manipulation tasks performed by an anthropomorphic robotic hand
equipped with soft fingertips
become more accurate at modeling the real world, the number of possible
applications for such systems increases. One example is the dynamic
simulation of grasping and manipulation tasks, enabling interaction
with virtual objects. However, accurate grasping simulation requires
modeling subtle physical phenomena not taken into account by
traditional multi-rigid-body dynamic systems. Here we present examples
of simulated manipulation tasks using an analytical model for soft
fingertip contacts. This allows fingertips to apply frictional torque,
a key factor in the ability to create stable grasps using only two
fingers. Our approach also captures subtle phenomena, such as the
effect of local object geometry on the space of frictional forces that
can be applied at the contact, as well the coupling between tangential
force and frictional torque.
Simulation is performed at interactive rates, allowing the user to send force control commands to the motors controlling the hand and also change camera position during the simulation. These movies run at approximately twice the speed of the actual simulation.
pinch grasp example
In this example, the hand successfully picks up a rectangular object and holds it in a vertical position. This is possible because the frictional forces applied at the contacts are able to counter the effects of gravity. Since the grasped object is locally planar, the fingertips create large contact areas which helps increase the magnitude of frictional torque supported at the contacts.
Complete movie ( 1.5 Mb mpeg4 avi )
pinch grasp example
The cylindrical object used in this case has similar material properties (friction coefficient, mass, inertia tensor) with the block used in the previous example, and the fingertips are applying identical normal forces. However, local curvature at the contact locations leads to smaller contact areas with the fingers, which in turn limits the magnitude of frictional torque. The fingers are unable to counter the effect of gravity, and the cylinder rotates around the axis of the contacts, eventually falling into the robotic palm.
Complete movie ( 1.8 Mb mpeg4 avi )
|The goal of the simulated task
is to obtain a stable grasp of the glass-shaped object by using finger
contacts around its circumference as well as contacts on the palm.
However, in the initial position (top left image) the surface of the
table prevents the execution of such a grasp. We have simulated a
control algorithm that uses two fingers to pick up the object and
rotate it above the palm (top row). This is possible only in the
presence of frictional torque applied by soft finger contacts. With the
palm facing up (top right image), the force applied by the fingers is
decreased allowing the glass to rotate while maintaining contact with
the fingertips until it hits the rigid palm of the robot (bottom left
image). The ring and little finger can now be closed around the glass,
creating a stable grasp (bottom right image).
Complete movie ( 5.4 Mb mpeg4 avi)
if you can not see these movies in your Media Player, try downloading
the DivX 6 codec from www.divx.com