Commanding robots through natural language enables a natural mode of interaction for end users.  Natural language is widely understood and can be used to describe a broad range of subjects and tasks.  Our system — dubbed RoboFrameNet — allows end-users to interact with robots through natural language and enables developers to integrate their work with the same.  In RoboFrameNet, natural language is grounded into actions through an intermediary called semantic frames, which describe scenes being played out, such as an navigating, picking up an object, or following a person.

Publications and more information:

• RoboFrameNet’s ROS.org project page
• ICRA 2012 Submission: Provided upon request.
• RSS 2011 Workshop on Grounding Human-Robot Dialog for Spatial Tasks: Paper, Poster
• Research proposal (April 2011): Paper