Page 201 - Biomimetics : Biologically Inspired Technologies

Bar-Cohen : Biomimetics: Biologically Inspired Technologies DK3163_c006 Final Proof page 187 21.9.2005 2:56am

Robotic Mechanisms 187

Figure 6.8 MFI at University of California, Berkeley. (Image courtesy of R. Fearing, University of California,
Berkeley.)
Many other bio-inspired morphologies of locomotion exist in robots, from fish robots that can
swim like manta rays, to lobster robots, and to robots that tumble like tumbleweeds. The reader is
encouraged to investigate further the astonishing diversity of robotic locomotion that is currently
flowering in the world.

6.2.5 Grasping and Manipulation

Many robots need the ability to grasp and handle objects. On a large scale, this may be done in
the manner of hands, claws, or teeth, while at a smaller scale the task may be accomplished in the
manner of proteins or the sticky feet of flies or geckos.
Numerous issues play into grasping and manipulation. First, the manipulator element must be
guided to the object to be handled, a task that can require machine vision and a high resolution
control system. Second, the mechanical system must be able to reach objects with sufficient
flexibility and dexterity. And when the manipulator handles the object, the manipulator must
contain enough sensors and the right control software to apply just the right amount of force in
the object.
Many robots around the world have accomplished grasping and manipulation tasks, and
some are even developing the ability to learn manipulation tasks by learning and imitation such
as Ripley at MIT, and the juggling robots of USC Georgia Tech and MIT (Atkeson and Schaal,
1997).
Shadow Robotics of the U.K. has developed a robotic hand with 24 degrees of freedom (DOF),
driven by McKibben air muscles. Most of the joints are driven by an opposing pair of muscles,
permitting variable compliance at the joint; however, some of the finger joints are driven by a single
muscle with return spring. The Shadow hand operates at about half speed of a human.
The NASA Robonaut, a humanoid torso for outer space applications, boasts a pair of dexterous
arms enabling dual-arm operations, and 14 DOF hands that interface directly with a wide range of
tools. Using a humanlike model for autonomous grasping, the Robonaut relies heavily on feedback
from its robotic fingers and palm. Numerous sensor technologies are employed including piezo-
films, capacitive pressure sensors, and Force Sensing Resistor (FSR) technology. To achieve the
fine resolution of feedback, tactile sensors for grasping, and the smart algorithms for interpretation
of the signals, are systemically dubbed tactile perception. Robonaut’s manipulation capabilities are
being automated, and can also be run by teleoperation to enable delicate space operations via distant
human presence.