Grasp Planning & Force Computation for Dextrous Object Manipulation with Multi-finger Robot Hands

This paper deals with the problems of grasp planning and force computation that occur when objects have to be manipulated with dextrous multi-finger robot hands. Assuming that the initial contact points at the object surface are pre-selected, the joint motions to perform a desired operation according to a given object trajectory are determined. This is done under the consideration of the effect of rolling and slipping of the fingertips. Another point is the computation of appropriate grasp forces (internal forces) according to given object forces/moments needed to ensure a stable and secure grip configuration from which the joint torques are derived. This leads to an optimization problem that can be solved with two different approaches based on search procedures for finding the maximum (best fitting value). As major result, the force optimization problem could be generalized for the case of arbitrary robot hands and contact situations that are specified by the following parameters: number of fingers fno, number of joints per finger jno i, chosen contact models cmodi and null space dimension ndim. Especially, the resulting object motions and contact forces are demonstrated at a simulated example of a peg-in-hole insertion task with the Karlsruhe Dextrous Hand.