137091 research outputs found
Sort by
NASA High Speed Commercial Vehicle (HSCV) Propulsion System and NPSS Model Development
No full textPaper presented in a technical paper session at the 2025 AIAA AVIATION conference in Las Vegas, NV, discussing the development of a propulsion model for the NASA HSCV project, which develops concepts for supersonic commercial air travel.NASA’s High Speed Commercial Vehicle (HSCV) project aims to develop concepts and establish technology needs to enable supersonic commercial air travel for 50 passengers at cruise speeds in the range of Mach 2 – 4. This paper summarizes the propulsion content of the HSCV program. Specifically described within is the derivation of propulsion system requirements, the architecture concept, the development of a corresponding NPSS model, and trade study conclusions derived therefrom.NAS
Literature-based discovery to differentiate spectral neuropathology
No full textArtificial intelligence-enabled literature-based discovery (LBD) provides an advanced method for performing large scale, cross-domain analyses of disease. Unlike traditional systematic reviews, which use a narrow and siloed set of data sources, LBD integrates millions of text relationships from cross-domain data sources for a more comrehensive and objective assessment. Specifically, SemNet 2.0 is a LBD software that utilizes a knowledge graph of text relationships extracted from 33+ million PubMed studies and an unsupervised learning rank aggregation algorithm to rank the importance of user-defined targets (i.e. target nodes) to related biomedical concepts (i.e. source nodes). SemNet 2.0 had previously successfully been utilized for drug repurposing, mechanism identification, and adverse event identification. In the present study, artificial intelligence-enabled LBD with SemNet 2.0 and the companion visualizer, CompositeView, was innovatively used to differentiate spectral neuropathology. Spectral neuropathology includes multi-factorial diseases where symptoms and/or underlying pathological mechanism or dynamics have overlapping similarities. SemNet 2.0 was used to specifically compare and contrast the spectral neuropathology of Alzheimer's Disease (AD), Frontotemporal Dementia (FTD), and Amyotrophic Lateral Sclerosis (ALS). Comparisons were performed to identify overlapping and differentiating Unified Medical Langauge System (UMLS) node types of amino acids, peptides and proteins (AAPP); diseases or syndromes (DSYN); and pharmacological substances (PHSU). Using percentile-ranked normalized HeteSim relevance scores and composite scores, the top 1\% of relevant nodes are reported for AD, ALS, and FTD, as well as the intersections and unions between these spectral diseases. Top-ranking nodes were mapped to functional biological processes to more holistically assess differences in underlying spectral etiology. Finally, human-in-the-loop validation was used to apply relevant context to the top-ranked nodes and functional biological processes. In conclusion, LBD results illustrate that AD, ALS and FTD share a large degree of underlying pathology dynamics, and thus, likely comprise a multi-factorial neuropathological spectrum. Small differences in the network, through either the identified genetics, co-morbidities, or environmental differences, likely shape the underlying expressed disease phenotype. Finally, the results of the present LBD study provide prioritized, testable hypotheses for future clinical or experimental research to better understand, diagnose, and treat the overlapping, spectral neuropathology of AD, FTD, and ALS.M.S.Biomedical Engineerin
Low-Force Hand Interactions Induce Changes to Human Gait through Sensorimotor Engagement with Task-Relevant Information Instead of Direct Mechanical Effects
No full textThe motivation for this research is to develop intuitive low-force human-robot hand interactions to alter human walking. These types of interactions have the potential to improve human movement, enhance physical collaborations between human-human and human-robot partners, and facilitate the performance of physical tasks not previously possible. While current physical human-robot interactions aid walking primarily through direct mechanical effects – i.e. by applying large forces directly on human tissues/joints to propel locomotion or by supporting significant bodyweight, I take a different approach to improving walking through low-force interactions at the hands. This approach is inspired by subtle hand interactions between human partners that alter gait without explicit instructions or training. While concepts from human-human interactions have the potential to inspire human-robot interactions, the control strategies for modifying gait in human partners are not well understood, and there are no existing hand-contact robotic devices adequate for testing controllers based on human-human hand interactions during walking.
Through a series of human-human and human-robot studies, I demonstrate my central hypothesis that low-force hand interactions can induce people to change their own gait through sensorimotor engagement with task-relevant haptic information instead of relying on direct mechanical effects. I demonstrate the feasibility for hand interactions to induce intended changes to gait by showing a) improvements to balance and desired changes to step frequency in human-human interactions and b) desired changes to gait coordination in human-robot interactions. I demonstrate that these gait changes do not rely solely on mechanical effects by showing that a) hand forces remain below 30N, b) mechanical power transfer at the hands is not sufficient for directly propelling walking, and c) gait changes occur only when humans expect task-relevant information from hand interactions.
This research establishes a scientific framework for examining human sensorimotor control of hand interactions during walking. I develop novel experimental paradigms, analysis methods, computational models, and a physical device – a robotic emulator – to enable investigations of human control strategies. My results provide principles of haptic communication and sensorimotor engagement for physical collaborations between human-human and human-robot partners as well as greater understanding of how hand interactions influence walking within an individual.
This work also has broad engineering applications for improving human walking and performance of hand interactions during walking. I develop a hand pHRI controller specifically to alter gait coordination and provide general guiding principles for designing effective and intuitive low-force hand interactions during walking. Such controllers have many potential applications, such as physical assistance and rehabilitation (e.g. robotic walkers), industrial manufacturing (e.g. human-robot load-transportation), physical education (e.g. teaching dance), and recreation.Ph.D.Biomedical Engineerin
Efficient modeling of 3D shape of women during gestation to assess risk of cephalopelvic disproportion (CPD)
No full textCephalopelvic disproportion (CPD) is a mismatch in the size of the maternal pelvis and the fetus, which often leads to obstructed labor. Most cases of CPD require C-section for successful delivery and in low resource settings like Ethiopia, there is a lack of adequate facilities with the infrastructure or the expertise to perform a C-section. Currently, obstructed labor is known to account for 11 – 22 \% of maternal deaths in Ethiopia. Early assessment of the risk of CPD would enable women in these settings to access the proper healthcare services and improve overall maternal health. This thesis aims to develop an algorithm that would use longitudinal shape modeling to analyze, in real-time, 3D scans of pregnant women and assess the risk of CPD-related obstructed labor at the earliest possible stages of gestation. The longitudinal shape model would be trained on 3D scans of pregnant women across different periods of gestation and would be optimized to run on devices with low computational power. The prognostic value of the model for assessing the risk of CPD would be compared to anthropometric measurements. This model is envisioned to be used by nurses and midwife personnel as part of point-of-care tools for routine antenatal care in low-resource settings.Ph.D.Bioengineerin
A Machine Learning Model for Cellular Microscopy Segmentation
No full textDespite the number of new options biologists have to automate the task of
cell segmentation, virtually none have caught on and the majority of biomed-
ical research labs rely on human annotators to collect and analyze microscopy
data. The goal of this paper is to facilitate the process of cell segmentation
for researchers by providing a simple solution. We aim to build upon the ad-
vancements of SAMCell and increase its usability for common use in biomedical
research laboratories through the improvement of the model and development
of an accessible lightweight User Interface for users to make use of SAMCell in
a manner that minimally disrupts lab procedure and resources.
A modern challenge of cell culture analysis is a lack of standard metrics.
Scientists with experience in cell research form intuitions regarding cell culture
analysis, but their ability to confirm results are limited. If this project is suc-
cessful in attracting users in biomedical research, a secondary goal is to leverage
SAMCell, and this new ability to extract concrete cell masks, to standardize
the methods by which certain cell culture metrics are obtained.UndergraduateComputer Scienc
Parametric Carrier Sizing for Pressurized Rover Resupply Operations
No full textAIAA ASCEND 2025, Las Vegas, NVThe Artemis program aims to establish a lunar base near the south pole of the Moon, with plans that potentially include a surface habitat and other surface elements, such as a pressurized rover. To ensure continuous operation, these elements require regular cargo resupply that includes consumables necessary for mission success that are not installed as part of the vehicle. Resupply cargo will be delivered via specialized logistic carrier modules, whose design can benefit greatly from understanding the impact of high-level design alternatives. This study introduces a parametric framework for sizing logistic carrier modules, with emphasis on the impact of these high-level design choices. The framework outlines structural sizing requirements to accommodate resupply consumables, along with the thermal and power subsystems needed to maintain temperature stability and supply power during independent carrier operation. Trade-offs between larger, integrated carriers, capable of resupplying both the habitat and a
pressurized rover, and smaller specialized carriers for pressurized rover resupply are evaluated. The framework’s capabilities are then demonstrated through sensitivity analysis and design space exploration for a use case mission
Modulation of the Electronic Structure and Redox Properties of Lanthanide Imidophosphorane Complexes
No full textLong characterized by their similar chemical properties and ubiquitous 3+ oxidation state, the lanthanides have recently been shown to exhibit a broader range of molecular oxidation states than previously thought possible. In 2019, terbium was reported in the 4+ oxidation state in independent reports utilizing imidophosphorane or siloxide ligand frameworks. Despite the only slightly more positive redox potential, Pr4+ proved more difficult to isolate in either of the ligands that initially supported Tb4+. In this thesis, the impact of the alkali metal cation on the redox properties of Ce3+ complexes is investigated. Additionally, a rare structurally authenticated example of Pr4+ in a low-coordinate tetrahedral environment is reported and analyzed by various physical characterization methods, including synchrotron X-ray absorption near-edge spectroscopy for the first time in a molecular complex. Further ligand development for f-element complexes is explored using La3+, due to its closed-shell electronic structure, which facilitates characterization by methods such as nuclear magnetic resonance spectroscopy. In sum, significant advances in tetravalent lanthanide chemistry are presented, and the importance of new ligand design and tuning is exemplified.Ph.D.Chemistry and Biochemistr
Deep Learning for Object Classification in Order Picking
No full textWe propose a method for image classification in warehouse order picking using ResNet models to learn representations of the objects used in the order picking task. We apply iterative clustering techniques using hue-saturation histograms to assign ordered labels to each pick within a picklist to generate training labels for each frame, and we train the ResNet models using these labels. Thus, our data pipeline allows models to be trained from unordered picklist labels without any prior knowledge about the objects other than the number of unique object classes.
We observe a per-frame accuracy of 92.4% using the ResNet-10 model, a significant improvement over the baseline of 56.6% using only hue-saturation histograms. Additionally, we observe a 100\% accuracy on a per-pick level, suggesting that the model is unlikely to make many incorrect predictions within a single pick sequence, and thus would be able to perform well in a realistic warehouse scenario where it only needs to make a single prediction per pick.UndergraduateComputer Scienc
Improving Scalability of Deep Learning Accelerators Using 3D IC
No full textOur work investigates and addresses the critical scalability challenges that flexible Deep-Learning (DL) accelerators face in both interconnect and memory system design. As DL models grow increasingly diverse in their shapes, operators, and data reuse patterns, rigid architectures such as systolic arrays struggle to maintain high utilization, leading to inefficient execution on emerging workloads. Flexible accelerators, such as SARA and MAERI, overcome these utilization bottlenecks through reconfigurable compute arrays and adaptable dataflows. However, this flexibility introduces substantial physical design overhead, particularly in wirelength, area, power, and critical path timing.
To address these bottlenecks, we propose architectural and physical design methodologies that leverage 3D integration to improve the scalability of flexible DL accelerators. First, we tackle the scalability limitations of flexible interconnects by identifying key architectural bottlenecks—namely, topology complexity and switch logic depth and introducing Logic-on-Logic 3D partitioning techniques. Our methods yield up to 3x improvement in timing, over 75% throughput gains, and reduced energy costs compared to 2D flexible designs.
Second, we analyze memory system and datapath scalability issues specific to hierarchical accelerators like SARA. These architectures suffer from excessive memory port proliferation and datapath congestion as PE count increases. To address this, we apply 3D Memory-on-Logic integration and improved macro placement strategies. Our design demonstrates up to 1.24x runtime speedup, 1.4x improvement in EDP, and 1.3x reduction in area compared to traditional 2D design baselines.
These contributions validate our work's central argument: 3D integration is a powerful enabler of scalable, high-performance, and energy-efficient flexible DL accelerators. By co-optimizing architecture and physical design and integrating realistic simulation data and PPA metrics, this work presents a practical and future-facing blueprint for accelerator scaling in the post-systolic era.M.S.Electrical and Computer Engineerin
Evalutation of the Life Cycle Costs and Performance of Geotechnical and Stormwater Infrastructure
No full textAs urbanization accelerates and global population increases, the necessity and importance of highway infrastructure has rapidly increased. Infrastructure serves to improve the quality of life, supports community needs, and increases economic benefits through increased access to resources. However, indiscriminate development can lead to severe environmental and ecosystem issues such as deforestation, greenhouse gas emission, and wildlife destruction. Even though environmental concerns have been rising globally, decision-makers have historically focused primarily on the economic aspects of infrastructure design, construction, and maintenance. A balance of economic, social, and environmental impacts is an important factor for sustainable development.
The work performed in this dissertation focused on the sustainability of transportation roadside infrastructure, particularly geotechnical assets including mechanically stabilized earth (MSE) retaining walls and stormwater best management practices (BMPs). Life cycle assessment (LCA) and life cycle cost analysis (LCCA) were conducted in the planning and design to construction phases to identify sustainable design options for assets to support decision-making. Additionally, the performance of assets and the environmental and cost impacts of asset management was evaluated during the operation and maintenance (O&M) phases. The work provides insights to minimize significant environmental, economic, and social impacts with low expected expenses and extend the life expectancy of assets.
The work showed that the impacts of planning and design and construction phases were significant when compared to the O&M phases in the life of MSE walls. In particular, the selection of backfill materials and machinery operation were identified as the most important factors influencing sustainability. In addition, in stormwater BMPs, the treatment and removal of contaminants from stormwater runoff during O&M phases can effectively offset the construction impacts of stormwater infrastructure. To optimize performance, well-planned construction and long-term management are essential.Ph.D.Civil Engineerin