48440 research outputs found
Sort by
Bee-ing Adaptable: Copenhagen Beekeepers vs Shifting Seasonality
No full textClimate Change related shifts in seasonality have affected various aspects of beekeeping. This project explored how Copenhagen beekeepers perceive shifting seasonality and adapted their beekeeping practices in response to environmental changes. We explored seasonality, beekeeping practices, pests and diseases, and their impact on honeybees. Our team used methods such as surveys, interviews, site assessments, document analysis, and participant observation to understand Danish beekeepers' challenges and changes throughout the beekeeping year. We identified themes in their approaches to sustainable beekeeping in a changing environment. Our team developed a model to predict how bee colonies are impacted by pest populations
Physically Active Youth: Website Re-Design
No full textPhysically Active Youth (PAY) is a holistic non-profit afterschool program in the Katutura township of Windhoek, Namibia. Through ethnographic observations, focus groups, and semi-structured interviews, we explored the positive impact that PAY has on all of its stakeholders. To showcase this profound impact that the program has on its student learners, volunteers/staff, and alumni, we created a redesigned website for PAY using the software Framer. The website pages, key features, and content were all informed with background literature research and the social science data collected consulting with the PAY community. To ensure their digital footprint can be maintained continuously into the future, we created a manual walking through the website and its content as well as linking to online resources from the Framer company
Breast Thermal Patterning in Response to Reproductive Hormones and Exercise
No full textBreast thermography has potential as a non-invasive monitoring device for breast health, but its clinical utility in reproductive and exercise physiology remains limited by a lack of normative data on how hormones and physical activity modulate breast surface temperature. In this study, we recruited five healthy, regularly menstruating women (ages 21–23) and monitored them over one complete menstrual cycle (30 days). Daily first-morning urine samples were tested for indicators of estrogen (estrone-3-glucuronide (E3G)), progesterone (pregnanediol-3-glucuronide (PdG)), luteinizing hormone (LH), and follicle-stimulating hormone (FSH). Resting breast temperature was measured biweekly using infrared thermography and skin-mounted thermistors. Once weekly, participants completed a 20-minute exercise protocol at 80–90% of age-predicted maximal heart rate, with post-exercise breast temperature recorded over time. Exercise induced a clear, statistically significant change in breast surface temperature from baseline (unpaired t = 3.64, p = 0.0003; paired t = 2.75, p = 0.0066). Repeated-measures ANOVA revealed cyclic variation in resting temperature across menstrual phases (F(3,511) = 10.29, p < 1×10⁻⁶) and significant inter-individual differences (F(4,511) = 29.57, p < 1×10⁻⁶). A global multiple linear regression of breast temperature on LH, E3G, PdG, and FSH yielded an R² = 0.117, with PdG emerging as the strongest single predictor (β = 0.144). Sensor-specific regressions showed modest regional sensitivity, with some locations achieving R² = 0.150 for PdG. Mixed-effects modeling (random intercepts for ParticipantID) rendered all hormone coefficients non-significant (p > 0.98), underscoring the importance of personalized baselines. Pearson correlations suggested PdG’s thermogenic role (r = 0.234). Models predicting post-exercise temperature change from hormone levels performed poorly (all R² < 0), suggesting that these hormones do not explain behavior of thermal changes in the breast following exercise. These findings suggest that progesterone-related and exercise-induced thermal patterns produce distinct breast thermal signatures, but also highlight substantial inter- and intra-individual variability. This supports the need for personalized calibration in clinical thermography and points toward future integration of machine-learning approaches to interpret spatial–temporal temperature patterns for reproductive health monitoring. These results establish baseline breast thermal variability associated with normal hormonal cycling and exercise, providing reference data that could improve the development of wearable thermographic monitors for monitoring fertility and breast health
Mining Backscatter Correlations for Soil Moisture Approximation
No full textThe ability to assess sub-surface soil volumetric water content (VWC) has many advantages in agriculture, construction, and paleohydrology, both as a means to warn of vehicle immobilization and addressing potential floods. The proficiency to accurately categorize VWC at a stand-off to re-route and avoid probable hazards is paramount. In this work, we constructed a soil testbed, saturated under varying controlled stages until capillary action is achieved. The soil is probed with an air-coupled step-frequency continuous-wave (SFCW) radar operating in the UHF-L band continuously throughout the duration of the experiment. Mono- and bi-static reflection, transmission, and attenuation radar backscatter data is collected to find the optimal forward looking antenna configuration to acquire low clutter hydrological parameters. Correlations were mined and integrated with machine learning to develop a stand-off VWC approximator. This equation ultimately obtained a soil moisture prediction average accuracy of 91.5% at 10 inches beneath the soil surfac
Saranga: milliWatt Ultrasound Navigation On Palm-Sized Aerial Robots for Visually Degraded Scenes
No full textTiny palm-sized aerial robots possess exceptional agility and cost-effectiveness in navigating confined and cluttered environments. However, their limited payload capacity directly constrains the sensing suite on-board the robot, thereby limiting critical navigational tasks in GPS-denied wild scenes. Common methods for obstacle avoidance use RGB cameras and LiDAR, which become ineffective in visually degraded conditions such as low visibility, dust, fog or complete darkness. Other sensors, such as RADAR, have high power consumption, making them unsuitable for tiny aerial robots. Inspired by bats, we propose Saranga, a low-power ultrasound-based perception stack that localizes obstacles using a dual sonar array. We present two key ideas to combat the low signal to noise ratio: physical noise reduction and a deep learning based denoising method. In the first idea, we find an optimal and practical way to block propeller induced ultrasound noise on the weak echoes. In the second idea, we generate and train a denoising neural network to utilize long-horizon for finding signal patterns under extreme amounts of uncorrelated noise. We generalize to the real-world with no real data for training. For the first time ever, we enable a palm-sized aerial robot to navigate in visually degraded conditions of smoke, darkness, and snow in a cluttered environment with thin and transparent obstacles using only on-board sensing and computation. We provide extensive real-world results to demonstrate the efficacy of our approach
Enhancing Engagement Through Sustainability Education
No full textAs Earth's natural resources dwindle, the promotion and adoption of sustainable living practices are vital. In collaboration with the Appalachian Mountain Club (AMC), we designed four 'green tours' to showcase sustainability features at AMC lodges in the White Mountains and Maine regions. We collected data from six expert interviews, four focus groups with AMC staff, 19 visitor surveys, and three tour pilots. Using this data, we created a series of visuals, including charts, tables, and matrices, to extract trends and patterns, enabling us to identify the key features of engaging educational tours. Subsequently, we developed four professional-looking, brochure-style self-guided tours using Canva, as well as accompanying scripts for a staff-guided tour variation
AI in Education
No full textAs AI becomes more accessible and widely used by both students and educators, it continues to evolve rapidly, leaving many lecturers behind. This study examines the application and impact of generative AI tools in contemporary education among lecturers and students to develop recommendations for future AI use in education. To address this, semi-structured interviews with lecturers at the ZHAW Wädenswil campus in Zürich, Switzerland, were conducted, examining attitudes toward AI and its classroom roles. Student surveys were distributed to judge students' opinions on AI use amongst themselves and lecturers, as well as how they are currently using AI. This culminated in our recommendations for AI adaptation and use at ZHAW Wädenswil, which transcend the barrier between lecturers and students
A Comparative Stock Market Simulation of Trend-Following and Mean-Reversion Trading Strategies
No full textThis project conducted a simulation-based comparison of trend-following and mean-reversion trading strategies to determine which is more effective in the uncertain 2025 market. Operating within a controlled four-week framework, both strategies were applied to a common portfolio of six U.S. large-cap stocks, starting with 1,150.80 (+1.15%) from a single, successful trade. In contrast, the more active trend-following strategy ended with a smaller, unrealized gain of $601.86 (+0.60%) across four open positions. The outperformance is attributed to the prevailing market conditions during the simulation, which were characterized by short-term volatility and choppy, directionless trading, an environment better suited to capitalizing on price oscillations than sustained momentum. The study concluded that strategy performance is highly regime dependent and demonstrated that in a market lacking strong directional trends, a patient, risk-averse approach can be more profitable than a strategy designed to ride momentum
Disentangling heterogeneity in risk and protective factors of longitudinal substance use using machine learning latent profile analysis
No full textThis project applied machine learning–based latent profile analysis to adolescent cannabis use data to identify long-term cannabis usage patterns. Combining demographic, behavioral, and social indicators, in addition to automation tools, this approach modeled distinct usage trajectories and various risk profiles, highlighting factors that predict sustained use versus decline. This study demonstrates the value of incorporating machine learning techniques with advanced statistical and computational methods for uncovering hidden subgroups in public health data, as well as provides insights that can enable targeted prevention and intervention strategies
AdvancedPBL Newsletter, Spring 2025
No full textQuarterly email newsletter created by the WPI Center for Project-Based Learnin