|
Robotic Manipulation Contact: Frank van der Stappen
Description
Computer-supported design of industrial part handling processes can lead to a
substantial cost reduction in automated manufacturing. We study efficient algorithms
for the design of low-cost and robust solutions to common part handling tasks such as
part feeding and singulation (in which parts are oriented and separated, respectively),
grasping or fixturing (in which parts are immobilized by fingers or fixturing elements),
and assembly and disassembly.
Research of recent years shows that sequences of simple physical actions by simple manipulators using little or no sensory input can often accomplish the same manipulation tasks as the anthropomorphic robot hands and arms that are normally encountered in factories. As an example, a simple gripper consisting of two parallel jaws is able to orient any prismatic part by a sequence of squeeze actions. The sequence is computable from the shape of the part and does not rely on sensory data. Manipulation systems based on simple hardware elements bear a promise of increased reliability, speed, and suitability for automated planning, much lower cost, and rapid reconfigurability. Algorithms for the automated design of these reliable low-cost manipulation systems rely on a thorough understanding of the behavior of parts subjected to seemingly simple physical actions such as pushing, dropping and touching parts. Our focus is to analyze the relation between physical actions and their impact on a part and to use this knowledge to derive efficient algorithms for determining sequences of simple actions that will accomplish a higher-level manipulation task. We want our algorithms to be complete, meaning that they must report a solution if one exists. Geometry is a major parameter in the definition, modeling, and solution of manipulation problems. Interaction in a virtual environment often takes place through manipulation of the objects that are present in the environment, either by a user or by a computer-controlled creature. Insight in manipulation of objects will not only aid the automated design of manipulation systems; it will eventually also lead to more natural behavior of computer-controlled virtual humans in virtual environments. A better understanding of the effects of physical actions will result in a better match between the actions performed and the higher-level task these action aim to accomplish. Manipulation plays a crucial role is surgery simulation. The aim is to provide surgeons with virtual patients and surgical tools, and allow them to use those, rather than real surgical tools on real patients, to rehearse a difficult surgical intervention. At the core of such a simulation is a system that computes deformations as well as the changes to the object caused by the user's cutting actions. A separate page is devoted to our work on cutting in deformable objects. The Center for Geometry, Imaging and Virtual Environments closely collaborates with the group of Professor Ken Goldberg at the Department of Industrial Engineering and Operations Research at the University of California at Berkeley.
Current projects
Immobilization of non-rigid structures
Immobilization is a fundamental concept to manufacturing processes. It refers to
preventing any motion of a part, despite the possible application of external
forces and torques. Immobilization is of importance to grasping by
multi-fingered robot hands,
and crucial to fixturing, where a part is held by fixturing elements to be subjected
to for example a machining operation. Immobilization of rigid objects is a
fairly well-studied problem, but hardly any results exist for non-rigid
objects, such as assemblies, deformable objects, or mechanisms.
CLAMP (Complete Algorithms for Manipulation Planning)
Most results for manipulation systems based on simple hardware elements
apply to perfectly shaped planar polygonal parts. This NWO-funded project
takes up the
challenge to tackle this idealization as real industrial parts are
three-dimensional, often curved, and manufactured to tolerances. It
studies these issues in the context of two important manipulation
tasks: part holding - which is concerned with (partially or completely)
constraining the motion of part - and part feeding - which is concerned
with orienting parts.
Past projects
Fixture planning: geometry and algorithms
In this project we studied efficient complete algorithms for
computing grasps involving point and line contacts, and modular fixtures
consisting with locators, clamps, and so-called edge fixels, whose
placements are constrained by an underlying regular grid of holes.
The results have been reported in several papers and
in the PhD thesis of Chantal Wentink.
People
The following people are currently involved in this research area:
Publications
Facilities
For implementation and testing we have a
lab
available with modern top-end PC workstations with stereo
viewing facilities. This lab was partially donated by Microsoft.
Student projects
There are ample opportunities for students to do master projects in this area. The list above gives a
number of research themes. The projects can both be theoretically oriented and more experimental in nature.
The lab is available for the students to work in. If you are interested, contact
Frank van der Stappen.
For more information see the
education page.
webmaster: Frank van der Stappen
|
