from the MerativeTM MarketScan® Research Databases from 2003 to 2017 with CPT codes for implant-based breast reconstruction and augmentation. ICD codes were then used to identify all events and conditions of a patient’s history. Hyperparameter-tuned …

Amp: Single-Shot Ultra-Wide Fisheye-To-Cubemap Pnp Pose Estimation, Ryan M. Raettig, Richard R. Nyquist, Scott L. Nykl, Clark N. Taylor, Christine M. Schubert Kabban

Faculty Publications

Estimating the position and orientation of a rigid object from an image is critical for situational awareness in robotics and autonomous systems. This study explores relative pose estimation using an ultra-wide fisheye camera for unmanned aircraft inspection vehicles. Ultra-wide fisheye lenses introduce radial distortion and capture features beyond the rectilinear image plane, rendering rectilinear Perspective-n-Point (PnP) algorithms inadequate. Designing a bespoke ultra-wide fisheye localization algorithm requires consideration of both the feature detection method and the pose estimator itself. This study proposes a novel method that combines (1) a fisheye-to-cubemap reprojection, (2) a You Only Look Once (YOLO) convolutional neural network …

Applying Machine Learning Methods To Laser Acceleration Of Protons: Synthetic Data For Exploring The High Repetition Rate Regime, John J. Felice, Ronak Desai, Nathaniel Tamminga, Joseph R. Smith, Alona Kryshchenko, Christopher M. Orban, Michael L. Dexter, Anil K. Patnaik

Faculty Publications

Advances in ultra‐intense laser technology have increased repetition rates and average power for chirped‐pulse laser systems, which offer a promising solution for many applications including energetic proton sources. An important challenge is the need to optimize and control the proton source by varying some of the many degrees of freedom inherent to the laser‐plasma interactions. Machine learning can play an important role in this task, as our work examines. Building on our earlier work in Desai et al. 2024, we generate a large ∼1.5 million data point synthetic data set for proton acceleration using a physics‐informed analytic model that we …

Exact And Approximate Conformal Inference For Multi-Output Regression, Chancellor Johnstone, Eugene Ndiaye

Faculty Publications

It is common in machine learning to estimate a response y given covariate information x . However, these predictions alone do not quantify any uncertainty associated with said predictions. One way to overcome this deficiency is with conformal inference methods, which construct a set containing the unobserved response with a prescribed probability. Unfortunately, even with a one-dimensional response, conformal inference is computationally expensive despite recent encouraging advances. In this paper, we explore multi-output regression, delivering exact derivations of conformal inference p-values when the predictive model can be described as a linear function of y . Additionally, we introduce a multivariate …

Feasibility Evaluation Of Secure Offline Large Language Models With Retrieval-Augmented Generation For Cpu-Only Inference, Erick Tyndall, Torrey J. Wagner, Colleen Gayheart, Alexandre Some, Brent T. Langhals

Faculty Publications

Recent advances in large language models and retrieval-augmented generation, a method that enhances language models by integrating retrieved external documents, have created opportunities to deploy AI in secure, offline environments. This study explores the feasibility of using locally hosted, open-weight large language models with integrated retrieval-augmented generation capabilities on CPU-only hardware for tasks such as question answering and summarization. The evaluation reflects typical constraints in environments like government offices, where internet access and GPU acceleration may be restricted. Four models were tested using LocalGPT, a privacy-focused retrieval-augmented generation framework, on two consumer-grade systems: a laptop and a workstation. A technical …

Impact Of Retrieval Augmented Generation And Large Language Model Complexity On Undergraduate Exams Created And Taken By Ai Agents, Erick S. Tyndall, Colleen Gayheart, Alexandre Some, Joseph Genz, Torrey J. Wagner, Brent T. Langhals

Faculty Publications

The capabilities of large language models (LLMs) have advanced to the point where entire textbooks can be queried using retrieval-augmented generation (RAG), enabling AI to integrate external, up-to-date information into its responses. This study evaluates the ability of two OpenAI models, GPT-3.5 Turbo and GPT-4 Turbo, to create and answer exam questions based on an undergraduate textbook. 14 exams were created with four true-false, four multiple-choice, and two short-answer questions derived from an open-source Pacific Studies textbook. Model performance was evaluated with and without access to the source material using text-similarity metrics such as ROUGE-1, cosine similarity, and word embeddings. …

Computer-Automated Systems And Methods For Using Language Models To Generate Text Based On Reading Errors, Scott Sosso, Siyu Chen, Ciara Figliuolo, Jack Mostow, Marlies Goes

AFIT Patents

A computer-implemented system and method generate personalized text based on statistics derived from input received from a user representing the user's attempts to decode graphemes into phonemes. Such statistics may be measured and recorded at the grapheme-phoneme level, and may include substitutions, insertions, deletions, and correct utterances of phonemes by the user when reading text. A language model may be trained based on characteristics of the user, such as the user's age and/or reading grade level, and the personalized text may be generated after such training of the language model. Generating the personalized text may include generating a text creation …

Contract Quality Feature Extraction Using Llm, Aaron C. Washington

Theses and Dissertations

This study explored the potential insights generated from linguistic complexity measurements and large language model (LLM) based assessments on the quality of contract documents. By combining structured True/False prompts with log-probability analysis and ambiguity scoring, the study introduced novel contract-quality assessment methods. Results support a feature-driven approach to contract evaluation, one that offers automated, scalable insights for triaging risk and improving drafting practices. These assessment methods contribute to the growing field of legal natural language processing by offering modular tools for effective contract analysis.

Global Sporadic-E Prediction And Climatology Using Deep Learning, J. A. Ellis, Daniel J. Emmons, M. B. Cohen

Faculty Publications

Sporadic-E (E_s) is an ionospheric phenomenon defined by strong layers of plasma which may interfere with radio wave propagation. In this work, we develop deep learning models to improve the understanding of E_s, including the presence, intensity and height of the layers. We developed three separate models. The first, building off earlier work in (J. A. Ellis et al., 2024, link in AFIT Scholar, 10.1029/2023sw003669), includes only the main features from radio occultation (RO) measurements. The second adds to that time, date, location, geomagnetic and solar indices, solar winds, x-ray flux, weather and lightning. A …

Exploring Emotion Classification Of Indonesian Tweets Using Large Scale Transfer Learning Via Indobert, Connor Shaw, Phillip M. Lacasse, Lance E. Champagne

Faculty Publications

Business, political, and other social structures create strong motivation to understand the attitudes, motivations, feelings, and emotions of a population of interest. Social media is a rich source of self-disclosed information by individuals from all walks of life about virtually every domain of the human experience, but the vast quantity of data is impossible to effectively analyze without advanced natural language processing algorithms. This research creates a transfer learning based emotion classification model for Indonesian language Twitter data. Transfer learning consists of two steps: pre-training and fine tuning. Three variations of Indonesian Bidirectional Encoder Representations from Transformers (IndoBERT) are tested …

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 …

Accelerated Multiobjective Calibration Of Fused Deposition Modeling 3d Printers Using Multitask Bayesian Optimization And Computer Vision, Craig S. Ganitano, Benji Maruyama, Gilbert L. Peterson

Faculty Publications

Proper process parameter calibration is critical to the success of fused deposition modeling (FDM) three-dimensional (3D) printing, but is time-consuming and requires expertise. While existing systems for autonomous calibration have demonstrated success in calibrating for a single objective, users may need to balance multiple conflicting objectives. Herein, an easily deployable, camera-based system for autonomous calibration of FDM printers that optimizes for both part quality and completion time is presented. Autonomous calibration is achieved through a novel, multifaceted computer vision characterization and a multitask learning extension to Bayesian optimization. The system is demonstrated on four popular filament types using two distinct …

Afit Generative Ai Teaching Guidebook Synopsis, Afit Faculty Learning Community, Mark G. Bateman, Brett J. Borghetti, Allen W. Dukes, Nicholas C. Francis, Mike Frick, Bobbie Oh, Kevin Patterson, Hiren J. Patel, Mark G. Reith, Erick S. Tyndall, Teresa M. Walton, Torrey J. Wagner, Timothy S. Wolfe, Jonathan Zemmer

AFIT Documents

The main objective of this work was to bring together various perspectives on how to envision incorporating Gen AI capabilities into the learning environment and identify some best practices for their implementation. Any instructor who is interested in these capabilities but does not necessarily have a technical background can find pragmatic use of the examples provided. While the examples have a wide range of applicability, they are meant to serve as a starting point for educators to explore what would be beneficial to their educational environment, from traditional classroom settings to online continuing education courses.

Editorial: Artificial Intelligence For Smart Health: Learning, Simulation, And Optimization, Bing Yao, Nathan Gaw, Hyo Kyung Lee

Faculty Publications

With rapid developments in medical sensing and imaging, we now live in an era of data explosion in which large amounts of data are readily available in clinical environments. The fast-growing biomedical and healthcare data provide unprecedented opportunities for data-driven scientific knowledge discovery and clinical decision support. Our Research Topic aims to catalyze synergies among biomedical informatics, machine learning, computer simulation, operations research, systems engineering, and other related fields with three specific goals: (1) develop cutting-edge data-driven models to accelerate scientific knowledge discovery in biomedicine using healthcare data collected from laboratory systems, imaging systems, and medical and sensing devices; (2) …

Personalized Learning Path Problem Variations: Computational Complexity And Ai Approaches, Sean A. Mochocki, Mark Reith, Brett J. Borghetti, Gilbert L. Peterson, John Jasper, Laurence D. Merkle

Faculty Publications

E-learning courses often suffer from high dropout rates and low student satisfaction. One way to address this issue is to use personalized learning paths (PLPs), which are sequences of learning materials that meet the individual needs of students. However, creating PLPs is difficult and often involves combining knowledge graphs (KGs), student profiles, and learning materials. Researchers typically assume that the problem of creating PLPs belong to the nondeterministic polynomial (NP)-hard class of computational problems. However, previous research in this field has neither defined the different variations of the PLP problem nor formally established their computational complexity. Without clear definitions of …

Using Machine Learning To Predict State Compliance With International Legal Obligations For Registration Of Space Objects: Comparative Performance Of Logistic Regression And Dense Neural Network Models, Jonathan K. Sawmiller

Student Publications

Approximately 12% of satellites and other objects launched into outer space have not been registered with the United Nations (UN) as required by international law. To predict whether States will register a launched space object and understand what factors influence a registration decision, data from a UN online index of space objects was used to train and select the best machine learning model. After preparation, the dataset had 1938 datapoints with 11 features, with categorical features simplified and converted to binary.

Multiple variations of classical logistic regression models were compared to multiple variations of dense neural network models. The best …

Coarse-Gridded Simulation Of The Nonlinear Schrödinger Equation With Machine Learning, Benjamin F. Akers, Kristina O. F. Williams

Faculty Publications

A numerical method for evolving the nonlinear Schrödinger equation on a coarse spatial grid is developed. This trains a neural network to generate the optimal stencil weights to discretize the second derivative of solutions to the nonlinear Schrödinger equation. The neural network is embedded in a symmetric matrix to control the scheme’s eigenvalues, ensuring stability. The machine-learned method can outperform both its parent finite difference method and a Fourier spectral method. The trained scheme has the same asymptotic operation cost as its parent finite difference method after training. Unlike traditional methods, the performance depends on how close the initial data …

Machine Visual Perception For Autonomous Docking Maneuvers, Derek B. Worth

Theses and Dissertations

This dissertation presents a novel approach to autonomous docking using machine learning for visual perception, particularly during probe and drogue aerial refueling. Autonomous vehicles have become pervasive in both civilian and defense sectors, and their ability to interact with their surroundings and each other autonomously is critical for future operations. Traditional methods relying on signals or inertial sensors face significant limitations such as interference, jamming, and drift. This research focuses on developing a computer vision-based solution to overcome these limitations. A novel pipeline, termed relative vectoring, is introduced, which utilizes dual object detection and machine learning to estimate relative positions …

Data Supporting Research On Personalized Learning Paths, Sean Mochocki, Mark Reith

Faculty Publications

Personalized Learning Paths (PLPs) are a key application of Artificial Intelligence in E-Learning. In contrast to regular Learning Paths, they return a unique sequence of learning materials identified as meeting the individual needs of the students. In the literature, PLPs are often created from knowledge graphs, which assist with ordering topics and their associated learning materials. Knowledge graphs are typically directed and acyclic, to capture prerequisite relationships between topics, though they can also have bidirectional edges when these prerequisite relationships are not necessary. This data package provides a primarily un-directed knowledge graph, with associated repository of open-source learning materials that …

Measuring The Presentation Of Supporting Content For A Set Of Learning Objectives Throughout The States Of An Educational Game Tree, Michael L. Hastriter Jr.

Theses and Dissertations

In an era of evolving warfare, the Department of Defense (DoD) recognizes the value of serious games as immersive tools for teaching critical concepts. This thesis introduces a pioneering framework tailored to enhance learning objectives through the presentation of educational game content. This addresses the unique needs of the DoD and other educators who use games by providing measurements to assess educational games. This researches investigates how instructors might assess games as potential teaching tools. It establishes a five-phase process, providing a framework to assess educational games against predefined learning objectives and informing future game development. This thesis demonstrates games …

Opportunities And Limitations: Integrating Narrative Ai Into Game-Based Assessment Creation And Evaluating Student Impacts, Kevin B. Patterson

Theses and Dissertations

The Department of Defense (DoD) has identified the need for a technically proficient workforce in the areas of science, technology, engineering, and mathematics. To meet this need, the DoD is actively seeking innovative technology capable of creating workforce development opportunities that are both accessible and effective. Educational research indicates serious games provide a potential avenue to achieve this goal. Unfortunately, a limited number of tools that simplify the game development process and leverage artificial intelligence are available. Content-generating artificial intelligence might help reduce instructor and game-based assessment designers' workloads while promoting individualized learning in students. This research presents a novel …

Reinforcement Learning For Team Based Air Combat Maneuvering Decisions With Directed Energy Weaponry, Joshua D. Combs

Theses and Dissertations

Leveraging the Advanced Framework for Simulation, Integration, and Modeling (AFSIM) we investigate the use of reinforcement learning (RL) techniques for imbuing AUCAV agents with high-quality behaviors for the within-visual-range air combat maneuvering problem (ACMP). We formulate the 2v2 WVR ACMP as a Markov decision process wherein friendly AUCAVs are equipped with DEW capabilities and operate with 6 degrees of freedom. We utilize the Double Deep Q-Network RL algorithm, which centrally trains two friendly AUCAVs and employ a phased learning approach, initially exposing the AUCAVs to a dense reward environment for early training, followed by a sparse reward environment to encourage …

Artificial Intelligence And Perception: An Empirical Study, Anthony J. Neil

Theses and Dissertations

This thesis investigates the impact of adjusting artificial intelligence explainability levels’ outputs on user perception. The overarching study extends within the Explainable Artificial Intelligence (XAI) domain. It examines a spectrum of variables, including performance, cognizance, familiarity, transparency, system bias, and the overall impact of AI, to understand their collective and individual effects that enable effective professional use in an organization. The study aims to illuminate the relationship between the degree of explainability provided by large language models such as ChatGPT, Bard, and Bing AI and the performance of these models when tasked with XAI adjustments.

Standardization Of Risk Classifications For Unmanned Space Vehicle Missions, Collin A. Gwaltney

Theses and Dissertations

This paper seeks to model risk classification levels (A-D) for 122 Space Vehicle programs. Models include multinomial logistic regression as well as random forest, a machine learning technique based on decision trees. We use independent variables (IVs) which are theoretically correlated to risk class for the regression and one random forest model. We then include all IVs and allow the random forest technique to use those which provide the most information on risk class before paring down the number of IVs to only 7. We show that the accuracy of predictions increases from 62% to 87% by using random forest …

Random Forests For Detecting Weak Signals And Extracting Physical Information: A Case Study Of Magnetic Navigation, Mohammadamin Moradi, Zheng-Meng Zhai, Aaron Nielsen, Ying-Cheng Lai, Aaron P. Nielsen

Faculty Publications

It has been recently demonstrated that two machine-learning architectures, reservoir computing and time-delayed feed-forward neural networks, can be exploited for detecting the Earth’s anomaly magnetic field immersed in overwhelming complex signals for magnetic navigation in a GPS-denied environment. The accuracy of the detected anomaly field corresponds to a positioning accuracy in the range of 10–40 m. To increase the accuracy and reduce the uncertainty of weak signal detection as well as to directly obtain the position information, we exploit the machine-learning model of random forests that combines the output of multiple decision trees to give optimal values of the physical …

A Random Forest-Based Q-Learning Algorithm: Toward Interpretable Artificial Intelligence, Victor R. Rae

Theses and Dissertations

A growing demand exists for interpretable artificial intelligence models, leading to extensive research efforts to enhance the explainability and transparency of policies generated by reinforcement learning (RL) methods. This research develops random forest-based RL algorithms as a logical progression in this academic pursuit. The algorithms are evaluated using three standard benchmark environments from OpenAI gym — CartPole, MountainCar, and LunarLander — and compared to implementations of the Deep Q-learning Network (DQN) and Double DQN (DDQN) algorithms for various metrics, including performance, robustness, efficiency, and interpretability. The random forest-based algorithms exhibit superior performance to both neural network-based algorithms in two out …

A Model-Based And System-Theoretic Approach To Design Advanced Autonomy For Air Dominance Missions: A Loyal Wingman Case Study, Elizabeth S. Pennington

Theses and Dissertations

In contested air environments, safe coordination between decision-makers is paramount. Although the Department of Defense (DoD) prioritizes the development of Artificially Intelligent (AI) wingmen for air combat, a lack of methodology exists to design safe, holistic coordination between human and autonomous wingmen in the same environment. This thesis delivers a framework using Systems Theoretic Process Analysis Extended for Coordination (STPA-Coord) to analyze and design holistic coordination for the Loyal Wingman concept in an Air Dominance mission. STPA-Coord is a safety and hazard analysis process that uses Systems Theory to analyze and design coordination between decisionmakers in a system-of-systems architecture. Using …

Natural Language Processing Analysis Of Online Reviews For Small Business: Extracting Insight From Small Corpora, Benjamin J. Mccloskey, Phillip M. Lacasse, Bruce A. Cox

Faculty Publications

Receiving and acting on customer input is essential to sustaining and growing any service organization, particularly a small family business whose livelihood depends on strong relationships with its customers. The competitive advantage offered by advanced analytical approaches for supporting decisions is not trivial, and enterprises across virtually all domains of society are investing heavily in this emerging discipline. Natural Language Processing (NLP) is a subset of computer science that employs computational approaches to analyze human language; it is effective at extracting insight from text data but frequently requires large corpora to train its models, in the scale of thousands or …