48440 research outputs found
Sort by
The Kevin Love Fund: Auditing and Expanding to Ensure Diversity and Representation
No full textThe Kevin Love Fund (KLF) is a non-profit organization founded by Kevin Love to work towards destigmatizing mental health. The goal of this project was to audit KLF's video library; identify gaps in the current stories; and compile recommendations to augment the cultural responsiveness of the library. This project also contributed additional videos from the Worcester Polytechnic Institute (WPI) student population to fill identified gaps. This project was carried out in a three step methodology; the students coded and audited KLF's curriculum video library, constructed a recommendations list of additional stories, and held KLF curricula-based workshops for the WPI student population. Through the audit, this student group was able to develop multiple interactive indices of present stories containing both experiences and identities. From these indices, 15 key experiences and identities were submitted to KLF's Education Team for prioritization in future video submission soliciting
Scaling Climate Action in Major German Cities Through Change Clubs
No full textThe climate crisis is one of the most pressing issues of the modern day, and nations around the world have taken varying levels of action to slow and alleviate its catastrophic effects. In recent years, the German government has been a leader in the realm of climate action and has taken steps to reach complete carbon neutrality. Although Germany’s legislation is effective in some areas, it falls short of making the shift needed to restore the environment. Community climate action and activism is necessary to pick up the slack that lawmakers leave behind. To create unity and multiply the effects of individual action, the non-profit organization Change Clubs uses small group gatherings to enhance the effects of community-based, citizen-centered climate action through the formation of “Change Clubs.” Recently, Change Clubs has initiated efforts to expand their reach into more urban areas, such as our host city of Berlin, Germany. To amplify the organization’s presence and impact in Berlin, our project team conducted archival and secondary research, interviews with key individuals in climate activism, participant observation in climate-focused events, and action research by engaging with the target audience of new members. We developed three project deliverables to give the organization insight into how to best establish themselves in Berlin: a database of potential partner organizations, recommendations for effectively establishing connections with the target audience and potential partners, and a bank of innovative growth ideas for expanding the organization’s public reach. With these deliverables, Change Clubs has a customized, research-based path for continued expansion of their mission in Berlin
Digital Tool Garage for civil societies in Taiwan and East Asia
No full textRecognizing the growing need for strong digital security, we bridge the gap between awareness of threats and the ability to protect against them. Through analysis of the needs and challenges of civil society organizations (CSOs), especially those supporting human rights defenders, we developed a software testing methodology, evaluated 21 digital tools, and published our findings in a publicly accessible GitHub repository. By offering CSOs accessible, thoroughly vetted tools, our repository acts not just as a resource, but as a lifeline—advancing the Open Culture Foundation’s mission to use open technology and collaboration to strengthen digital civil society against today’s threats. In the face of escalating digital threats, our toolkit empowers East Asian civil societies and human rights defenders to protect their digital rights and continue their vital work safely
Magnetic Imaging of Current Flow in MXenes using Quantum Sensors
No full textMXenes are a class of low-cost two-dimensional (2D) transition metal carbide, carbonitride, and nitride materials with exceptional electronic, optical, chemical, and mechanical properties. They have high electrical conductivity, offer exceptional strength and stiffness, are biocompatible, and have scalable synthesis methods. As such, they have enormous promise for revolutionizing energy storage, smart textiles, flexible electronics, medicine, 5G/6G communications, ultra-fast sensors, environmental remediation, and more. The electrical conductivity of MXenes is influenced by their chemical composition, the configuration of their surface terminations, and the physical MXene geometry. Additives, coatings, or defects in MXenes can be used to tune their electronic properties. However, with this, it becomes critical to understand the spatial current flow dynamics in different MXenes. In this work, we applied a combined experimental and computational approach to spatially resolve current flow in MXenes. We used quantum sensors to image magnetic fields generated from electric current flow through Ti3C2Tx MXenes. Specifically, we used nitrogen-vacancy (NV) centers in diamond to collect vector magnetic field data with micron-scale spatial resolution. Using this magnetic field data, we implemented a procedure based on the spectral inversion of Biot-Savart law to reconstruct the current density from the magnetic field. Applying this approach, we studied current flow through pure MXene samples with different geometries and physical defects, a MXene sample composed with sodium tripolyphosphate (TPP) for stability, and a MXene sample composed with silk fibroin (SF) for biocompatibility. The data showed that current flowed around physical defects, that the addition of TPP did not fully compromise current flow, and that textured current density was present in the MXene and SF composite material. These results experimentally demonstrate a new application for NV diamond magnetometry to study current flow in MXenes. Additionally, this new robust technique for high-resolution current imaging in MXenes can be used to aid in the future engineering of spatial current-flow dynamics or detection of micron-scale defects in MXenes
Use of a Virtual Reality System to Characterize the Adhesive Properties of Coatings of Ligand-Conjugated PEG-Coated Nanoparticles
No full textA series of experiments were conducted in order to determine whether varying the density of a ligand-conjugated PEG-coated nanoparticle designed to specifically target triple negative breast cancer cells would affect its adhesion effectiveness to these types of cells. To this end, a Virtual Reality (VR) headset was used to visualize and scrutinize the process. To achieve the desired outcome, the Narupa VR Molecular Simulation Software developed by Intangible Realities Laboratory was utilized. The programs that support the VR simulation had to be optimized to correctly display the molecular interactions of the ligand-conjugated PEG-coated nanoparticles. After conducting multiple simulations, it was found that increasing the density of the coating yielded a measurable increase in its adhesion effectiveness to breast cancer cells. Incorporation of the VR system considerably facilitated the understanding of the ligand coating molecular interactions, thereby simplifying the complexity of the process. In particular, the VR system proved its value beyond manufacturing applications in the Biomedical field. A number of improvements to the VR system are also suggested in an effort to further improve the value of the VR system for the specific targeting of breast cancer in Biomedical applications
Enhancing Reuse Assessment at the Amager Ressource Center
No full textWith the ever-growing popularity of circular economies in Denmark, direct reuse is being integrated more into daily life, lengthening the lifespan of items that would otherwise be thrown away. Organizations such as Amager Ressource Center (ARC) facilitate direct reuse but struggle to accurately report the amount of reuse occurring as user engagement grows. We worked with ARC through their administrative office and eight of their recycling centers to determine the efficacy of their current means of reuse assessment and ascertain areas of improvement through participant-observation, employee interviews, and customer surveys. Our recommendations include an artificial intelligence object detection and weight estimation model, customer self-checkout kiosks, and standardized employee reporting
Herd Highlights, Issue 29, May 1, 2025
No full textBi-weekly email newsletter for undergraduate and graduate students created by WP
Advancing Image-Based Assistive Technologies In Robotic Surgery: Perception, Targeting, and Planning
No full textRobotic surgery has evolved rapidly and advanced modern medicine over the decades. It is often minimally invasive and is usually guided by medical imaging and instrument navigation systems. Studies have shown that robotic-assisted surgery can improve procedural precision and outcomes, minimize patient trauma for faster recovery, and reduce surgeon fatigue. Despite numerous accomplished innovations, many robotic procedures still face challenges in fusing real-time imaging with intelligent motion control. This dissertation responds to these challenges across 3 key domains: image-based perception, needle targeting, and treatment planning. Firstly, a markerless 6 DOF needle pose tracking technique was developed by integrating deep learning-based keypoint detection with point-wise registration, enabling robotic suturing. Building on this, the dissertation extends to Magnetic Resonance Imaging (MRI)-based needle localization and targeting for Laser Interstitial Thermal Therapy (LITT), employing an MR-compatible robot that takes intraoperative feedback to improve precision. Finally, once the thermal applicator is positioned, a Finite Element Method (FEM)-based interactive treatment planning toolkit and a Reinforcement Learning agent can assist surgeons for conformal planning of directional hyperthermia. Collectively, the advancements highlight the integration of perception, targeting, and treatment planning in robotic surgery, emphasizing the promise of image-based assistive technologies in translating image information into robot actions. This dissertation contributes to more effective, patient-specific interventions and shows the potential of intelligent medical robotics
A Multi-Scale Framework for Climate-Adaptive Passive Thermal Enclosures Integrating Bioinspiration, Optimization, and Machine Learning
No full textThe growing demand for heating and cooling in the built environment underscores the need for advanced thermal enclosures capable of autonomously adapting to dynamic climatic conditions for energy saving while maintaining thermal comfort. Conventional static materials fail to maintain optimal thermal performance throughout the year, while active systems rely on external energy input, undermining their sustainability. To address these challenges, this Ph.D. research develops a comprehensive multi-scale framework that integrates bioinspired surface engineering, multi-objective numerical optimization, and machine learning–based predictive modeling for the optimal design of passive adaptive thermal enclosures. At the material scale, micro- to sub-micrometer bioinspired surface morphologies inspired by species such as the Saharan silver ant and the Morpho butterfly were designed to achieve tunable optical properties for thermal regulation. Using coupled electromagnetic–thermal numerical modeling (FDTD for spectral response and finite-element heat-transfer simulations for temperature regulation), we demonstrated that these morphologies enhance radiative adaptability by dynamically shifting solar absorptivity and emissivity in response to environmental conditions. The targeted thermal performance included maximizing solar heat gain in cold temperatures (to support passive heating), enhancing mid-infrared radiative cooling under hot and clear conditions, and achieving balanced heat exchange in temperate climates. The resulting designs demonstrated the ability to dynamically shift optical properties, enabling the enclosure to reduce heating energy demand and cooling loads while maintaining year-round thermal comfort in different climates. At the system scale, multilayer thermal enclosures combining radiative, conductive, and storage adaptive layers were developed and optimized for diverse climatic regions (Fairbanks, Phoenix, and Baltimore) and applications (building envelopes and cold storage). Multi-objective optimization was performed to minimize annual energy consumption and overall layer thickness while ensuring material feasibility. The results revealed strong climate–function dependencies, where sorbent storage layer enabled superior energy efficiency in cold regions, and adaptive conductivity layers played a key role in compact, high-performance systems. Trade-offs between energy reduction and thickness were identified, offering valuable design guidelines for practical implementation. Finally, a machine learning model was trained on a database of approximately 9,000 optimized cases from the previous computational work to predict annual energy consumption and required layer thickness under new boundary conditions. The trained model achieved high predictive accuracy and reduced computation time from several days to seconds, transforming the optimization process into a real-time predictive and prescriptive design tool. Collectively, this research aims to bridge the gap between bioinspired material innovation, system-level optimization, and data-driven generalization. It establishes a scalable and predictive design framework for passive adaptive thermal enclosures, enabling climate-responsive energy regulation without active control. The developed approach provides a foundation for the next generation of high-performance building envelopes and sustainable thermal management systems across multiple applications and climates
Using Fine-Tuned and Context-Aware Large Language Models for Requirements Technical Debt Evaluation in Systems Engineering Work Products
No full textRequirements Technical Debt (RTD) represents one of the most consequential yet difficult to detect forms of engineering debt, manifesting as ambiguous, incomplete, or inconsistent requirements that propagate costly downstream failures. In systems engineering (SE), RTD accumulates within document-heavy artifacts such as Systems Engineering Management Plans (SEMPs), where manual peer review—while essential—remains time-intensive, inconsistent, and prone to human oversight. This thesis investigates whether Large Language Models (LLMs), when grounded in expert reasoning patterns and authoritative standards, can replicate the analytical capabilities of Subject Matter Experts (SMEs) to detect and classify RTD at scale. The research employs a two-phase mixed-methods approach integrating qualitative expert knowledge elicitation with computational implementation. Phase I captured reasoning heuristics from 14 SE practitioners through structured interviews and surveys, synthesizing their cognitive processes into a machine-readable Requirements Debt Detection Guide. This formalized knowledge served as the foundation for Phase II: the development and validation of the SEMP Requirements Debt Analyzer (SRDA), an AI-assisted evaluation system implemented using AWS Bedrock, Claude 3 Sonnet, and Retrieval-Augmented Generation (RAG) architecture. SRDA combines Chain-of-Thought (CoT) reasoning with standardsbased contextual retrieval to produce transparent, traceable RTD assessments grounded in INCOSE, IEEE, NASA, and ISO guidance. Phase II validation with two senior SMEs demonstrated strong system performance across five evaluation dimensions, achieving an overall mean score of 4.03/5.00. SRDA exhibited particularly high performance in interpretability and traceability (4.17/5.00) and workflow impact (4.50/5.00), with evaluators characterizing the system as a valuable screening tool positioned between junior and senior engineer capabilities. Qualitative feedback highlighted SRDA’s effectiveness at identifying high-level deficiencies and accelerating review workflows while acknowledging limitations in capturing subtle, context-dependent judgment. The findings establish SRDA as a technically sound proof-of-concept that demonstrates LLMs can replicate surface-level requirements analysis when properly grounded in expert knowledge and authoritative standards. The system’s value lies not in replacing human expertise but in amplifying it—automating tedious scanning tasks, scaling systematic analysis, and enabling SMEs to focus cognitive effort on interpretive judgment and contextual reasoning that remain uniquely human capabilities