2024-2027 Pre-Doctoral Fellows

Crowd analysis is a field that seeks to understand the behavior and movement patterns of crowds, which is crucial to innovating new techniques to mitigate risks from movement activities of high density population areas. The aim of this research is to develop and train machine learning models to predict and minimize potential crowd dynamics to reduce the risk of injury to individuals within crowds from video streaming data, with interdisciplinary approaches to modeling crowd movement with physics based computational crowd dynamics simulations and leveraging human factors from behavioral psychology. The technical foundation for the research is accomplished with a machine learning methodology that incorporates simulated data and video data to develop new physical laws to be encoded in a crowd model. Experimentation with this model aims to provide new insights on the behaviors of pedestrian crowds, which can be utilized to design safer spaces for urban planning and event coordination.

Faculty Mentors:

Daniel Barbará, Professor, Department of Computer Science, College of Computing and Engineering

Rainald Löhner, Professor, Department of Physics and Astronomy, College of Science

Gerald Matthews, Professor, Department of Psychology, College of Humanities and Social Sciences

Chronic pain affects 20% of adults and is a leading cause of disability, significantly impacting healthcare costs and workforce productivity. This project addresses the critical need for improved diagnostic tools and treatment strategies for chronic pain by integrating patient narratives with Natural Language Processing (NLP) and generative Artificial Intelligence (AI). It comprises three tasks: developing a Motivational Interviewer system that leverages Large Language Models (LLMs) to embody MI techniques within a structured conversational framework.  Secondly, creating a Patient Simulator that generates realistic patient interactions by combining linguistic, medical, and behavioral profiles to engage in synthetic MI sessions for comprehensive testing. Lastly, integrating and refining the Motivational Interviewer and Patient Simulator through Reinforcement Learning from Human Feedback (RLHF) and AI Feedback (RLAIF) ensures effective and empathetic therapeutic dialogues. This project democratizes access to advanced pain management techniques, promoting equitable healthcare. By merging clinical practice, AI, and psychology, we advance both AI technologies and patient outcomes, addressing health disparities and improving engagement in chronic pain management.

Faculty Mentors:

Kevin Lybarger, Assistant Professor, Department of IST College of Engineering and Computing

Siddhartha Sikdar, Professor Department of Bioengineering College of Engineering and Computing

Samuel Acuña, Assistant Professor, Department of Bioengineering College of Engineering and Computing

Lynn Gerber, Director for Research Medicine Service Line, Inova Health System

The increasing demand for assistive technologies underscores the importance of advanced service robots in households. Existing mobile robots face significant challenges in these settings. Due to their inability to predict the long-term side effects of their immediate action, these robots execute tasks that seem optimal in the short term but are costly for subsequent tasks, leading to inefficient performance.

Our research addresses these issues by developing a mobile robot with anticipatory planning capabilities that enable robots to prepare for future tasks. By integrating task and motion planning with learning, we aim to achieve scalability and efficiency. Incorporating Large Language Models (LLMs) will enhance human-robot interaction, allowing robots to adapt to user instructions and preferences. These refinements will create more autonomous and efficient service robots exhibiting common-sense-like behaviors, improving the daily lives of individuals requiring assistance and pushing the boundaries of domestic robotics toward more proactive and self-sufficient technologies.

Faculty Mentors:

Gregory Stein, Department of Computer Science, College of Engineering and Computing

Xuesu Xiao, Department of Computer Science, College of Engineering and Computing

Ziyu Yao, Department of Computer Science, College of Engineering and Computing

Elizabeth Phillips, Department of Psychology, College of Humanities and Social Sciences