Abstract
Robot learning from demonstrations is critical in industrial and service robotics applications, where robots must acquire diverse skills and adapt to human workflows in complex, unstructured environments without relying on exhaustive manual programming. Although robots have been widely deployed in industrial settings, acquiring new skills and adapting to novel tasks remain significant challenges due to (1) the large volume of demonstration data typically required, (2) the limited reusability of previously learned knowledge, and (3) the high cost of collecting high-quality demonstrations. This dissertation aims to address key challenges in robot learning from demonstrations in complex and unstructured environments by making robot skill acquisition more effective, efficient, reusable, and adaptable through human–robot interaction. The first topic in this dissertation is the development of a human–robot interaction platform that enables XR-based robot teleoperation, providing an intuitive and efficient interface for robot control and demonstration collection. The second focal point is efficient robot learning from demonstrations, where the reusability of action models is improved through the decomposition of complex manipulation skills into modular components and their recombination for generating new actions. The third core area investigates how robots can effectively acquire skills from accessible human demonstrations, which is achieved through an egocentric video-based framework that extracts human hand motions from one-shot first-person demonstrations to support robot learning. The fourth topic focuses on tactile-enhanced robot manipulation, where robot tactile data is utilized to enhance physical-world data collection and improve the robustness and precision of contact-rich manipulation.
Publication Date
5-2026
Document Type
Dissertation
Student Type
Graduate
Degree Name
Computing and Information Sciences (Ph.D.)
Department, Program, or Center
Computing and Information Sciences Ph.D, Department of
College
Golisano College of Computing and Information Sciences
Advisor
Rui Li
Advisor/Committee Member
Denis Cormier
Advisor/Committee Member
Yunbo Zhang
Recommended Citation
Dengxiong, Xiwen, "Towards Effective and Efficient Robot Learning from Demonstrations via Human-Robot Interaction" (2026). Thesis. Rochester Institute of Technology. Accessed from
https://repository.rit.edu/theses/12691
Campus
RIT – Main Campus
