Tactile sensing and stiffness control with multifingered hands

In this paper we analyze and experimentally measure the benefits of tactile sensing for stiffness control in the presence of uncertainty. There are two parts of the manipulation process where tactile information is helpful: determining the initial object pose as the fingers close to grasp, and tracking the object as it rolls and slides against the fingers during manipulation. The paper begins with an analysis of uncertainty associated with the initial orientation of a grasped object. We then review the object stiffness control algorithm (Salisbury, 1985), and derive an expression for the errors in the commanded stiffness based on object orientation uncertainty. Experimental execution of a peg insertion task demonstrates that tactile sensing can effectively determine the initial orientation of the object. Tactile sensing can also be used to track object rolling, but update rates are slow which can limit the maximum stiffness.