Robotics Institute
Seminar, April 28, 2006
Time and Place | Seminar Abstract | Speaker Biography | Speaker Appointments
Making the Most of Minimalism in
Distributed Manipulation
Mauldin
Auditorium (NSH 1305)
Refreshments 3:15 pm
Talk 3:30 pm
Abstract |
Distributed manipulation involves
manipulating objects through many points of contact, or even a continuum of
contact. Typical work in this field
involves planar manipulation of an object using distributed traction forces. This provides great flexibility and power in
manipulation for both single and multiple objects, but poses two problems: 1)
How to produce the large number of forces required, and 2) How to decide what
to do with the huge number of input degrees of freedom. Much work has been done towards answering
each of these questions, although typically in a way which requires a huge
number of independently controlled actuators.
This talk focuses both on methods of producing planar force fields which
are useful for manipulation and simple to implement in hardware. The first part describes the use of planar
air flow fields for object manipulation which relies on the natural shape of
flow fields from a small number of flow generators. While the shapes of the resulting fields must
respect the flow dynamics, a class of useful fields for manipulation can be
generated, albeit with multiple equilibria.
Sequences of these fields can bring planar objects to an equilibrium
pose without sensors. The second part of
the talk describes the class of quadratic potential fields, with a host of
useful properties such as reduced number of equilibria, independence of object
shape, and ease of prediction of equilibrium pose under operations such as
superposition of fields. Until recently,
however, there was no method, short of an array of independently controlled
actuators, to implement these fields. By
examining the divergence properties of quadratic fields, and of the type of
planar air flow fields generated in the first part of this talk, a method of
producing quadratic fields with simple shaped regions of airflow is
developed. These methods are extended to
applications involving the superposition of manipulation fields for more
flexible manipulation including manipulating along trajectories. The third part
of this talk examines methods of using electric fields through delectrophoretic
forces to produce similar fields both in 2-D and 3-D on a near-micro scale.
Speaker Biography |
Jonathan Luntz obtained his B.S. in Mechanical Engineering
from the State University of New York at
Speaker Appointments |
For
appointments, please contact Peggy Martin (pm1e@andrew.cmu.edu)
The Robotics Institute is part of the School of Computer Science, Carnegie Mellon University.