Chapter 3
Visual grasp algorithm
A new method for a fast visual grasp of unknown objects, using a camera mounted
on a robot in an eye-in-hand configuration, it is presented in this chapter. This
method is composed of an iterative object's surface reconstruction algorithm and
of a local grasp planner evolving in a synchronized parallel way.
The reconstruction algorithm makes use of images taken by a camera carried
by a robot. First, a rough estimation of the object position and dimensions is
performed through the use of a new preshaping technique, then an elastic recon-
struction ellipsoidal surface is virtually placed around the object and the fingers
of the robotic hand are suitably placed on it. Afterwards, the reconstruction sur-
face iteratively shrinks towards the object's center of mass until some parts of
the surface intercept the object's visual hull, dragging out the fingers attached on
it. Between two steps of the reconstruction process, the grasp planner moves the
fingers according to some grasp quality measures, floating on the current recon-
struction surface at an imposed (security) distance.
The fingers keep moving until a local optimum with respect to the chosen grasp
quality measure has been reached, and then a new estimation of the surface of the
object provided by the reconstruction process is considered. Moreover, a controller
should ensure that the motion of the fingers follows the desired trajectories guided
by the indices for the grasp. The whole process ends when the object is completely
reconstructed and the grasp planner performs the last evaluation of the grasp
quality measure, finding the final configuration of the hand and performing the
execution of the grasp.
Quality measures considering both hand and grasp proprieties are adopted:
the directions of the finger motion leading toward grasp configurations which are
not physically reachable, or causing collisions or loss of hand's manipulability, are
discarded. A new heuristic grasp quality method has been here proposed, but it
is worth noticing that many other quality measures may be chosen in substitution
of the proposed one, without effecting the general framework.
23
Summary :
Chapter 3 Visual grasp algorithm A new method for a fast visual grasp of unknown objects, using a camera mounted on a robot in an eye-in-hand configuration, it is presented in this chapter. The whole process ends when the object is completely reconstructed and the grasp planner performs the last evaluation of the grasp quality measure, finding the final configuration of the hand and performing the execution of the grasp.
Tags :
quality,fingers,surface,reconstruction,objects,new,object,hand,isual,measures,planner,process,algorithm