麻豆传媒社区

In this project, we will use commercial-off-the-shelf wireless sensors to collect data about the physical environment, where human subjects perform daily activities. The idea is when wireless signals will interact with and bound off the physical environment including human subjects. Different movements will result in different signal patterns.

We will analyze the collected data to discover data patterns and understand their correlations with human activities using AI tools, such as machine learning, deep learning, and data visualization. We will investigate the opportunities and challenges in leveraging the discovered data patterns and correlations for healthcare applications, such as sleep quality assessment, gait pattern analysis, fall detection and fall risk prediction.

Students will develop research skills, such as surveying literature, identifying problems, designing experiments, analyzing results, presenting findings and insights.

Students will gain hands-on experience with sensor systems.

Students will gain hands-on experience with cutting-edge AI technologies.

Students will develop a good understanding of methodologies for conducting scientific research.

Students will develop teamwork and leadership skills from this project.

Students should meet with the advisor weekly, during which students should communicate their progress with the advisor.

Students are expected to summarize their results and the status of the planned action items. 

Students are expected to discuss and identify the planned action items (e.g., literature review, ideation, experiment design, algorithm implementation, evaluation and analysis) for the next week.

Students are expected to reach milestones in a timely manner.

Hybrid

Dr. Zongxing Xie, zxie1@kennesaw.edu 

Nanomedicine which uses nanoparticles for therapeutic and diagnostic applications has emerged with great potential to improve therapeutic and diagnostic efficacy. Among various nanoparticles developed for nanomedicine, magnetic nanoparticles have attracted attention due to their unique magnetic properties that can be applied for cancer thermal therapy and disease diagnostics using imaging modalities and biosensors. Although magnetic nanoparticles have shown promising properties for these applications, the constraints generated by a complex biological system hinder their performances and delay the translation of this promising technology. To overcome this challenge, this project proposes to develop artificial intelligence (AI) models that output optimal design spaces and properties of magnetic nanoparticles considering the constraints. Furthermore, this project will build a prototype magnetic sensor to detect magnetic nanoparticles for diagnostic applications.  

The student who performs the AI model task will learn AI model development and analysis using python programming for biomedical applications.

The student who performs the sensor task will learn sensor construction and processing of sensor generated signals.  

The student will conduct literature review, work on AI model/sensor construction, and analysis.

Hybrid

Dr. Asahi Tomitaka, atomitak@kennesaw.edu 

Artificial intelligence, particularly Large Language Models (LLMs), plays an important role in improving decision-making in real-world applications, such as question answering and classifications. The LLM model embeds an extensive amount of knowledge and responds based on human instructions/prompts. This human-AI teaming partnership yields better decision-making outcomes in complex information environments. The Retrieval-Augmented Generation (RAG) system utilizes the LLM knowledge to retrieve key requested information from external data. However, the quality and factual trustworthiness remain challenging and require forming a team with humans for better alignment and validations. In this project, we plan to develop a framework to form teams for enhanced RAG performance and improve the quality of the retrieved information. We plan to develop two-tiered human-machine collaboration, i) Interactive human curation of retrieved data, ii) Personalized retrieval models that adapt to user behavior. In this project, students will configure and experiment with pre-built RAG pipelines (e.g., LlamaIndex, Haystack) to investigate and analyze how human-in-the-loop review can improve retrieval quality and reduce hallucinations.

• Develop skills in reading and synthesizing scientific literature on AI and RAG.
• Gain real-world experience configuring and evaluating retrieval-based AI systems.
• Learn experimental design, parameter tuning, and evaluation methodology.
• Improve communication and scientific writing skills.
• Present research findings at a poster session or technical seminar.

• Read papers, technical articles, and framework documentation.
• Run pre-configured RAG pipelines with the provided datasets.
• Adjust retrieval and prompt parameters; run controlled experiments.
• Annotate and evaluate retrieved results for accuracy and trustworthiness.
• Meet weekly (in person or virtually) to report progress and discuss findings.

Hybrid

Dr. Md Abdullah Al Hafiz Khan, mkhan74@kennesaw.edu 

Abdul Muntakim, amuntaki@students.kennesaw.edu 

Abm. Adnan Azmee, aazmee@students.kennesaw.edu 

Francis Nweke, fnweke@students.kennesaw.edu 

With the advancement of Artificial Intelligence (AI) technology, it is used in every part of our lives, e.g., remote sensing, healthcare, finance, security systems, autonomous vehicles, cybersecurity, transportation system, text generation, and human-AI Teaming. While AI has achieved state-of-the-art performance in these domains, its black-box nature remains a critical challenge. Malicious actors can utilize adversarial attacks to access and manipulate the AI-based system. The attackers can use adversarial example attacks to craft a poison sample in the inference phase to predict incorrect information or perform a model poisoning attack to steal confidential data. To address these limitations, we must enhance trust in machine learning models.

In this project, students will create trustable machine learning frameworks to make the model secure, explainable, confidential, and private. Students will explore the potential security vulnerability primarily in AI models, but this can be extended according to research findings. Students must read research papers, government websites, newspapers, and white papers to prove their hypothesis. Students will learn and apply machine learning algorithms and adversarial attacks to demonstrate their hypotheses empirically, starting from a foundational level.

• Investigate cutting-edge AI Technologies.
• Learn machine learning algorithms and machine learning library, e.g., PyTorch and TensorFlow
• Develop scientific writing skills by reading, critically analyzing, and discussing key scientific conference papers with mentors.
• Improve presentation skills by participating in conference/workshop venue based on the research outcome.
• Advance communication skills based on regular meeting with mentors.
• Collaborate with other graduate/undergraduate students.

• Attend weekly scheduled meeting in-person/online, present weekly outcome and receive mentor鈥檚 feedback.
• Read research papers and summarize the finding in a scientific paper format.
• Learn Machine learning library, e.g., PyTorch and TensorFlow
• Execute and modify (if needed) a sample ML/AI source code and generate and analyze results.
• Implement algorithm and execute experiments to get hands-on experience in cutting edge research.
• Assist mentor writing research papers or submitting grants.

Hybrid

Dr. Kazi Aminul Islam, kislam4@kennesaw.edu 

Dr. Md Abdullah Al Hafiz Khan, mkhan74@kennesaw.edu 

Deep learning (DL) models have become central to many vision-based applications, from facial recognition and autonomous driving to medical imaging. However, as these models are increasingly deployed on edge computing platforms such as IoT devices and microcontrollers, new challenges arise regarding both efficiency and security. Unlike large-scale cloud-based systems, edge devices are constrained by limited computing power, memory, and energy availability, which creates a need to deploy lightweight DL architectures for processing. At the same time, adversarial attacks, such as adversarial patches, pose a unique risk in such scenarios, as they can be applied to physical objects and remain effective across different viewpoints and lighting conditions. 

This combination of resource constraints and physical-world vulnerabilities creates an urgent need for new methods to evaluate and improve the robustness of lightweight models operating in edge environments, a challenge that will be studied in the current First-Year Scholars project. To explore the challenges of this setup, the student(s) will develop and deploy lightweight Convolutional Neural Networks (CNNs) models on an Arduino device using TensorFlow Lite. The project will also focus on the generation of physical adversarial patches, produced from both lightweight and standard CNN models, and evaluate their transferability across different model architectures running on the Arduino device. Beyond attack analysis, the project will also investigate simple defenses that can improve resilience without exceeding the strict computational limitations of edge hardware. 

1. Learn to train and evaluate lightweight deep learning models for edge devices.
2. Gain hands-on experience with adversarial attack and defense techniques on ML and hardware.
3. Gain practical skills in deploying and testing models on Arduino hardware.
4. Develop presentation skills and scientific research skills.

1. Review research papers on adversarial attacks, defenses, and lightweight models for edge computing.
2. Train and evaluate lightweight and standard Deep Learning object detection models on resource-constraint edge hardware.
3. Generate and test printed adversarial patches to study real-world attack transferability.
4. Document results and prepare progress updates for weekly meetings.
5. Prepare project poster presentation and demo for the annual symposium.

Hybrid

Dr. Michail Alexiou, malexiou@kennesaw.edu 

This project will provide the student with the opportunity to contribute to a research study examining the classification of misinformation originating from both human and artificial intelligence (AI) sources. With the increasing prevalence of AI-generated text, particularly in the context of news media, there is a growing need to understand the extent to which AI-generated fake news differs from human-generated fake news in terms of linguistic characteristics and detection difficulty.

The student will utilize publicly available, pre-annotated datasets comprising true and fake news articles produced by both human authors and generative AI systems. The primary objectives of the project are twofold: (i) to systematically compare the classification performance of machine learning models in detecting fake news across AI and human sources, and (ii) to identify the most discriminative linguistic and semantic features contributing to accurate classification within each source type.

The student will implement a range of natural language processing (NLP) methodologies to preprocess text data, extract salient features, and apply both classical machine learning algorithms (e.g., logistic regression, random forests) and advanced deep learning models (e.g., transformer-based architectures). Furthermore, the student will apply model interpretability techniques, including SHAP (SHapley Additive exPlanations), to quantify feature importance and elucidate the linguistic signatures most indicative of misinformation across distinct generative sources. This project will provide comprehensive training in applied machine learning, interpretable AI methodologies, and critical analysis of textual data in the context of contemporary misinformation challenges.

• Learn how to work with text data, including basic cleaning, formatting, and preparation of datasets for analysis.
• Gain experience using R / Python programming tools for natural language processing (NLP), such as tokenization, word frequency analysis, and simple text feature extraction.
• Become familiar with running basic machine learning models for classification and evaluating model accuracy.
• Acquire skills in scientific paper writing and conference presentations.

• Attend regular weekly research meetings to discuss progress, receive feedback, and set goals for upcoming tasks.
• Review relevant background literature on misinformation detection, AI-generated text, and natural language processing techniques.
• Conduct data preprocessing tasks, including cleaning, formatting, and organizing text data for analysis.
• Implement and evaluate machine learning models using R / Python.
• Document all procedures and findings in well-organized progress reports.
• Prepare summary figures, tables, and visualizations for inclusion in the final publication or conference presentation.
• Contribute to the drafting of a research poster and present findings at the Symposium of Student Scholars.

Hybrid

Dr. Dhrubajyoti Ghosh, dghosh3@kennesaw.edu 

Cancer remains one of the most pressing health challenges of our time, and early detection is often the key to saving lives. Pathologists traditionally diagnose cancer by carefully examining histopathological slides under a microscope, but this process is time-consuming, labor-intensive, and sometimes subjective. In today鈥檚 era of Artificial Intelligence (AI), we have an extraordinary opportunity to use computers to assist doctors by analyzing medical images faster and more consistently.

This project invites first-year students to participate in an exciting and meaningful research experience where they will work with histopathological images, which are high-resolution pictures of tissue samples used to detect cancer. Students will learn how to process these images using modern AI tools, focusing on computer vision methods that allow machines to 鈥渟ee鈥� and recognize cancerous patterns. They will also explore Natural Language Processing (NLP), which is the field of AI that enables computers to generate understandable written reports from complex data. The ultimate goal is to build systems that can not only detect cancerous regions in medical images but also generate clear, human-readable summaries to support doctors and patients.

To ground the research in real-world impact, students will study multiple cancer types, such as breast cancer, colorectal cancer, lung cancer, and prostate cancer. For example, the BreakHis dataset provides nearly 8,000 breast cancer images across different magnifications; the NCT-CRC-HE-100K dataset includes over 100,000 colorectal cancer images; and the Cancer Genome Atlas (TCGA) hosts whole-slide images for lung and prostate cancers along with clinical data. Students will work with these publicly available resources, ensuring they gain hands-on experience with authentic medical datasets used by researchers worldwide.

Resources will include high-performance computing systems, Python-based AI libraries (PyTorch, TensorFlow, Scikit-learn), and specialized medical imaging tools (3D Slicer, SimpleITK, etc.).

By the end of this project, students will not only contribute to an important area of healthcare research but also develop transferable skills in data science, programming, image analysis, and scientific communication. This research combines three of today鈥檚 most impactful areas (AI, medical image processing, and NLP), giving students a truly interdisciplinary experience at the forefront of innovation.

Students participating in this project will gain valuable technical, analytical, and professional skills. Specifically, they will:

• Programming & Data Science: Learn the fundamentals of Python programming and how to use AI libraries (e.g., PyTorch, TensorFlow, Scikit-learn) for data analysis.
• Medical Image Processing: Understand how to handle high-resolution histopathological images, preprocess data, and apply computer vision techniques for cancer detection.
• Artificial Intelligence & Machine Learning: Gain exposure to supervised learning, convolutional neural networks (CNNs), and methods for model evaluation (accuracy, sensitivity, specificity).
• Natural Language Processing (NLP): Learn how AI systems generate clinical-style text reports, bridging raw computational output with human-friendly language.
• Ethics & Interdisciplinary Thinking: Reflect on the ethical challenges of AI in medicine, including privacy, bias, and the importance of human oversight.
• Research Communication: Develop the ability to explain findings clearly through written reports, oral presentations, and visual posters.

These outcomes are directly transferable to careers in healthcare, data science, computer science, and biomedical research. More importantly, students will leave the program with the confidence that they can tackle complex, real-world problems using computational methods, which is an empowering outcome for first-year scholars beginning their academic journey.

Students will engage in weekly, hands-on research activities under faculty guidance. Typical duties will include:

1. Learning and Training: Completing guided tutorials on Python, AI libraries, and image processing methods.
2. Data Handling: Downloading, organizing, and preprocessing histopathological image datasets (e.g., resizing, normalization, and annotation review).
3. AI Model Development: Training and testing machine learning models for cancer detection (e.g., distinguishing benign vs. malignant breast tumors using BreakHis data).
4. NLP Report Generation: Assisting in building simple systems that convert AI outputs into text-based summaries of cancer findings.
5. Critical Analysis: Reviewing model results, identifying errors, and discussing strategies for improvement.
6. Team Meetings: Participating in weekly research group discussions to share progress, troubleshoot challenges, and plan next steps.
7. Communication: Preparing short reports or presentations summarizing weekly progress and results.

By the end of each semester, students will contribute to a collective research report and potentially present findings at KSU鈥檚 Undergraduate Research Showcase or similar venues.

Face-to-Face

Dr. Muhammad Imran, mimran2@kennesaw.edu 

As healthcare systems become increasingly digitized, they face growing risks from cyberattacks that can compromise sensitive patient data, disrupt hospital operations, and even threaten lives. From ransomware attacks shutting down hospitals to medical device vulnerabilities, healthcare cybersecurity is one of the most urgent and high-impact areas in information technology today.

In this project, students will explore how cyber threats affect electronic health records (EHRs), medical imaging systems, wearable health devices, and telemedicine platforms. We'll look at the common vulnerabilities in healthcare IT systems and the unique challenges that make the healthcare industry a prime target for attackers. Students will study real-world incidents, examine emerging threats, and review defense strategies such as encryption, access control, network segmentation, and zero-trust architecture.

The main deliverable of this project is a survey paper written by the student that synthesizes existing research and provides insights into best practices for protecting healthcare data and systems. The paper will aim to educate peers and potentially contribute to a future publication or grant proposal. Students will be guided through the entire research process, from identifying relevant literature to organizing findings and improving scientific writing.

• Gain an understanding of core cybersecurity principles in healthcare systems.
• Learn to conduct a thorough literature review on technical topics.
• Improve skills in research organization, academic writing, and citation.
• Develop the ability to analyze real-world cyberattacks and defense strategies.
• Enhance public speaking and presentation skills through weekly updates and symposium participation.
• Explore interdisciplinary topics that combine IT, public health, and ethics.

1. Read and summarize 2鈥�3 academic or industry papers on healthcare cybersecurity each week.
2. Maintain a running annotated bibliography of relevant sources.
3. Meet weekly with the mentor to present findings and receive feedback.
4. Draft sections of the survey paper and revise based on mentor comments.
5. Attend occasional guest seminars or virtual talks related to healthcare IT or security (optional but encouraged).

Hybrid

Dr. Liang Zhao, lzhao10@kennesaw.edu 

This project, Smell-GPT: Enabling Language Models to Perceive Smell via Electronic Noses, explores a novel intersection between sensor-based olfaction and artificial intelligence. The goal is to introduce undergraduate students鈥攑articularly first-year students鈥攖o hands-on research that combines physical sensing, data collection, and large language model (LLM) reasoning in an accessible and engaging way.

Electronic noses (e-noses) use sensor arrays to detect and characterize odors by capturing the chemical signatures of volatile compounds. While such sensors are increasingly used in food safety, environmental monitoring, and health diagnostics, interpreting their outputs remains a technical challenge. In parallel, large language models like GPT-4 have demonstrated remarkable capabilities in understanding structured prompts, performing classification tasks, and reasoning over unfamiliar data when given examples.

In this project, students will collect e-nose data from a variety of common odor sources (e.g., vinegar, lemon, coffee, garlic) in a controlled setting. The focus will be on building a small, labeled dataset representing these odor categories. Rather than training complex machine learning models, students will explore how this raw or lightly preprocessed sensor data can be formatted and presented to language models for classification using prompt-based approaches.

For example, students will design structured prompts that present the LLM with sensor readings from known samples and ask it to predict the class of an unknown sample. By experimenting with different prompt styles, sample orders, and data representations, students will gain hands-on experience in prompt engineering and understand how LLMs can be used for basic reasoning tasks鈥攅ven in domains they were not explicitly trained for.

This project emphasizes accessibility, creativity, and foundational skill-building. Students will learn about sensor technology, data formatting, supervised classification, and natural language prompts, while also reflecting on the broader implications of sensory AI. Deliverables may include a small annotated dataset, a summary of prompt-based experiments, and a poster or demo showcasing how AI models can begin to "understand" smells.

By focusing on a tangible and creative task with minimal technical barriers, this project serves as an ideal entry point into interdisciplinary AI research鈥攅quipping first-year students with both curiosity and confidence.

Students working on this project will develop technical and computational skills through hands-on research in sensor data processing and artificial intelligence. This project is ideal for students with an interest in programming, machine learning, or computational science.

Participants will gain practical experience in the following areas:

• Sensor Interfacing & Data Collection: Students will operate an electronic nose (e-nose) to collect time-series data from various odor sources. They will learn how to sample, store, and label sensor readings consistently for downstream analysis.
• Python Programming & API Usage: Students will write Python scripts to interface with the e-nose and to interact with large language model (LLM) APIs (e.g., OpenAI鈥檚 GPT-4). They will learn to format input data, send structured prompts, and process model responses.
• Data Preprocessing & Feature Engineering: Students will perform basic data preprocessing such as normalization, smoothing, and dimensionality reduction (e.g., PCA) to extract meaningful features from sensor data.
• Prompt Engineering & Model Evaluation: Students will design and test structured prompts for LLMs, evaluating classification performance using accuracy, precision, recall, and confusion matrices. They will conduct experiments with few-shot and zero-shot prompting techniques.
• Scientific Computing & Documentation: Students will document their methods and results using Jupyter notebooks, version control (Git), and clear code commenting. They will summarize findings in technical posters or reports.

By the end of the project, students will have developed a basic end-to-end AI pipeline鈥攆rom physical sensing to computational inference鈥攁nd will have quantitative evidence of their work through small datasets, classification metrics, and reproducible scripts. This experience will prepare students for future research in AI, data science, or engineering.

Students participating in this project will engage in a variety of hands-on and learning-based activities each week to support the research project.

Each week, students will meet in person for a 1-hour research meeting with the faculty advisor or a graduate student research assistant. In these meetings, students will report their progress, discuss any challenges, and receive feedback and guidance for the next steps. These meetings are also opportunities to learn how research teams collaborate and solve problems.

Throughout the week, students will work on the following tasks:

• Data Collection and Processing
  Students will collect odor data using an electronic nose device and prepare it for use in the project by organizing and labeling it carefully.
• Python Programming
  Students will write and modify simple Python scripts to help build the data pipeline. This includes formatting the data and sending it to an AI model for interpretation.
• Model Evaluation and Experimentation
  Students will test how well the AI model performs in recognizing different smells. They will run experiments and track how accurate or consistent the results are.
• Reading and Independent Learning
  Each week, students will be expected to read short articles, tutorials, or book chapters related to the project. These resources will help them build skills in areas like data handling, AI concepts, or sensor basics.
• Documentation
  Students will keep records of their work, including code, results, and observations, to help with communication and reproducibility.

This project provides a consistent and supportive weekly routine designed to grow students鈥� skills in programming, data analysis, experimentation, and research communication.

Hybrid

Dr. Taeyeong Choi, tchoi3@kennesaw.edu 

This research project explores the potential of wearable exoskeletons to support nurses in physically demanding tasks and reduce the risk of injury and burnout. Nurses often perform repetitive lifting, bending, and patient transfers that place significant strain on the body, contributing to high rates of musculoskeletal injuries and fatigue. As the healthcare system continues to face growing demands, innovative solutions are needed to protect the well-being of frontline workers. Exoskeletons are wearable devices designed to provide mechanical support to the body, particularly the back and arms, to assist with heavy lifting and prolonged physical effort.

This study aims to evaluate the usability, comfort, and effectiveness of several exoskeleton models in realistic nursing scenarios. The focus will be on how these devices affect muscle activity, movement patterns, and perceived workload. First-year student assistants will play an essential role in supporting the data collection process. Responsibilities will include preparing the research lab, assisting with the setup and calibration of equipment such as motion sensors and video cameras, and supporting participants as they complete tasks with and without exoskeletons. Students will help attach sensors, monitor participant safety, record observations, and distribute surveys assessing comfort and ease of movement.

Technical Skills:

• Setting up and calibrating advanced equipment such as motion capture systems, muscle sensors, and video cameras.
• Safely attaching sensors and monitoring participants during simulated nursing tasks.
• Collecting and organizing biomechanical data, as well as distributing and compiling participant surveys.
• Following established research protocols to ensure accurate and reliable data collection.

Professional and Transferable Skills:

• Teamwork and collaboration by working closely with faculty, graduate students, and peers in the HOPE Lab.
• Communication skills through interacting with study participants and sharing observations with the research team.
• Problem-solving and adaptability while assisting with experimental sessions and equipment setup.
• Data management and organization, skills that are useful in both academic and professional settings

Each week, student assistants will be actively involved in supporting the research process within the HOPE Lab. Their duties will include preparing the lab for study sessions by setting up equipment, organizing materials, and ensuring a safe environment for participants. Students will assist with attaching and calibrating motion sensors, muscle sensors, and video cameras, learning step-by-step how to handle advanced research tools.

During study sessions, students will help guide participants through simulated nursing tasks, such as lifting weighted mannequins or transferring patients, both with and without exoskeletons. Responsibilities will include monitoring participant safety, recording observations, and distributing short surveys to collect feedback on comfort, usability, and ease of movement. Students will also learn how to troubleshoot basic issues with equipment under the supervision of faculty and graduate mentors.

Outside of data collection sessions, students will assist with organizing, labeling, and cleaning datasets to ensure accuracy and completeness. They may also help with simple data entry and review of participant surveys. Regular team meetings will provide opportunities for students to share observations, ask questions, and contribute ideas for improving study procedures.

Face-to-Face

Dr. Maria Valero de Clemente, mvalero2@kennesaw.edu 

Dr. Valentina Nino, lvallad1@kennesaw.edu 

Data-driven scientific applications may su铿�er from significant overhead for loading huge amount of scientific data from storage servers to compute nodes, leading to CPU/GPU stalls and longer workflow execution time. The emerging computational storage devices offer the opportunity to reduce the overhead of data transfer by offloading computation to the devices for in-storage computing. However, the existing IO stack only achieves sub-optimal performance for computational storage devices for three reasons. (1) The random-access pattern of data-driven applications makes the workloads cache unfriendly. (2) No e铿�ective IO interface exists for users to utilize computational storage devices. (3) There is a lack of coordination among IO layers on when and where to execute the offloaded computations. In this project, we propose a set of new designs focusing on HDF5-compatible interface, cross-layer runtime for executing offloaded kernels, and co-scheduling of IO and computation to address these issues. We will implement them in open-source software 位-HDF5. Real-world scientific applications will be able to use 位-HDF5 to access computational storage devices with performance improvement.

1. Learn the state-of-the-art techniques used in file and storage systems and systems for AI.
2. Learn about the simulation, emulation, and hardware platforms for storage system evaluation.
3. Learn the entire process of science from making observations and generating hypotheses through publishing scientific papers.
4. Increase communication skills by presenting papers in lab meetings and increase the ability to work in a research team.
5. Increase critical thinking and data interpretation skills.

1. Students will help develop and evaluate the 位-HDF5 software.
2. Students will meet faculty advisors to discuss the project progress.
3. Students will attend weekly group meetings.

Hybrid

Dr. Xuechen Zhang, xzhang54@kennesaw.edu 

Augmented Reality (AR) smartglasses let us see digital information directly in our line of sight. Imagine walking outdoors while navigation directions, maps, or helpful tips about your surroundings appear seamlessly in front of you. Recent advances are making AR headsets more practical than ever: lighter designs make them comfortable to wear for longer periods, Vision Positioning Systems (VPS) provide centimeter-level tracking accuracy so your headset knows exactly where you are, and powerful artificial intelligence (AI) acts like a 鈥渟econd brain,鈥� ready to assist with everyday tasks.

This project, not limited to labs, will explore how to use the latest AR headsets for outdoor, on-the-move, real-world scenarios. We will combine technologies like VPS, AI, and creative interaction techniques, along with elements of game design, to create engaging and useful applications. You might: 

• Design AR navigation that gives hands-free, real-time directions to your destination
• Build interactive field guides to explore historical landmarks, plants, or wildlife
• Create safety systems that warn about hazards in construction zones or busy streets
• Invent collaborative outdoor games where the entire campus becomes your game board

As a first-year scholar on the project, you will have the opportunity to design, prototype, and test AR applications that go beyond the lab. You will work with cutting-edge AR devices, learn about spatial computing, experiment with interaction methods, and explore how AI can make AR experiences more personalized and intelligent. You will also test your prototypes with real users and collect data to evaluate your designs using psychology-based research methods. 

The goal is to discover what today鈥檚 AR technology can (and can鈥檛) do in real-world situations and to imagine new ways it could enhance daily life. Along the way, you鈥檒l gain hands-on experience in creative problem-solving, teamwork, user experience design, and emerging technologies that are shaping the future.

• Design and build functional augmented reality prototypes.
• Apply interdisciplinary methods from computer science, psychology, design, and engineering.
• Engage in the full research process from question formulation to literature reading, building prototyping, data collection and analysis, and interpretation.
• Present original work at research symposia, conferences, or in publications.

• Design, develop, and refine AR prototypes aligned with project goals and milestones.
• Review and summarize relevant research literature to inform design and evaluation.
• Conduct user testing and capture observations and feedback.
• Submit biweekly reports documenting progress, challenges, lessons learned and planned next steps.
• Attend biweekly meetings with the faculty mentor and team to review progress and plan next steps.

Hybrid

Dr. Brooke Zhao, yzhao20@kennesaw.edu 

Injury is the leading cause of death among children ages 0鈥�3 in the U.S. (with more than 2,300 deaths and over 1.3 million non-fatal emergency department visits from unintentional injuries in 2023). This calls for an urgent need to address home hazards through effective parent-focused safety training programs. Current approaches, typically pamphlets and short videos, offer limited learning engagement and effectiveness. By contrast, embodied experiential learning through immersive virtual reality (IVR) and real world scenario simulations can significantly enhance both engagement and training outcomes.

During the 2024鈥�2025 academic year, our faculty and student team successfully developed a proof-of-concept IVR training prototype. Based on feedback from our user studies, we now aim to expand the prototype with advanced features designed to improve user experience and training effectiveness.

The goal of the proposed project is to develop a Phase 2 prototype, ParentSHIELD-MR-AI, by implementing the following features and evaluating their effectiveness:

1. Designing a semi-realistic 3D interior environment for the VR mode of the prototype.
2. Integrating machine learning (ML) algorithms, such as computer vision, and large language models (LLMs) APIs into a mixed reality (MR) environment for hazardous object recognition, detection, and intelligent AI tutoring.
3. Creating intuitive 3D user interfaces (3DUIs) to support diverse user tasks during training.
4. Conducting a pilot user study to evaluate the system鈥檚 design and effectiveness.

1. Research capability of doing literature review, user study data collection and analysis.
2. Software development skills of creating immersive interactive experiences in Unity for educational and training purposes
3. Interactive design skills in 3D user interfaces.
4. Game level design skills to create 3D virtual environments.

Students will work on assigned project prototype development work and give weekly reports of the progress during co-PI meetings. The co-PIs will share supervision responsibilities equally,

Hybrid

Dr. Allison Garefino, agarefin@kennesaw.edu 

Dr. Lei Zhang, lzhang24@kennesaw.edu 

Dr. Melissa Osborne, mcowart3@kennesaw.edu