Dexterous Manipulation

Abstract

The main reason for using a multifingered articulated hand as end-effector of a robotic arm is to endow the system with the ability of manipulating a grasped object with dexterity (Ma and Dollar 2011). While the motion of the robot arm is used to move the robotic hand in the workspace, the fine motion of the fingers can reposition and reorient the object within the hand while guaranteeing a stable grasp of the object. By “within the hand,” we mean with respect to the palm where the fingers of the hand are attached. For a successful manipulative operation, the robotic grasp has to resist external forces, i.e., be force-closure, and to allow dexterous manipulation, i.e., be configured in such a way that the fingers can accommodate arbitrary in-hand object motions (Murray et al. 1994). In this essay, we will refer to multifingered robotic hands. A similar reasoning, however, applies when considering several cooperative manipulators (Caccavale and Uchiyama 2016). As underlined by Okamura et al. (2000), to properly formulate the dexterous manipulation (DM) problem, an object-centered point of view must be adopted. Given an object, grasped by a robotic hand, the aim is to move the object from a pose A to a pose B with respect to the palm, by imposing a suitable motion to the hand fingers, as shown in Fig. 1a. To deal with the DM problem, a suitable mathematical model of the grasp needs to be defined, and some common assumptions are made to allow a precise analysis of the situation (Li et al. 1989a; Murray et al. 1994): (i) the object is a rigid body in contact with a rigid link robot, (ii) accurate models of the fingers and object are given, and (iii) the object is grasped by the distal phalanges of the fingers. The first step to compute the required fingertip forces from a desired force/torque wrench on the object is to consider the grasp static equilibrium