Title: Improving Learning from Demonstration in Real-World Scenarios

 

Date: Monday, May 22nd, 2023

Time: 1:00pm EST

Location: Kendada Building, Room 210

Zoom: https://gatech.zoom.us/j/97821389293

 

Erin Botti

Robotics PhD Student

School of Interactive Computing

Georgia Institute of Technology

 

Committee

Dr. Matthew Gombolay (Advisor) – School of Interactive Computing, Georgia Institute of Technology

Dr. Sonia Chernova – School of Interactive Computing, Georgia Institute of Technology

Dr. Charlie Kemp – Biomedical Engineering Department, Georgia Institute of Technology

Dr. Agata Rozga – School of Interactive Computing, Georgia Institute of Technology

Dr. Maya Cakmak – Computer Science & Engineering Department, University of Washington

 

Abstract

To realize a vision of assistive robots in the home (e.g., care robots for older adults), robots will need the ability to learn new skills and adapt to their users and environment. Everyone's home has a different layout, everyone has differing needs, and people's preferences and needs will change over time as they age. One solution to this challenge is Learning from Demonstration (LfD), a paradigm that enables novice end-users to teach robots new skills based on human demonstrations of a task. However, LfD is not foolproof.  The real-world presents challenges from suboptimal teachers, communication errors, and unstructured environments. I aim to develop improvements to LfD to account for robot failure, teacher suboptimality, and target end-users. First, I conduct a human-subjects experiment evaluating a person's perceptions of robot failure when the person is the one teaching the robot via LfD. Since robot failure can occur due to demonstrator suboptimality, I then develop a novel architecture to learn from suboptimal demonstrators' feedback.  I propose a framework for robot failure recovery, where the robot learns the person's preferences through human interventions. Lastly, I propose to evaluate LfD with a target population of older adults.