Mon, Nov 20, 2017 @ 09:00 AM - 11:00 AM
Thomas Lord Department of Computer Science
Monday, November 20th, 9 a.m. to 11 a.m, RTH 406
PhD Candidate: David Inkyu Kim
Title: Learning affordances by interactive perception and manipulation
Robots can plan and accomplish various tasks in unknown environment by understanding underlying functionalities of objects around. These attributes are called affordances, describing action possibilities between robot and objects in the environment. Affordance is not an universal property due to its relative nature, therefore must be learned from experiences. Such learning would involve predicting affordances from perception, followed by interactive manipulation. Learned affordance models can be directly applied to robotic tasks as the model describes how to manipulate and what the consequence will be.
In the presentation, methods to learn affordances with interactive perception and manipulation will be introduced. For the developed affordance models, extensive experiments were performed to verify the models and its application to robotic tasks.
Gaurav S. Sukhatme
Satyandra K. Gupta
Audiences: Everyone Is Invited
Contact: Lizsl De Leon