Abstract:
Our shared dream is to have robot humanoids with hands complete similar tasks that humans do. While there are a few robot hands available today, the popular opinion is that they are difficult to use, expensive, and hard to obtain which precludes their ubiquitous usage. We argue that this is not an inherent problem of robot hands, but rather that these robot hands are not built using the right design principles. In this thesis, we introduce a new class of robot hands that are significantly more dexterous, lower cost, and easier to use than prevailing robot hands. They are fully open-sourced and are continuing to serve as a significant entry point for many people in the dexterous manipulation research community.

Furthermore, how can robot hands imitate the human brain’s ability to complete dexterous tasks in a human-like fashion? While most robots used today have fewer than 10 degrees of freedom, a humanoid with two hands has over 50 degrees of freedom with many points of contact with the environment. This high dimensionality makes data-efficient learning extremely difficult. To rectify this, we leverage internet-scale human experience from the web as training data. Because robot hands have a similar morphology as human hands, we can directly learn from retargeted human motion to teach robots with significantly more data. In this thesis, we find that this begins to unlock the generalizable, human-like behavior we seek.

Committee:
Prof. Deepak Pathak (advisor)
Prof. Nancy Pollard
Prof. David Held
Murtaza Dalal

—