48440 research outputs found
Sort by
Community Gardening at the Cubuy-Lomas Community Center
No full textIn Puerto Rico, where natural and man-made disasters are common and jobs are scarce, community support and opportunities are vital. This project worked with the communities of Cubuy and Lomas in Canóvanas through a partnership with Id Shaliah, a grassroots nonprofit. Working with the Cubuy-Lomas Community Center, we helped develop a small community garden and social space. The initiative supports residents struggling with unemployment or health issues by offering them a chance to grow crops and contribute to the center. This garden promotes resilience, self-sufficiency, and solidarity—offering a sustainable model of hope and empowerment in the face of adversity
Soft Substrates, Small Molecules: Exploring Stem Cell Differentiation
No full textSubstrate micropatterns can mimic extracellular matrix (ECM) cues to direct stem cell differentiation via mechanotransduction. Using 3D printed resin molds, the team fabricated microgrooved PA substrates of stiffness 4-40 kPa to mimic the native cardiac microenvironment and promote cardiomyocyte differentiation. This project evaluated the induction of induced pluripotent stem cells (iPSCs) from primary mouse embryonic fibroblasts (MEFs), and differentiation of the mouse myoblastic cell line C2C12 for proof of concept. We also tested cardiomyocyte differentiation from a mouse ES cell line. The iPSC induction from MEFs using small molecules was unsuccessful. C2C12 cells exhibited robust myogenic differentiation on culture plates. The mESCs differentiated to contracting cardiomyocytes on culture plates. C2C12 differentiation was successful on PA gels of 4-23 kPa, providing proof of concept. The team tested this model with mESCs differentiating into cardiomyocytes. Attachment of mouse embryoid bodies (EBs) was observed on PA gels of 10-23 kPa. However, minimal signs of differentiation into cardiomyocytes were observed. Our study validates the utility of micropatterned hydrogels for cell growth and differentiation
Investigating the Effects of Fibrotic Stiffening on Lymphatic Endothelial Cell Growth and Barrier Integrity Using Photocrosslinked Collagen Matrices
No full textFibrosis is an integral part of many chronic diseases including kidney disease, cancerous tumors, and lymphedema. Lymphangiogenesis—new lymphatic capillary growth—can be triggered by fibrosis-related tissue stiffening and soluble factor signaling that occurs under disease conditions. New lymphatic vessel density can be used as a predictor of the severity of fibrotic progression, but despite the consistency of lymphatic capillary growth during fibrosis, tumors and kidney disease are treated with anti-lymphangiogenic therapies while lymphedema therapies promote lymphangiogenesis. The differences in these therapeutic approaches likely arise, in part, from a conflation of lymphangiogenesis (growth) and lymphatic capillary function (barrier integrity). Current preclinical in vitro models are inadequate when separating the two processes and do not match outcomes observed in vivo. Moreover, less is known about how tissue stiffening generally impacts lymphatic capillary growth and function, which is important when studying fibrosis. Current in vitro models have not yet included physiologically relevant stiffnesses to represent disease states, often opting to use soft hydrogel extracellular matrices (ECM) that are more representative of tissue stiffness during developmental stages. Studies also tend to focus on lymphatic vessel growth without fully considering and assessing function (e.g., barrier integrity) as a distinct outcome. Moreover, while some information is known about how ECM stiffening regulates lymphatic vessel growth via mechanosensitive molecules and growth factor receptor expression, less is known for whether those pathways play a role in stiffness-mediated changes in function. Therefore, improved in vitro models with controlled biophysical properties are needed to systematically investigate stiffness mediated outcomes and the distinct contributions to lymphatic capillary growth and function. The work presented in this dissertation describes the use of methacrylated type I collagen (PhotoCol®, Advanced BioMatrix) as a culture substrate that can be stiffened with photo-crosslinking as an investigative tool to study lymphatic endothelial cell responses (cell morphology, cellular junctional plasticity) within a well-plate format and separately study the formation and function of capillary-like lymphatic vascular structures using a unique microfluidic device. We first showed that PhotoCol® could achieve physiologically relevant stiffness values ranging from normal to pathological tissue stiffness levels (~0.5 – 6 kPa shear storage modulus) by altering the photoinitiator solution used for photo-crosslinking (photoinitiators: Lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP), Irgacure 2959 (IRG), and Ruthenium/Sodium Persulfate (Ru/SPS)). Results showed that Ru/SPS offered the greatest dynamic stiffness range and highest overall stiffness compared to other photoinitiators, producing stiffness values that allowed us to systematically study human dermal lymphatic endothelial cell (HDLEC) responses related to capillary growth (i.e., morphology) and function (i.e., vascular endothelial (VE)-Cadherin cellular junction formation). Our quantitative morphological analysis demonstrated our ability to produce HDLECs with a fibrotic phenotype—larger with more irregular borders and increased VE-Cadherin thickness (junction zippering). Subsequently, we investigated the effects of these physiologically relevant stiffnesses on lymphangiogenic sprouting within a microfluidic device. The design of this device (established by Wang et al., 2020), unlike similar microfluidic devices, allows for full exposure of HDLECs to the ECM for studying direct sprouting and migration into the ECM, as well as cellular junctions and vessel permeability. Stiffer ECMs supported increased vascular stability and sprouting of capillary-like structures, while HDLECs within softer ECMs were more migratory with decreased cellular junction formation. This result highlights the role of ECM stiffness in directing the balance between migratory and proliferative phenotypes during vessel formation, which further dictates lymphatic capillarity integrity in normal and disease states. Finally, the underlying mechanisms by which ECM stiffness influences LEC junctional plasticity and subsequent barrier integrity was investigated by modulating LEC mechanosensing via yes-associated protein (YAP) inhibition and vascular endothelial growth factor (VEGF)-A and VEGF-C binding to VEGFR2 and VEGFR2. Overall, this work increases the field’s understanding of the unique contributions of tissue stiffness to lymphatic capillary growth and function under conditions that better reflect the fibrotic environment observed in disease states. By developing an in vitro system that is capable of producing fibrotic LEC and vascular phenotypes, we can move toward future work that investigates ways to effectively target lymphatic vasculature for therapeutic intervention
CALEB - A UNITY VISUAL NOVEL PLUGIN
No full textCALEB is a custom-built visual novel (VN) and cutscene development toolkit for Unity, designed to lower the barrier of entry for narrative-focused game development. Born from frustration with existing engines and a lack of suitable tools, CALEB evolved into a scripting and prefab system that allows creators to control dialogue, visuals, audio, and gameplay logic from simple .txt scripts. This project documents the journey from concept to execution: a story of failure, growth, and reinvention. It highlights the development process, major features, collaborative use of AI, and the feedback from real users. Evaluated in this paper are both the technical components and the design philosophy behind CALEB, as well as its potential for continued growth as a publicly available tool for developers
Flow Me if You Can: Uncertainty is My Co-Pilot
No full textNavigating nano-scale UAVs in forest-like, cluttered environments presents unique challenges. Their compact form factor imposes strict constraints on onboard sensing, power, and compute, limiting the feasibility of traditional perception and control pipelines. Moreover, passive monocular vision alone makes it difficult to react to fast-changing surroundings in time. To address these limitations, we draw inspiration from biology—specifically the agile, gaze-driven behavior of hummingbirds—to develop an active perception framework that tightly couples motion and sensing. We propose a bio-inspired navigation strategy that uses an uncertainty-aware FlowMotion network to estimate dense optic flow, inter-frame camera motion, and aleatoric flow uncertainty from monocular images. These perceptual cues are used to train a hierarchical reinforcement learning (HRL) agent that actively controls both vehicle motion and camera yaw, enabling agile, perception-driven navigation in complex environments. To bridge the gap between simulation and reality, we introduce VizFlyt, an open-source hardware-in-the-loop platform that leverages 3D Gaussian Splatting for photorealistic rendering at 100 Hz. When deployed on real nano-UAVs within forest-like settings, our trained HRL policy achieves a 96% success rate in obstacle avoidance and gap-crossing tasks—demonstrating a scalable, compute-efficient, and robust approach to embodied AI navigation in the wild
Designing The Conformable Lumen Assessment Robotic Assistant: A Soft Bodied Mobile Robot for Pipeline Inspection
No full textPipe infrastructure is an extremely important issue for modern day society. Pipes transport necessary materials such as clean water, oil and gas, and hazardous waste. Failure in these networks can lead to millions of dollars in damages and can endanger communities depending on the payload. In order to ensure the health of these structures, pipes need routine inspection to identify any potential failures. The inside of pipes especially need inspecting due to failures at pipe joints and internal corrosion and cracks due to transported material. In order to keep repair efforts to a minimum, non-destructive testing (NDT) is typically preferred. The purpose of this thesis is to update and test a compact wall-press pipe inspection robot with a novel design centered around using origami inspired continuum modules. The Yoshimura module allows this design to have an extensive level of flexibility similar inspection robots do not possess. A variable suspension system and front viewing camera allow it to traverse and inspect a multitude of different pipes. To determine the efficacy of the design, tests involving speed, battery life, verifying the kinematic model, and recording the time it takes to navigate different sized pipes and joints were preformed
Applications of Machine and Deep Learning for Continuous, Cuffless Blood Pressure Monitoring
No full textAccurate and reliable estimation of cuffless, continuous BP estimation is a powerful prognosticator of key diseases; but is based on a complex non-linear pulsatile flow-pressure relationship requiring good signal and feature extraction, an efficient machine learning algorithm and personalization to improve accuracy and lower variability. Our first aim in developing such an end-to-end BP monitoring system used a design of experiments methodology to identify optimal values for the key factors in designing a PPG sensor with a high signal to noise ratio. We then developed the theory and demonstrated in both healthy (n=21) and diseased subjects (n=31) how multimodal feature datasets can be combined with various machine learning algorithms to estimate BP using a calibration-free method in our second aim. While the BP bias was < 5 mm Hg, the standard deviation (SD) was > 8 mm Hg for SBP and MAP - not within the AAMI criteria. In our third aim, we used a catch-22 feature extractor and flow-triggered personalization on a large ‘diseased subject’ AAMI dataset (n=1525) with three machine learning algorithms. We obtained a bias< 5 mm Hg and a SD<8 mm Hg for both SBP and DBP—well within the AAMI criteria. In summary, this dissertation demonstrates that a high-fidelity PPG signal, using catch-22 algorithm for feature extraction, coupled with a “flow-triggered” personalization (key contribution) on a large dataset, and use of an optimal machine learning algorithm can improve the accuracy and reliability of continuous, cuffless BP estimation, with great promise of clinical adoption
Procedural Generation in Houdini with a Focus on Vernacular Architectural Design
No full textFor my Master of Science project in Interactive Media and Game Development (IMGD), I created a procedural generation tool in Houdini based on vernacular architecture design. My goal for this project was to research procedural architecture generation and architectural design relationships, as well as how to implement architecture generation in Houdini. The tool can create variations of two styles of buildings that are placed procedurally on a desert and tundra inspired terrain. For the Adobe style generator, I added variation through the creation of the floor plans. With the Norwegian style generator, I focused on creating every model in Houdini and facing each house towards roads on the terrain. Lastly, I was able to combine both generators into the same terrain. If there were more time, I would use what I've created to make a game or educational tool
Mindful Machines: Enabling HCI Research Exploring Novel Brain-Driven Interaction Paradigms for Collaboration
No full textWith advances in the capabilities and affordability of brain sensing technologies, brain data is increasingly being integrated as input into interactive systems. Such data can give insight into the cognitive and affective states of users, augmenting their capabilities and enriching interactions, as well as informing user-centered design and evaluation of innovative interfaces. However, there is a dearth of user-friendly tools supporting the development of brain-computer interfaces, which typically requires a high level of time and expertise. The aims of this dissertation are to develop and evaluate such tools and methods to make working with brain data accessible, and to demonstrate the utility of brain signals in novel interaction paradigms for collaboration
Investigating the Robustness of Knowledge Tracing Models in the Presence of Student Conceptual Drift
No full textKnowledge Tracing (KT) has been an established problem in the educational data mining field for decades, and it is commonly assumed that the underlying learning process being modeled remains static. Given the ever-changing landscape of online learning platforms (OLPs), we investigate how concept drift and changing student populations can impact student behavior within an OLP through testing model performance both within a single academic year and across multiple academic years. Four well-studied KT models were applied to five academic years of data to assess how susceptible KT models are to concept drift. Through our analysis, we find that all four families of KT models can exhibit degraded performance, Bayesian Knowledge Tracing (BKT) remains the most stable KT model when applied to newer data, while more complex, attention based models lose predictive power significantly faster. To foster more longitudinal evaluations of KT models, the data used to conduct our analysis is available at https://osf.io/hvfn9/?view_only=b936c63dfdae4b0b987a2f0d4038f72