Mason Journals (George Mason Univ.)
Not a member yet
3256 research outputs found
Sort by
A Vision-Language-Action Model Approach for Geospatially Guided Autonomous Navigation
No full textAutonomous navigation in dynamic environments requires robust computer vision to ensure obstacle avoidance along precomputed paths. In geospatial-based robotics, understanding visual input is essential to making route decisions, avoiding collisions, and reaching desired waypoints. While single-model approaches like YOLOv5 offer object detection, they often struggle with ambiguous or low-confidence visual inputs, especially in cluttered spaces. This project builds upon Google PaLM-SayCan and the vision-language action model paradigm, applying their core idea of grounding language models in decision-making to improve perception and precision using compact multimodal models. Furthermore, existing vision systems often lack contextual understanding and adaptive decision-making for human -centered environments. Instead of relying on large-scale LLMs, we explore the use of lightweight, robust models - most notably SmolVLM2, fine-tuned on COCO (Common Objects in Context) images for better contextual reasoning in visual scenes. Our multi-model perception pipeline integrates YOLOv8 for initial object detection, while SmolVLM2 acts as a fallback validator when detection confidence is low. We implement a fusion model of classic and generative techniques, creating a decision module that selects among 4 possible actions: proceed, detour, stop and query. Route planning is based on Dijkstra-generated waypoints, with the fused vision system influencing live adjustments. Overall, this approach achieved 88% detection precision and 91% obstacle avoidance success. The method is able to improve obstacle avoidance by 27%, and achieve significantly better throughput compared to YOLO-only inference. This work demonstrates that compact VLAMs can significantly improve the perception layer in autonomous navigation, especially in human-centric geospatial environments
Quantification of murine brain samples to characterize the correlation between the LTA4H expression and aged-related neuroinflammation biomarkers
No full textThe leukotriene A4 hydrolase (LTA4H) enzyme is implicated in age-related neuroinflammation, which may contribute to Alzheimer’s Disease (AD). LTA4H is an enzyme that plays a dual role in inflammation: 1) as an epoxide hydrolase (EH), LTA4H catalyzes the hydrolysis of leukotriene A4 (LTA4) to leukotriene B4 (LTB4) as an inflammatory response, and 2) as an aminopeptidase (AP), LTA4H catalyzes the hydrolysis of the tripeptide Pro-Gly-Pro (PGP) as an anti-inflammatory response. The two amino acid residues, Q136 and D375 are crucial for LTA4H activity. A mutation in Q136N demolished LTA4H AP activity while maintaining LTA4H EH activity, whereas a mutation in D375N had an exact opposite effect. For this study, we compared protein and lipid profiles in four groups of mice: young (less than 10 months), aged (over 18 months), LTA4H AP knockout, and LTA4H EH knockout. We utilized high resolution mass spectrometry technique to quantify LTA4H and other associated proteins in mice brain tissue. In addition, various lipid profiles associated with LTA4H were qualified with a lipidomics approach to detect inflammatory markers of, either AP pathway or EH pathway. We identified a significant difference in LTA4H expression between young and aged mice. This study will help support current literature suggesting neuroinflammation in AD
A Spatiotemporal Transformer Architecture for Long-Term Post-Wildfire Vegetation Recovery Forecasting
No full textThe increasing frequency and scale of wildfires in regions like California present a critical challenge for ecological management. Predicting post-fire vegetation recovery is essential for restoration efforts, yet it remains a complex problem due to the interplay of static topographical features and dynamic climatic conditions. Traditional remote sensing models, often based on solely statistical analysis or Convolutional Neural Networks, struggle to capture these long-range spatiotemporal dependencies effectively. To address this, a novel spatiotemporal transformer architecture for forecasting vegetation regrowth was utilized. The model leverages a Vision Transformer (ViT) backbone to extract rich spatial features from multi-modal data, including Sentinel-2 imagery, burn severity, and topography. A temporal transformer encoder-decoder then processes a 12-month sequence of these features alongside climatic variables to predict the following 24 months of Normalized Difference Vegetation Index (NDVI). The decoder utilizes a unique querying mechanism, combining a spatial hint (the last known NDVI map) with learnable temporal embeddings to dynamically generate each future monthly prediction. This approach was trained on a comprehensive dataset spanning 41 historical wildfires across California. The model achieved a preliminary Structural Similarity Index Measure (SSIM) of 0.58 and a Mean Squared Error (MSE) of 0.15 on the validation set. While initial visualizations confirm the model is just beginning to learn temporal dynamics, these results indicate a foundational capacity for capturing complex spatiotemporal patterns. Further training is expected to significantly enhance its ability to forecast dynamic changes. By integrating a ViT with a temporal transformer, our approach provides a powerful new tool for land managers, offering more accurate insights into ecosystem recovery trajectories and enabling more effective post-fire environmental planning
Tracking Pollution Downstream: Linking Agricultural Land Use to PFAS and Microplastic Pollution in U.S. Streams
No full textAgricultural land use is a major source of non-point pollution, which contributes to the spread of emerging pollutants, such as per- and polyfluoroalkyl substances (PFAS) and microplastics. Despite increasing concern, the relationship between agricultural land-use patterns and the occurrence of these pollutants in aquatic systems remains under-explored. This study investigates these links by analyzing environmental monitoring data from agricultural stream sites across the U.S, including PFAS concentrations, microplastic assessments, and related environmental variables, as well as land-use descriptors categorized by agricultural activity. Preliminary analyses feature exploratory data visualizations and correlation assessments to characterize pollutant distributions. Spatial regression models are applied to quantify the effect of land-use factors on contaminant variability. To improve interpretability and accessibility, a Shiny-based interactive dashboard is developed to visualize spatial patterns and the dynamic trends of the key variables in the study. Together, these tools aim to improve insight into the spatial structure of emerging contaminants and support reproducible, data-driven environmental monitoring in aquatic agricultural ecosystems
Redefining Autism Care: Innovative Solutions for Autism Diagnosis and Intervention
No full textAutism is a neurological disease that has serious effects on a child's social, cognitive, and emotional intelligence. It is optimal for children with autism to be diagnosed at a young age and receive treatment/intervention to enhance these traits. However, the intervention methods for children with autism remain largely ineffective in the United States. Traditional methods are often slow, resource-intensive, and yield inconsistent developmental outcomes, resulting in a 68% graduation rate and an average hourly income of $8.10, which is significantly lower compared to other disabled groups. Technology, over the years, has allowed researchers to understand the flaws of the developmental system for autistic children and use innovative measures to fix it. This paper conducts a cross-study examination of existing research to highlight improvement through innovative measures regarding intervention methods and early diagnosis, as well as systemic barriers that hinder a quality developmental process. Through extensive innovation using machine learning, researchers have built predictive models that have broken the age barrier of early diagnosis, establishing an intelligent screening system that analyzes brain imaging and genetic data to diagnose a child with autism at a few months old instead of the late age of two years old. Early interventions include Naturalistic Developmental Behavioral Interventions (NDBI) and Applied Behavior Analysis (ABA), which are effective in promoting the social and cognitive development of autistic children. Researchers have also highlighted socio-economic aspects that impact an autistic child's development, such as the classroom environment and systemic barriers like limited clinician access, household income, etc. Future directions include the conceptualization of a low-cost virtual persona and AI agencies designed for personalized intervention and continuous evaluation, acknowledging the latest AI technologies. This approach aims to bridge existing gaps in autism education by combining technological innovation with evidence-based practices to improve long-term developmental outcomes and reduce the cost
CORE: A Novel Cost-Effective Operational ROV Explorer
No full textThe high cost of commercial Remotely Operated Vehicles (ROVs) presents a significant barrier to widespread underwater research and infrastructure monitoring. CORE presents a novel, open-source, low-cost ROV designed for accessibility and mass deployment. The prototype improves upon traditional six-thruster designs by utilizing an efficient four-thruster configuration augmented with four servos. This innovative approach maintains full six-degree-of-freedom (6-DOF) motion while drastically reducing the manufacturing cost from 1,000 per unit. Another mechanism is a linear actuator system that adjusts the battery's position, changing the center of gravity of the ROV, thus providing pitch control without thruster power. A key performance enhancement is the doubling of vertical thrust capacity, as all four thrusters can be dedicated to heave maneuvers, which provides superior stability and payload capacity for various mission profiles. The control system simplifies piloting through a sophisticated algorithm featuring single-axis automatic stabilization. With a compact 15.5-inch long by 5-inch diameter frame, the ROV is highly portable and field-deployable in under 15 minutes. Its modular architecture is built around 3D and laser cut parts, a Raspberry Pi, integrating a sensor suite that includes an IMU, pressure sensor, acoustic modem, and a leak sensor. The design can be controlled through ethernet cable or acoustic modem. The versatile platform is ideal for applications including critical infrastructure inspection, environmental monitoring, aquaculture management, and search and rescue
Training for Instructors Teaching Inclusive Postsecondary Education Program Students
No full textThe passage of the Higher Education Opportunity Act in 2008 triggered an increase in inclusive postsecondary education (IPSE) programs across the United States, giving students with intellectual and developmental disabilities the opportunity to enroll in typical college courses. A multiple case study of IPSE programs at five institutions of higher education was conducted. Interviews of program directors and other program affiliates examined their roles in the development, components, and implementation of training programs, followed by observation of faculty development sessions and analysis of the documents used for training. Findings indicated no unified approach to faculty training, but similarities existed in the processes and resources used. Components included an introduction/overview of the program, inclusive instructional practices, and accommodations and modifications to course material
A Focus on Healthy Living Between Natural Supports and Young Adults with Intellectual Disability in College
No full textCollege students with and without intellectual disability (ID) are challenged with maintaining healthy lifestyles since it may be their first time living away from home. Natural supports (or peer mentors) can work together to promote healthier options on college campuses. This article shares how one college program promoted healthier lifestyles across three areas (i.e., nutritional understanding, dietary intake, physical activity). Dietary intake and physical activity were monitored across a 6-month period. Preliminary findings indicated 9 of 10 students with ID increased nutritional knowledge, 10 of 10 increased servings of fruits and vegetables consumed, and 6 of 10 increased cardiovascular endurance.
 
Qualitative Analysis of Peer Mentoring in an Inclusive Postsecondary Education Program for Students with Intellectual Disabilities
No full textAlthough peer mentors play a vital role in supporting college students with intellectual disability enrolled in inclusive postsecondary education programs, their use has not been explored extensively. This qualitative study examined the experiences and perspectives of university students serving as peer mentors in an inclusive postsecondary education program. Nineteen peer mentors participated in semi-structured interviews. Results provide additional insight as to why individuals choose to become peer mentors, and as a result of their mentoring experience, what the peer mentors learned about themselves and gained from their experience. Recommendations for recruiting peer mentors are also provided
