48440 research outputs found
Sort by
Introducing Computational Thinking to Pre-Adolescent Students through a Robotics Workshop
No full textThis research explores the development and implementation of a computational thinking (CT) curriculum for pre-adolescent students through a four-day robotics workshop. Our curriculum taught block-based programming for the Experiential Robotics Platform (XRP) while incorporating the Creative Computational Problem Solving (CCPS) model and visual-based storylines. Our findings indicate that CT was challenging for this age level, and the CCPS model was difficult to implement class-wide. Notably, we established that the low-cost Experiential Robotics Platform is viable for teaching younger age groups
Improving Community Engagement with Sustainable Development Goals at WPI
No full textThis project investigated the methods of a Copenhagen-based NGO called Dreamtown, and how they effectively teach both youth and adults about the UN Sustainable Development Goals (SDGs). By building on the efforts of Dreamtown, we hoped to incorporate stronger SDG education and involvement within the WPI community. This culminated in a set of activities and proposed curriculum changes that aim to make WPI students more familiar with the SDGs, and get them more involved with the sustainability efforts happening on campus
Shine Initiative
No full textWe worked in collaboration with the Shine Initiative to explore data collection methods, the prevalence of mental health education, and to recommend and begin implementation of a new tool to collect non-clinical mental health data. We interviewed people from eleven organizations and collected survey results from counselors in a local school district to gain insight on mental health education and data collection methods. We analyzed six different data collection tools and concluded that PowerBI was a good fit for Shine. We began implementation of PowerBI for Shine and developed an instruction manual to ease the organizations’ transition to this new data collection framework
MonoEye - Exploring Monocular and Multi-View Neural Reconstruction Pipelines for Robotic Grasping
No full textRobotic manipulators typically rely on RGB-D sensors or stereo vision to perceive depth for grasping and manipulation. While effective, these sensors suffer from limitations such as high cost, minimum depth range, noisy outputs, and incompatibility with lightweight or mobile systems. Motivated by these practical challenges, this thesis explores whether monocular vision—using only RGB images—can support 6-DoF grasping in real-world robotic settings. We investigate two monocular pipelines: (1) single-image depth estimation using a fine-tuned depth prediction model, and (2) multi-view 3D reconstruction via Gaussian Splatting. Both are integrated into a unified grasping framework that uses LangSAM for object segmentation and AnyGrasp for 6-DoF grasp prediction. This modular design enables consistent downstream processing across different depth sources. Through experiments in both simulation and real-world environments, we evaluate grasp success, runtime performance, and failure modes. Our results show that monocular pipelines can approach the performance of traditional RGB-D systems, while offering better adaptability in cases involving reflective, transparent, or cluttered objects. These findings suggest a viable pathway for lightweight, low-cost, and deployable robotic perception without dedicated depth hardware
Symbolic Model Extraction of Collective Behaviors in Homogeneous and Heterogeneous Swarms
No full textIn this work, I present a novel approach to symbolic regression tailored specifically for modeling both homogeneous and heterogeneous swarm behaviors. Swarm behaviors, as observed in animals like ants, bees, and birds, are a significant source of inspiration in research and engineering. However, constructing symbolic models from observational swarm data remains challenging due to the complexity of the interactions involved. Existing methods for simplifying collective behaviors often require extensive manual testing and creativity. To address this, I introduce the first symbolic regression approach capable of supporting a wide range of interactions within a swarm. The method is composed of two phases. The first phase employs a modified Graph Neural Network (GNN), which I call the GNN Multiplexer, to learn and capture the unique interactions within the swarm. This phase generates neural networks that approximate the relationships between any two individuals in the swarm. In the second phase, a modified nested evolutionary algorithm, Macro-Micro Evolution, leverages the data generated by the GNN Multiplexer to generate simple symbolic models that approximate these relationships and are more easily interpretable by humans. I validate this method through a series of case studies, including lattice formation experiments, flocking simulations, and real-world fish schooling data, assessing the performance of both the overall approach and its individual components
Resource Scheduling for Concurrent Mixed-Priority Workloads on General Purpose GPUs
No full textAs general purpose GPUs become more powerful, it becomes more difficult for any given GPU application to make use of the entirety of its computational resources. A common solution to this problem is concurrency, where multiple GPU applications are executed together on a single GPU so that more of its resources are utilized. However, this poses a novel challenge: latency-sensitive applications may experience lower responsiveness and higher turnaround times if the concurrently executing applications are occupying the resources that it needs when it arrives at the GPU. The main premise of this dissertation is that with fine-grained control of GPU resources and awareness of resource saturation points, we can leverage the colocation of thread blocks from different tasks on the same streaming multiprocessors to increase system utilization while protecting the quality of service requirements of latency-sensitive tasks. We first provide a characterization of the performance of currently available concurrency mechanisms on NVIDIA GPUs, analyzing how their lack of flexibility and coarse-grained resource control can lead to unpredictable performance and reduced system utilization. Next, we present a set of kernel profiles, hardware mechanisms, and scheduling components that are designed to reduce resource contention between tasks and effectively prioritize the execution of high priority latency-sensitive tasks over best-effort low priority tasks. Finally, we outline a thread block scheduling policy and register oversubscription mechanism that allow the GPU to maintain responsiveness for high priority tasks while also allowing opportunities to improve colocation
The User Experience System (UXS): Integrating Systems Engineering and Systems Thinking to Define, Operationalize, and Empirically Test Its Correlation with Commercial Success
No full textDespite decades of research into new product development, between 35% and 45% of new products still fail commercially, underscoring a persistent gap in understanding the systemic drivers of success. While prior studies attribute failures to factors such as weak positioning, poor timing, or inadequate investment, these explanations do not account for high-profile cases where technically sound, well-marketed products nevertheless failed. This research introduces the concept of a User Experience System (UXS): a lifecycle-oriented, system-of-systems framework that strategically integrates non-functional requirements (e.g., reliability, maintainability, usability) to optimize end-user outcomes. Although scholars have implied the importance of systemic experience, they have not defined a UXS, nor has any empirical study tested whether UXS integration predicts commercial success. The primary aim of this research was to determine whether a statistically significant correlation exists between the degree to which a product, system, or service is designed from a UXS perspective and its commercial success. Confirming such a relationship would provide actionable guidance for development teams, with implications for profitability, market resilience, and end-user satisfaction. This work contributes to the body of knowledge by: 1. Defining and operationalizing the UXS construct. 2. Empirically demonstrating its correlation with commercial success. By bridging systems engineering, systems thinking, and human-centered design, this research offers a new lens for both scholars and practitioners, advancing the theory and practice of product development while addressing a critical gap in the literature
Machine Learning Prediction of Clinical Youth Mental Health
No full textRecent increases in prevalence rates of anxiety and depression among middle school student populations has called for the urgent need to research these groups to develop effective targeted strategies to treat mental illness.This study was able to use a Random Forest Classifier (RFC) to predict varying levels of anxiety and depression among different subpopulations using the Revised Child Anxiety and Depression Scale shortened version (RCADS-25) which was given to middle school students in a Massachusetts school district. Through the effective RFC analysis (accuracy >90%) we were able to find diverse symptom importance for different subpopulations and severity levels. These findings highlight the need for personalized intervention to treat anxiety or depression in middle school students
Improving Walkability for Seniors in Upham's Corner
No full textIn the Dorchester neighborhood of Boston, Massachusetts, lies the bustling neighborhood of Upham’s Corner. In this project, our team evaluated walkability for seniors in Upham’s Corner and surrounding neighborhoods using a walk audit tool. Additionally, we orchestrated a focus group with local senior advocates to understand their motivations and barriers to walkability. Through this research, we found that although seniors value walking for their contribution to their quality of life, a lack of safety and poor infrastructure maintenance deters them. This report evaluates the walkability of Upham’s Corner, incorporates community feedback from local senior advocates, and presents our recommendations for potential improvements
Exploration of Attention Models and Implementation of Diffus
No full textWe compare attention-based and diffusion-based models for genotype-to-phenotype prediction. After reproducing two attention models, Rijal et al.’s shallow transformer and Arcadia Science’s multi-output canonical transformer, we confirm that Arcadia’s model achieves better accuracy across traits. We then implement a conditional denoising diffusion probabilistic model (DDPM) using the same yeast dataset. While diffusion models offer flexibility and uncertainty modeling, they underperformed attention models in accuracy and stability. Our findings suggest attention-based models, especially those using multitask learning, remain the most effective for phenotype prediction, though diffusion offers promise for future generative and hybrid approaches in biological modeling