This project will develop a better understanding of how people express their intent in dexterous tasks, such as opening a jar of peanut butter, scooping some, and spreading it on a slice of bread, model their preferences for how task actions should be performed, and create multimodal interfaces that map high- and low-level task instructions to dexterous task primitives.

Using a multimodal interface, a user instructs a dexterous robot system to perform a desired task based on the user’s preferences