The Developmental Organization of Dexterous Robot Behavior

Abstract

Hands are levers of influence on the world that made intelligence worthhaving. Precision hands and precision intelligence coevolved in the human lineage, and the fossil record shows that hands led the way.

Steven Pinker, "How the Mind Works"

"Dexterity" is the bridge between a physical and an intellectual relationship to the world. Humans have evolved to be bipedal with hands free for manipulation and defense. This evolutionary choice has had a tremendous impact on the co-evolving human brain as well as the perceptual categories, motor synergies, control knowledge, and representations that support interactions with the world.

For robots to exploit the same potential that every infant enjoys, new challenges must be addressed for which there are no adequate computational theories. However, we may take some inspiration from the maturational structure demonstrated by newborn humans. Such developmental processes construct increasingly complex control strategies from a sequence of much simpler learning tasks. This talk is about computational accounts of these processes that can be used to program complex robot systems---in particular, dexterous robot manipulation systems. These technologies have a huge commercial potential and they promise to teach us something about our own brains as well. Examples will be drawn from experimental platforms at UMass including Dexter, a bimanual manipulation robot, and the UMass uBot-4, a new mobile manipulator concept.