Two-stage control via virtual desired values in soft-fingered manipulation with time delay

This article focuses on the modeling and control in soft-fingered grasping and manipulation. We observed how a pair of 1-DOF fingers, whose fingertips are similar to humans in that they have a soft and hemispherical pad with a buttressing hard back plate, grasped and manipulated a rigid object. Based on the observations, we propose a parallel distributed model that takes into account tangential deformation of the fingertips. Human can grasp and manipulate an object despite of a relatively large delay of several ten milliseconds in the nerve system. We thus propose a two-stage control law via virtual desired angles for soft-fingered grasping and manipulation, and show that the proposed control law works well despite of the delay in control loop.