Physical Sciences and Mathematics Commons^™

Interplanetary Trajectory Optimization With Reinforcement Learning, Shiloh Cuffe

Discovery Day - Daytona Beach

This project investigates the application of reinforcement learning (RL) to optimize low-thrust interplanetary trajectory design, focusing on the Earth-Venus transfer leg of the BepiColombo mission. Traditional trajectory optimization methods, such as patched conics and genetic algorithms, often require simplifying assumptions or complex optimization schemes. This work formulates the trajectory design problem as an optimal control problem (OCP) within a Markov Decision Process (MDP) framework, enabling an RL agent to learn efficient transfer strategies under realistic spacecraft constraints. The objective is to develop an autonomous guidance approach capable of replicating or improving upon established mission designs. The spacecraft is modeled as …

Feasibility Of High-Throughput Onboard Ai For Mars Rovers Under Solar Constraints, Aashman Gupta

Discovery Day - Daytona Beach

This project evaluates the feasibility of sustained onboard AI autonomy for a solar-powered Mars rover by directly linking solar energy availability to achievable compute performance. While Mars solar irradiance and edge computing performance have been studied independently, no unified framework currently couples surface power generation to autonomy throughput in an experimentally validated manner. The project will begin with a simulation of solar power generation for a 1 m² rover-mounted array across a Martian sol, accounting for seasonal variation, dust opacity, and array configuration (fixed versus sun-tracking). The resulting power profile will then be coupled to representative compute platforms running autonomy …

An Energy-Aware Meta-Learning Framework For Real-Time Lunar Rover Localization Via Adaptive Algorithm Selection, Jose Demedeiros, Garrett Seyler

Discovery Day - Daytona Beach

This work proposes an energy-aware meta-learning framework that selects the single most suitable localization algorithm for a lunar rover, per scene, using only monocular imagery and orbital maps. The goal is to achieve sub-meter accuracy while minimizing onboard compute and energy consumption. We assemble a suite of seven lunar-relevant algorithms spanning relative and absolute localization, including monocular ORB-SLAM3, LuVo homography-based visual odometry, Censible cross-view matching with orbital imagery, crater-based methods (LunarNav and ShadowNav), monocular horizon navigation with a DEM, and DROID-SLAM. Relative methods provide incremental motion updates, while absolute methods deliver global pose fixes; an Extended Kalman Filter fuses these …

A Qualitative Analysis Of Human-Ai Interaction Through Animated Shapes, Kavya Dipen Shah, Angel Hinojosa, Caroline Deck

Discovery Day - Daytona Beach

As technology becomes more advanced, it is important to understand how people perceive the intentions and abilities of machines. This project, conducted in the InTeRACT Lab, explores how we attribute humanlike qualities to different types of agents, ranging from animals to robots. The study analyzes data from an experiment where participants watched animations of moving triangles. Although the videos were identical, participants were told the shapes represented either humans, robots, dogs, or inanimate objects. While previous math-based data showed that these labels changed how people felt, those structured scales didn't allow for a natural, unbiased explanation of what people actually …

A System Safety Approach To Assuring Artificial Intelligence Enabled Functions In Civil Aviation, Evan Bear, Quinn Galen

Discovery Day - Daytona Beach

Artificial intelligence and machine learning techniques are increasingly proposed for use in safety-critical civil aviation functions including perception decision support and pilot assistance. Existing aviation safety and certification standards such as ARP4754A and DO-178C were developed under assumptions of determinism explicit requirements and complete behavioral specification which do not directly apply to learning-enabled systems. This mismatch has created uncertainty regarding how artificial intelligence enabled avionics can be safely assured and certified. This paper presents a system safety approach for assuring artificial intelligence enabled functions within existing aviation certification frameworks. In this approach safety assurance is based on explicitly identifying the …

A Hybrid Llm-Srgm Framework For Ai-Enabled Reliability Assessment In Safety-Critical Software Systems, Caleb Stone, Shrenik Jadhav

Discovery Day - Daytona Beach

Ensuring the reliability of software intensive and safety critical systems is a persistent challenge across aerospace, defense, transportation, and other mis- sion focused domains. Traditional software relia- bility growth models (SRGM) provide useful quanti- tative insight into defect discovery trends, but they rely mostly only on numerical failure data and do not use the rich contextual information contained in test logs, anomaly reports, and engineering notes. This paper presents a hybrid framework that com- bines semantic features extracted by a large lan- guage model (LLM) with a non-homogeneous Pois- son process (NHPP) based software reliability growth model. The LLM analyzes …

Cars Imass - Comparing Llm Vs Human Operator Effectiveness In Multi-Agent Swarm Coordination, Gatlin Nelson

Discovery Day - Daytona Beach

Title: Dual-Perspective Risk Analysis for Human-LLM Decision Comparison in UAV Swarm Navigation   Unmanned aerial vehicle (UAV) swarms operating in low-altitude wireless network environments encounter localized disruptions that degrade positioning and navigation metrics. These disruptions are modeled as geographic failure zones with defined boundaries. A UAV discovers a zone by entering it and observing degraded performance on its onboard systems. This work assumes that affected UAVs can autonomously retreat to safety using onboard sensors and focuses on the subsequent rerouting decision. Once recovered, the system generates candidate repositioning points surrounding the vehicle, each scored using Conditional Value-at-Risk (CVaR). A human operator …

Social Attributions Of Moving Shapes: Comparing Qualitative Analyses Of Humans Vs. Ai, Angel Hinojosa

Discovery Day - Daytona Beach

As technology becomes more intelligent, the relationship between humans and machines is rapidly shifting. Whether a machine is perceived as a capable partner or a source of wariness often depends on the intentions and abilities, we attribute to it. My research in the InTeRACT Lab seeks to empirically assess these perceptions by comparing how we view different nonhuman agents, ranging from animals to robots.   This study analyzes qualitative data from a task where participants viewed animations of moving triangles. While the videos remained the same, participants were told the shapes represented either humans, robots, dogs, or inanimate shapes. Previous quantitative …

A Survey On Machine Learning Applications For Operating System Fingerprinting, Siri Siqveland

Discovery Day - Daytona Beach

In the modern age of computers and interconnected networks, cybersecurity and cyber-attackers are evolving in tandem to exploit each other’s vulnerabilities. One technique used by both parties is Operating System Fingerprinting (OSF): with the knowledge of what Operating System a target system is running, innate vulnerabilities can be identified and patched or exploited. Historically, OSF utilizes two main methods: passive and active—the former trades accuracy with undetectability while the latter is generally more detectable but more accurate. However, recent work has combined OSF with Machine Learning (ML) to improve accurate identification. The work presented here is a survey for the …

Phaëthon System, Brady Roudabush, Lauren Gallo, Emelia Thompson, Jacob Woods

Discovery Day - Daytona Beach

Phaëthon System is the project name for the Search and Rescue Drone Initiative. This initiative will improve the current search and rescue drone industry by introducing new techniques to get through dense forest canopies and other places where an overhead view is not useful. The Phaëthon System uses a swarm of drones that can penetrate under the tree canopy to map and search with the utmost efficiency and safety for rescuers. A command drone is launched to survey the overall search area, and set up a communications and data link. The next component is then released, which is a swarm …

Stress-Triggered Automation Reliance, Jazmin Elek

Discovery Day - Daytona Beach

Automation is widely used in complex systems and includes any process that replaces human motor, sensory, or cognitive functions with machines or computers (Norman, 1996). As automation becomes more common, understanding how humans trust and interact with these systems is critical. Trust can be measured by whether users override automation or blindly follow its prompts (Norman, 1996).   Artificial intelligence (AI) introduces additional complexity by enabling systems to learn patterns from data it generates. AI performs tasks with the ability to learn from experience (NASA, 2024). AI builds internal databases that can mimic human-like responses (Norman, 1996). However, AI systems can …

Stormtrack: A Regime-Aware Classifier-Router Architecture For Multi-Horizon Kp Index Forecasting, John Rendleman

Discovery Day - Daytona Beach

STORMTRACK: A Regime-Aware Classifier-Router Architecture for Multi-Horizon Kp Index Forecasting Current algorithms in operational space weather face extreme difficultly predicting the Kp geomagnetic index beyond 24 hours, a lead time that is critical for protecting high-frequency communications and infrastructure. Most regression models are optimized for quiet conditions, which dominate the data, leading to systematic underpredictions of storm events that cripple space infrastructure. Probabilistic approaches and physics-based numerical models also falter due to the same class imbalance plaguing standard regressors at multi-day lead times. The ICARUS 6 architecture addresses this by splitting the forecasting component into quiet and storm regimes, which …

Accelerating Search And Rescue Response: A Simulation Study On The Dynamic Efficiency Of Flocking-Enabled Drone Swarms, Sophia Beckwith, Carys Del Prete

Discovery Day - Daytona Beach

This project explores how imitations observed in animal group behavior, specifically flocking in birds, can be applied to the functionality of autonomous drone systems to aid in search and rescue efforts. The goal is to demonstrate how incorporating code based on the Boids, Vicsck and predictive control linear algebraic mathematical models for drone flight controls and the collective behaviors of flocks will increase the efficiency of drone maneuvers, allowing them to reorganize and fill gaps when one is removed. A MATLAB-based simulation was developed to model the behaviors using research conducted on the symmetric and synchronized behaviors observed from flocks …

Measuring The Tokenization Premium: A Cost Audit For Underserved Language Communities, Avijit Roy, Proma Roy, Hrishitva Patel

Publications and Research

Large language models are increasingly deployed as general-purpose educational and technical assistance systems, but their basic infrastructure does not treat languages equally. One underexamined source of disparity is tokenization: semantically equivalent content can require substantially different token counts across languages, affecting API cost, latency, and usable context length before a model is even invoked. We introduce the Tokenization Equity Audit (TEA), a reproducible benchmark for measuring tokenization premiums in technical tutoring content. TEA evaluates three widely used tokenizers, GPT-4o’s o200k base, Qwen2.5-7B, and Mistral-7B, on a 120-item Python debugging corpus translated from English into Bengali, Hindi, Arabic, Tamil, and Yoruba. …

Demonstrating Superresolution In Radar Range Estimation Using A Denoising Autoencoder, Robert Czupryniak, Abhishek Chakraborty, Andrew N. Jordan, John C. Howell

Mathematics, Physics, and Computer Science Faculty Articles and Research

We apply machine learning methods to demonstrate radar range superresolution using a denoising autoencoder trained without supervision. Focusing on the estimation of a single physical parameter, the separation between two scatterers in the subwavelength regime, we constrain the network to a one-dimensional bottleneck layer with its size matched to the parameter dimensionality. We find that the bottleneck layer forms a reproducible, monotonic mapping with the true separation, showing that the network learns a low-dimensional representation directly aligned with the underlying physical parameter. We further show that this representation preserves the Fisher information of the signal, indicating that the network recovers …

Learning To Evaluate: Cost-Effective Model Evaluation On Unlabeled Data With Meta-Learning, Trinh Pham, Viet Huynh, Hongzhi Yin, Quoc Viet Hung Nguyen, Thanh Tam Nguyen

Research outputs 2022 to 2026

The rapid advancement of machine learning has led to an unprecedented expansion of model ecosystems, making it increasingly difficult to assess the reliability of newly released models on unseen and unlabeled data. Existing evaluation pipelines typically rely on costly annotation, repeated fine-tuning, or assumptions that do not generalize well to new models. We introduce MetaEvaluator, a cost-effective, model-agnostic framework for fast, label-free evaluation of unseen models across diverse architectures and modalities. MetaEvaluator meta-learns over a pool of reference models to acquire an effective initialization for accurate assessment of unseen models, thereby amortizing evaluation cost and eliminating the need for per-model …

South Africa's Inclusive Six-Dimensional Generative Ai Readiness Framework, Keegan Steyn, Ojelanki Ngwenyama, Sibongiseni Tunzelana

African Conference on Information Systems and Technology

Generative artificial intelligence (Gen AI) is transforming governance, knowledge systems, and economic activity, but current preparedness measures pay little attention to the socio-technical, institutional, and postcolonial contexts that shape country trajectories. To address this gap, the study employs a design science methodology to create the National Generative AI Readiness Assessment Framework (N-GAIRF), a six-dimensional diagnostic tool theoretically based on Sen's capability approach to development and incorporating Institutional Theory, Socio-Technical Systems Theory, National Innovation Systems Theory, and Postcolonial Information Systems critique. An example application to South Africa, compared to Zimbabwe and an indicative Organization for Economic Cooperation and Development (OECD) reference …

Clinical Utility Of An Fda-Authorized Artificial Intelligence Imaging Platform In Interstitial Lung Disease Diagnosis, Arjun Prakash Tambe, Ryan Boente, Gautam George, Fayez Kheir, Omid Tahamtani Omran, Kavitha Selvan

Division of Pulmonary, Allergy, and Critical Care Medicine Faculty Papers

Background/Objectives: The diagnosis of interstitial lung disease (ILD) is challenging and frequently delayed. Clinically accessible and minimally invasive diagnostic tools are needed to expedite the diagnosis of ILD while minimizing risk to patients. Fibresolve is an imaging artificial intelligence (AI) tool recently approved by the Food and Drug Administration (FDA) for use in ILD diagnosis and made available to clinicians. The objective of this study was to describe its utility in clinical practice. Methods: We conducted a prospective, observational study of patients across the United States (US) in whom Fibresolve was utilized during routine clinical practice between July 2024 and …

Dynamind: A Dynamic Learned Index For Update-Intensive Workloads, Jingxian Cheng, Yingfang Wang, Tianqing Zhu, Xu Yang, Ningning Cui, Jianxin Li

Research outputs 2022 to 2026

Learned indexes leverage machine learning models to approximate data distributions and predict key positions, offering better performance than traditional index structures such as B+Trees. As data in real-world applications evolve rapidly, the timely and efficient updating of learned indexes has become an increasingly important research problem, attracting growing attention in recent studies. However, under update-intensive workloads with frequent insertions and deletions, existing learned indexes cannot update the model in a timely manner. Moreover, they ignore the impact of deletions on model accuracy. These limitations lead to degraded prediction accuracy and increased query latency, undermining the core advantage of learned indexes. …

Ai And The Music Industry: Its Current Status And A Speculative Projection Of Its Evolutionary Trajectory, Rhett D. Morris, Clayton Rosati, Stefan Fritsch

Honors Projects

Music serves as one of society's biggest cultural outlets, allowing millions to share in what used to be a uniquely human form of expression. The commodification of music has built a huge industry full of companies and platforms that have used technology and property laws to shape music's relationship with the public. This study aims to look into the future to see how AI and its implementation could affect the structure of the music industry. To look into the future, this piece establishes two of the most pressing kinds of AI technology for the music industry and looks to contextualize …

Sentinel: Evaluating Occlusion-Centered Next-Best-View Selection Using Rgb-Derived Pseudo-Geometry, Paul Nassar

Master's Theses

Three-dimensional cameras provide direct geometric measurements, but their cost, weight, power requirements, and calibration constraints can limit their use in various lightweight or large-scale sensing systems. A potential alternative is to use conventional two-dimensional RGB cameras together with geometric reconstruction models that infer a partial three-dimensional representation from images. This thesis evaluates that possibility for next-best-view (NBV) selection through Sentinel, an occlusion-centered system for static, object-centric scenes with known camera poses and intrinsics. Sentinel converts source RGB observations into pseudo-geometry using monocular depth or point-map predictions, combines those predictions with camera-ray evidence, identifies occluded unknown regions, and selects a candidate …

Enabling Multi-Task Neural Network Inference On Heterogeneous Edge Devices, Redwanul Islam Arif

Master's Theses

Deep neural networks are increasingly required to run on the devices that generate the data. If such a device must perform more than one task, the standard practice is deploying one model per task, which makes memory grow linearly with task count, which is unacceptable when the entire budget is kilobytes. This thesis asks one question in three settings: how much capability can a network acquire without incurring deployment cost?

The first study takes an ImageNet-pretrained ResNet-18, sweeps the branch point across every residual stage and the classification-head depth across one, ten, and twenty layers, and deploys the resulting multi-head …

Leveraging Artificial Intelligence To Enhance Marine Biosecurity, Marnie L. Campbell, Chi T.U. Le, Jumana Abu-Khalaf, Craig D.H. Sherman

Research outputs 2022 to 2026

Marine biosecurity stands at the crossroads of innovation and necessity, with artificial intelligence (AI) offering promising tools to enhance risk assessment, surveillance, and response strategies for invasive species management. Despite the rapid growth of AI applications in marine science, there has been no comprehensive overview of its potential role in enhancing and supporting marine biosecurity. Our literature review aims to fill this gap, showing that AI use in this field remains limited and underdeveloped. Through our literature analysis, we identify key opportunities for AI research and innovation across the marine biosecurity continuum, while also highlighting both general and domain-specific challenges. …

Wevit: Weight-Entangled Vision Transformers With Class-Specific Attention For Weakly Supervised Semantic Segmentation, Narges Saeedizadeh, Seyed Mohammad Jafar Jalali, Burhan Khan, Shady Mohamed

Research outputs 2022 to 2026

Weakly Supervised Semantic Segmentation (WSSS) is a challenging task in computer vision, as it relies on limited supervision to generate precise object localization maps, often using Class Activation Maps (CAMs). Traditional methods struggle with balancing localization accuracy and scalability due to their reliance on fixed network architectures and handcrafted strategies. Neural Architecture Search (NAS), despite its proven success in optimizing network designs across tasks, has not yet been explored in WSSS due to the need for efficient weight sharing. To address these limitations, we propose WEViT, a novel framework that integrates NAS with transformers to optimize network architectures and generate …

Llm-As-A-Judge For Software Engineering: Literature Review, Vision, And The Road Ahead, Junda He, Jieke Shi, Terry Yue Zhuo, Christoph Treude, Jiamou Sun, Zhenchang Xing, Xiaoning Du, David Lo

Research Collection School Of Computing and Information Systems

The rapid integration of Large Language Models (LLMs) into software engineering (SE) has revolutionized tasks from code generation to program repair, producing a massive volume of software artifacts. This surge in automated creation has exposed a critical bottleneck: the lack of scalable and reliable methods to evaluate the quality of these outputs. Human evaluation, while effective, is very costly and time-consuming. Traditional automated metrics like BLEU rely on high-quality references and struggle to capture nuanced aspects of software quality, such as readability and usefulness. In response, the LLM-as-a-Judge paradigm, which employs LLMs for automated evaluation, has emerged. This approach leverages …

Efficient Test-Time Retrieval Augmented Generation, Hailong Yin, Bin Zhu, Jingjing Chen, Chong-Wah Ngo

Research Collection School Of Computing and Information Systems

Although Large Language Models (LLMs) demonstrate significant capabilities, their reliance on parametric knowledge often leads to inaccuracies. Retrieval Augmented Generation (RAG) mitigates this by incorporating external knowledge, but these methods may introduce irrelevant retrieved documents, leading to inaccurate responses. While the integration methods filter out incorrect answers from multiple responses, but lack external knowledge like RAG methods, and their high costs require balancing overhead with performance gains. To address these issues, we propose an Efficient Test-Time Retrieval-Augmented Generation Framework named ET2RAG to improve the performance of LLMs while maintaining efficiency. Specifically, ET2RAG is a training-free method, that first retrieves the …

Continuous Query For Top-K Maximal Sum Intervals Over Streaming Data, Zhongshuai Zhang, Xiaochun Yang, Baihua Zheng, Rui Zhu, Haomin Li, Bin Wang

Research Collection School Of Computing and Information Systems

The continuous identification of top-k maximal sum intervals using a sliding window over a data stream is a critical operation for applications in IoT and beyond. A maximal sum interval is a non-overlapping, contiguous subsequence with the maximal sum in a sequence of signed values. Existing algorithms are ill-suited for streaming contexts: they either exhaustively enumerate all intervals even for small k values, or depend on indexes that require frequent and costly restructuring. We propose a novel partition-based strategy. Our core insight is a partitioning scheme that guarantees that any maximal sum interval is fully contained within a single partition, …

Approximation And Learning-Based Algorithms For Influence Maximization In Multilayer Social Networks, Xueqin Chang, Ruize Liu, Qing Liu, Baihua Zheng, Yunjun Gao

Research Collection School Of Computing and Information Systems

Motivated by the observation that users in the real world often engage across multiple social networks simultaneously, we study the problem of influence maximization in multilayer social networks (Mlim), aiming to select a small set of nodes that maximizes the total influence spread across all layers. To this end, we introduce a hybrid propagation model that jointly captures layer-specific diffusion dynamics and probabilistic cross-layer propagation. Based on this model, we formally define the Mlim problem and establish its NP-hardness, monotonicity, and submodularity. To address the Mlim problem, we first propose a greedy baseline Mlim-Greedy, which achieves a (1-1/e) approximation. Since …

Left: Learnable Fusion Of Tri-View Tokens For Unsupervised Time Series Anomaly Detection, Dezheng Wang, Tong Chen, Guansong Pang, Congyan Chen, Shihua Li, Hongzhi Yin

Research Collection School Of Computing and Information Systems

As a fundamental data mining task, unsupervised time series anomaly detection (TSAD) aims to build a model for identifying abnormal timestamps without assuming the availability of annotations. A key challenge in unsupervised TSAD is that many anomalies are too subtle to exhibit detectable deviation in any single view (e.g., time domain), and instead manifest as inconsistencies across multiple views like time, frequency, and a mixture of resolutions. However, most cross-view methods rely on feature or score fusion and do not enforce analysis–synthesis consistency, meaning the frequency branch is not required to reconstruct the time signal through an inverse transform, and …

Timeradar: A Domain-Rotatable Foundation Model For Time Series Anomaly Detection, Hui He, Hezhe Qiao, Yutong Chen, Kun Yi, Guansong Pang

Research Collection School Of Computing and Information Systems

Current time series foundation models (TSFMs) primarily focus on learning prevalent and regular patterns within a predefined time or frequency domain to enable supervised downstream tasks (\eg, forecasting). Consequently, they are often ineffective for inherently unsupervised downstream tasks—such as time series anomaly detection (TSAD), which aims to identify rare, irregular patterns. This limitation arises because such abnormal patterns can closely resemble the regular patterns when presented in the same time/frequency domain. To address this issue, we introduce TimeRadar, an innovative TSFM built in a fractional time–frequency domain to support generalist TSAD across diverse unseen datasets. Our key insight is that …