Warning: This pages are still under construction.
The main way the human being modifies the environment -besides mere presence- is through manipulation. General agreement exists over the evidence that the hands and their dexterous use is one of the basic features that characterize the human being. Manipulation does just mean this, to act on something with the hands. If robots are intended to be skillful helpers of humankind, it is obvious that they have to be enabled with similar manipulation capabilities. Manipulation ranges from mere picking up and placing a wooden block to sophisticated handling of a tie to make a Windsor knot.
In any case, some planning is involved: the sequence of available elemental actions, as well as the unambiguous instantiation of their physical parameters, is far from being arbitrary. Success can only be achieved for specific orderings of such actions, and conditions are even more restrictive if some kind of optimality is pursued. Such a plan can be just inherited (hardwired or precoded), acquired by learning, or retrieved by the own robot if the means are provided to represent different action sequences and to evaluate their potential success. The necessary descriptional and procedural models for planning itself can be precoded or learnt, where both ways are not mutually exclusive.
Elemental actions: different kinds of grasping, whole objects or a part of it, other kinds of non-prehensile manipulation (pushing, pinching, hitting, pressing...)
Objects to manipulate appear in overwhelming diversity. Just look around and consider all the objects you manipulate along a single day. Think about the objects you have known along your life and try to imagine those which are still to come. Never wondered about what that extravagant tool exposed in the hardware store was for and how it was to be handled? Or if that exotic fruit should be peeled and how to perform this operation?
Short Survey on Manipulation Planning for Rigid Objects
Last modified on wednesday, 12 March, 2008