Date: Sept 29th, 2026
Time: 2 pm EST
Location: Klaus 1212
Virtual Meeting: https://gatech.zoom.us/j/6342692876?omn=93309114089
Nina Moorman
School of Interactive Computing
College of Computing
Georgia Institute of Technology
Committee
Dr. Matthew Gombolay - Advisor, School of Interactive Computing
Dr. Sonia Chernova - School of Interactive Computing
Dr. Agata Rozga - School of Interactive Computing
Dr. Mykel Kochenderfer - Stanford University, Department of Aeronautics and Astronautics
Dr. Nakul Gopalan - Arizona State University, School of Computing and Augmented Intelligence
Abstract
To successfully deploy assistive in-home robots in human populated environments, robots will need to be able to perform supplemental on-site learning to adapt to aging users and changing environments. It is intractable for a roboticist to be contracted each time the robot policy needs to be updated. Instead, the robot should be able to learn from the end user directly. The field of learning from demonstration (LfD) enables non-roboticist end users to teach a robot novel skills via demonstrations of the desired behavior, without necessitating programming knowledge. While LfD aims to democratize robot learning, barriers still exist. Novice users require and request guidance to be able to successfully teach a robot via LfD. This is particularly the case when determining how to improve upon a demonstration set whose resulting learned behavior is not as desired or expected. In my thesis, I aim to increase the accessibility of LfD for novice users by developing LfD systems that provide novel forms of guidance to the demonstrator.
]]>