Page 169 - Human Inspired Dexterity in Robotic Manipulation

A Grasping and Manipulation Scheme 167

system [6, 7]. The advantage of this method is that only proprioceptive sen-
sory information, such as information from encoders and rough location
information, is required for stable-object grasping. For the manipulation
of the grasped object, this method does not require precise information
about the object, such as its geometrical shape. Object manipulation is real-
ized using only information on the position and orientation of the object.
This kind of information can be acquired using markers or feature tracking
systems. Following the original proposal, this method has been extended to a
visual servoing method for object manipulation that is robust against tem-
porary loss of visual information when an occlusion occurs, or the object
otherwise goes out of sight [8]. However, the time-delay problem has
not been extensively discussed in previous research. In this chapter, a
new visual servoing method for object manipulation by means of a multi-
fingered hand-arm system is proposed. The new method is robust, not only
to the temporary loss of visual information, but also to time delays in visual
information. This method utilizes a virtual-object frame that is defined in
terms of the position and orientation of each fingertip, which are obtained
by proprioceptive sensors, similar to the previous method [8]. The differ-
ence between the new method and previous one is that in the new method,
the information regarding the measured-object frame that is obtained by the
visual sensor is not directly used in the feedback controller, but rather is used
in the design of a desired virtual-object frame. In other words, the new
controller is designed to manipulate the position and orientation of the
virtual-object frame toward their desired values as designed based on the
information of the measured-object frame.

9.2 MULTIFINGERED HAND-ARM SYSTEM GRASPING
AN OBJECT

The hand-arm system that we consider here is composed of one arm and
one multifingered hand. As a case study, we have focused on a hand-arm
system with a three-fingered hand component, in which the arm part has
five DOFs and the first finger has five DOFs and the other two fingers have
four DOFs. Namely, the hand-arm possesses a total of 19 DOFs as shown in
Fig. 9.1. Note that our control framework is not limited to apply for this
hand-arm system, it can apply for other hand-arm systems which have a
sufficient number of fingers and DOFs to achieve stable-object grasping
and stable manipulation of the grasped object. The symbols N a and N i
denote the numbers of DOFs of the arm and the ith finger, respectively,