Grasp planning for precision manipulation by multifingered robotic hand

Abstract

This paper describes a model-based grasp planner for an articulated multi-fingered robotic hand. The planner searches for an initial grasp from which a required continuous motion of an object is achieved using fingertips alone while keeping rolling contact with the object. An adaptive search strategy has been developed and implemented: the error analysis of the failed grasp is made to find the most promising search direction. A quality measure of manipulation is newly introduced and evaluated to select a better grasp among possible candidates. Simulation results are demonstrated.