137091 research outputs found
Sort by
Exploring the Limits of Behavior, Muscle Activity, and Wing Kinematics of Linear Flower Tracking Regime
No full textAnimals have a linear tendency when tracking objects. The same principle applies to the Manduca sexta, which has an overall linear behavior when tracking the unpredictable flower that they feed from. There are many subsystems that are involved including sensorimotor control, specifically visual and mechanical reception, muscle physiology, and wing kinematics. By looking at these individual system, the linearity of these subsystems can be analyzed to determine how they contribute the linearity of the hawkmoth’s behavior and explore the limits of this control using free flight tracking of wing kinematics and EMG and wing kinematic activity. First, by looking at the aerodynamics and sensorimotor systems, there exists a linear relationship within the aerodynamics across three incrementing sum of sine stimuli. There is a nonlinear tendency in the magnitude of the gain at higher frequencies of the largest stimulus amplitude for the sensorimotor system. This means that there must exists a nonlinear relationship elsewhere to cancel out and create an overall linear behavior. Looking at the relationship between wing kinematics and muscle activity is a preliminary start. Across the same stimulus, there exists no correlation between the both the right 3rd axillary and right DVM and the deviation and flapping angle. There is a relationship when looking at the bilateral latency for these muscles. The bilateral latency of the 3rd axillary muscle shows negative correlation for the timing and both the deviation and flapping angle, while the DVM bilateral latency shows a positive correlation for the timing and both the deviation and flapping angle. These results show us that there is variation occurring in the difference between the right and left muscles that lead to a variation in the wing activity.UndergraduateNeuroscienc
An Exploration of Titanium Oxide Hydrate/Polyalcohol Hybrids for Solar Energy Harvesting and Storage
No full textPhoto-rechargeable redox flow batteries are an emerging energy storage system that advantageously combines solar energy harvesting and storage capabilities into one device. This thesis explores inorganic/organic hybrids based on a liquid titanium oxide hydrate and various polyalcohols as novel photo-rechargeable redox flow battery electrolytes. Nuclear magnetic resonance spectroscopy suggested that the polyalcohol participates in the photoreduction mechanism of titanium oxide hydrate. Gas chromatography mass spectrometry on titanium oxide hydrate/glycerol hybrids enabled the development of a possible photoreduction mechanism for titanium oxide hydrate. Importantly, it is proposed that glycerol is consumed within this photoreduction reaction. Electrochemical and optical properties of titanium oxide hydrate/polyalcohol hybrids were also characterized. The voltage generated by these hybrids under UV illumination and their ability to store voltage without constant UV illumination were both dependent on the polyalcohol identity within the hybrid. Ultraviolet-visible spectroscopy and cyclic voltammetry suggested reasons for these electrochemical properties through potential differences in the stability of titanium(III). Overall, the knowledge presented in this thesis can promote a new class of photo-rechargeable redox flow battery electrolytes for more accessible solar energy generation and storage.UndergraduateChemical and Biomolecular Engineerin
The Physical Internet: Research Trend Analysis Through Natural Language Processing
No full textThe Physical Internet (PI) is a revolutionary concept in the logistics sector that has emerged over the past decade. It aims to develop an integrated network that emphasizes sustainability and efficiency to achieve interoperability and a collaborative global logistics system. To achieve this, numerous studies have been conducted addressing various aspects of the PI. However, understanding how PI research is currently developing and which directions it focuses on remains limited. Thus, a comprehensive analysis of past PI research is needed. This study aims to analyze the research trend of the PI by identifying the current state of research and under-researched aspects using natural language processing techniques. Relevant PI-related papers were obtained throughout the years from numerous journals and served as the dataset for this research. The data were reconstructed and filtered through natural language processing (NLP) to retain only relevant information. It is further analyzed using BERTopic, a machine learning technique based on Bidirectional Encoder Representations from Transformers (BERT). The analysis produced several topics that represent focus points in PI research. These topics highlight research trends and indicate areas that can be evaluated, offering better insights for future research related to the Physical Internet
Advanced characterization of transport and microstructural properties in membrane materials
No full textMembranes are employed in a variety of sustainable water treatment and energy conversion technologies. Advancing membrane materials for new separations includes unraveling the complex structure-property relationships that govern water and salt transport in conjunction with scaling processing methods. Material properties can be tuned by increasing salinity, adjusting humidity, or introducing filler particles to achieve desired transport metrics. These modifications impact water and salt diffusion as well as microstructural characteristics in the membrane materials, yet the detailed relationships between membrane performance, processing conditions, and transport properties remain to be fully characterized. Although extensive membrane research has been conducted, a need exists to better understand how compositional changes impact transport in membrane materials as well as resulting membrane properties. Understanding the structure- property-processing relationships of membrane materials is crucial for optimizing performance in practical applications.
Here, I examine three classes of membrane materials used in water treatment and energy conversion applications: polyamide reverse osmosis membranes, Nafion cation (proton) exchange ionomers, and graphene oxide-based nanofiltration membranes. Advanced characterization techniques quantify water diffusion coefficients and relaxation times in polymer membranes under varying environmental conditions. Additionally, aggregation behaviors and interlayer binding interactions between graphene oxide sheets and intercalant molecules are examined. Diffusion measurements reveal distinct transport mechanisms of water and salt in charged and uncharged polymer membrane materials. Aggregation behavior observed in graphene oxide (GO) dispersions and salt rejection measurements with GO membranes bridges the gap between fundamental material properties and the membrane performance.Ph.D.Mechanical Engineerin
Prefetching in CXL-based Server Processors
No full textThe memory subsystems of modern server processors have undergone significant architectural changes in recent years. Furthermore, emerging technologies such as CXL are poised to become widely adopted in the industry. This work studies how data prefetching in server processors is impacted by these trends and in particular focuses on latency-critical datacenter applications. We show that state-of-the-art data prefetchers are not effective in bandwidth-contained configurations that server processors typically operate. However, we show that data prefetching becomes more important in a CXL-based processor due to the increased memory bandwidth and increased memory access latency.UndergraduateComputer Scienc
Toward light-field interrogation of cell biology: Physics, computation, and systems
No full textUnderstanding how intracellular molecules and organelles are organized and dynamically mapped to functional activities requires tools for volumetric interrogation of single-cell systems with high spatiotemporal resolution and high throughput. Conventionally, optical microscopy techniques, such as wide-field microscopy, confocal microscopy, and structural illumination microscopy, produce orthographic views and acquire 3D information in a sequential or scanning fashion. In recent decades, the method of point-spread-function engineering has been rapidly developed, which allows for various unique features, such as non-diffracting beams for extended depth of focus, and axially variant beams for 3D imaging. On the other hand, light-field microscopy, known as one of the point-spread-function engineering methods, simultaneously records both the 2D spatial and 2D angular information of the light field with a microlens array, allowing for the computational synthesis of the volume of a specimen from a single camera frame. This 4D imaging scheme offers ultrafast and volumetric acquisition, minimum photodamage for time-lapse observation, and high scalability and design flexibility. These abilities have revolutionized functional imaging with a cellular-level, milliseconds spatiotemporal resolution across a significantly large depth of focus. Due to the increasing need for new microscopic techniques that can provide high-throughput, high-spatiotemporal-resolution multi-color volumetric imaging of subcellular anatomies and dynamics, this thesis examines the physics, computation methods, and systems of light-field techniques, together with a comprehensive solution for microscopic wavefront modulation and the innovation of integrating flow cytometry and deep learning advances.
The main theme of this thesis revolves around three aspects. First, we established a point-spread-function-engineering strategy for depth-extended high-resolution volumetric imaging. Specifically, we develop the entire microscopic system, which includes the optical setup construction, the wave-optics model for numerical simulations, and the imaging acquisition and processing programs. We built the Fourier light-field microscope by implementing a microlens array to allow for single-shot 3D information retrieval. Second, we applied optofluidic devices to the imaging system for high-throughput, high-resolution volumetric imaging. We constructed the light-field flow cytometer for single-cell screening by integrating microfluidics into the Fourier light-field microscopy. We implemented stroboscopic illumination for motion-blur suppression and imaging throughput improvement. Finally, we enhanced the imaging capability of the Fourier light-field system with deep learning. We effectively reduced reconstruction artifacts and accelerated the reconstruction process by incorporating deep neural networks for the task of 3D image reconstruction. The advancement offered by the Fourier light-field system presents a promising methodological pathway for broad cell-biological and translational investigations, with the potential for widespread adoption in various biomedical research fields.Ph.D.Biomedical Engineerin
Next Generation Innovators: A Solution to Workforce Development and Research
No full textPresented on March 13, 2025 in the TSRB 1st floor ballroom at 12:30 p.m.Debra Lam is the Founding Director of the Partnership for Inclusive Innovation, a statewide public-private partnership committed to investing in innovative solutions for shared economic prosperity. Debra serves on the board of the Community Foundation of Greater Atlanta and was most recently appointed by the U.S Department of Commerce to the Internet of Things Advisory Board.Clarence Anthony Jr. is the inaugural Workforce Development Program Manager for the Partnership for Inclusive Innovation. His goal is to develop initiatives that engage current and emerging talent in capitalizing on their strengths and developing new skills to build a strong workforce within the Southeast region.Runtime: 45:21 minutesTraditionally, research labs rely on, and benefits from, the work of undegraduate and graduate research assistants and visiting reseachers/scholars, with many being on track for the professoriate. However, higher education can benefit from engaging in increased opportunities that develop talent pools that are fluent in, and capable in building, public-private partnerships. This session will showcase how the Partnership for Inclusive Innovation’s Fellowship and Internship programs has leveraged early career talent to bolster research and economic development activities
Heterogeneous Integration for High-Speed mm-Wave Communication
No full textThe objective of the proposed research is to develop highly integrated interposers with embedded dies and thin film components such as antennas at sub-THz frequencies for the next generation 6G applications. The focus is on handsets that feature back-end high-speed processing for applications such as on-device AI. AI applications require specialized chips that need to communicate with each other through interconnections. Our focus here is to demonstrate high-speed communication between chips through non-TSV based integration in glass interposer. 3D\textsuperscript{+} integration is proposed for the integration of RF and baseband ICs in glass interposer with antennas to achieve a compact profile and low insertion loss between the chips. The proposed packaging architecture features both embedded chips and flipchips on glass interposer with short, vertical, high-bandwidth interconnects between the chips. Chip-to-interposer and chip-to-chip transitions in the proposed integration technology is modeled and characterized using specially designed test structures. The challenges with direct characterization of the individual transition are addressed by introducing a characterization method that uses multiple back-to-back transitions to extract the S-parameters of an individual transition. A broadband S-parameter characterization from near DC to 170 GHz of such interconnects is presented, using which eye diagram simulation have been performed to predict the maximum data rate and the bandwidth density supported. This work also discusses the design and characterization of integrated high-bandwidth end-fire antennas on glass interposer in D-band. Characterization of antennas operating at frequencies over 100 GHz entails challenges due to the unavailability of the coaxial cables at these frequencies. This work aims to address this problem by exploring probe-station-based methods of end-fire antenna characterization without using expensive characterization facilities that require robotic arms and custom assemblies.Ph.D.Electrical and Computer Engineerin
Gromov Wasserstein Framework and Survey
No full textIn biological research, analyzing experimental data for cell type classification is a challenging and time-consuming task. Gromov-Wasserstein optimal transport methods offer promising results, but they face scalability issues with large datasets and variable features. To address this, we propose a Python-based framework facilitating the comparison of Gromov-Wasserstein implementations. We review literature, outline methodology, and assess scalability and accuracy metrics using simulated data. Results highlight Sliced GW's efficiency and Sampled Sliced GW's promise compared to other variations. Our work underscores the need for scalable and accurate computational methods in biological research, with implications for machine learning and bioinformatics.UndergraduateComputer Scienc