The Gambit robot.
Click for video!
My expected graduation is June 2014.
My primary research interests are in robotics and natural language processing; in my work, I combine these interests to support research in human-robot interaction, or HRI. My work focuses on the problem of grounded language acquisition: extracting semantically meaningful representations of human language by mapping those representations to the noisy, unpredictable physical world in which robots operate.
More specifically, I work on combining probabilistic, grammar-based natural language processing with machine learning to transform human communication into a formal language that a robot can understand. I have looked at using this kind of language learning to learn how to follow navigation instructions (ISER 2012) or learn more about the world from human users by learning to extend a world model in tandem with learning a language parsing model (ICML 2012).
My research is conducted largely on the Gambit manipulator arm, which we developed in a collaboration among UW, Intel, and Alium Labs. Its first project was learning to playing chess robustly, without the common laboratory constraints on board placement, piece color and placement, lighting, and so on (ICRA 2011). Now, we use it in research on how naive users can control a robot using language and other input modes.