Major
Computer Science
Anticipated Graduation Year
2027
Access Type
Open Access
Abstract
General-purpose robotic manipulators will be maximally useful if they can be trained on existing data, such as pre-collected human footage. Recent advances in robot learning typically rely on visuals to mimic the high-fidelity movements of humans. Our goal is to model large datasets of human actions as sequences of goal-seeking behaviors, driving robotic motion by leveraging this understanding of human intent. This presentation will discuss the current state of robotic imitation learning, the advancements we have made over the past three months, and our future goals.
Community Partners
N/A
Faculty Mentors & Instructors
George K. Thiruvathukal Ph.D., Department Chair, Computer Science; Matt Hyatt, Ph.D. Student, Computer Science
Supported By
N/A
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License.
Towards a Unified Architecture for Jointly Learning Human and Robot Behavior
General-purpose robotic manipulators will be maximally useful if they can be trained on existing data, such as pre-collected human footage. Recent advances in robot learning typically rely on visuals to mimic the high-fidelity movements of humans. Our goal is to model large datasets of human actions as sequences of goal-seeking behaviors, driving robotic motion by leveraging this understanding of human intent. This presentation will discuss the current state of robotic imitation learning, the advancements we have made over the past three months, and our future goals.