Date of Award

1-2026

Document Type

Thesis

Degree Name

Master of Science (MS)

College/School

School of Computing

Department/Program

Computer Science

Thesis Sponsor/Dissertation Chair/Project Chair

Rui Li

Committee Member

Jiayin Wang

Committee Member

Michelle Zhu

Abstract

Perceiving another person’s emotional state is vital for effective human collaborations, allowing a person to adjust their behavior and emotional state to build trust and empathetic connections among those involved in the collaboration. However, in human-robot collaboration, robots usually lack the emotional intelligence humans have, instead being designed for specified applications. In a manufacturing setting, collaborative robots are programmed with repetitive movements, coordinated to improve overall productivity and efficiency in assembly tasks, but these preprogrammed actions may not align with the human collaborator’s expectations, or leave them uncomfortable or not safe, deterring future collaboration. To enhance human-robot collaborations to better adapt to the collaborative scene and to integrate a human’s emotional state in the decision-making process, this thesis studies friendly human-robot collaboration and develops an emotion-based behavior planning system to realize it. The developed system grants an integrated collaborative robot with emotion recognition and reaction capabilities to create a vivid and friendly human-robot collaboration process. It does so by utilizing a multimodal emotion recognition method based on visual-audio-physiological data in order to respond with artificial emotional expressions via a 3D simulated digital head model as well as change how it works alongside a human collaborator with an employed hierarchical reinforcement learning (HRL) approach to training the robot. Simultaneously, the system listens for vocalized requests for assembly parts and determines the optimal path to retrieve them using inverse kinematics. Human-robot co-assembly tasks are used as experimental scenarios throughout development for validating and evaluating our developed system.

File Format

PDF

Available for download on Sunday, February 20, 2028

Share

COinS