OPTICAL SENSING FOR FLUID-FILLED ROBOT FINGERTIPS
Dr. Nicola Ferrier
Harvard University
3:30pm Monday April 17 in Room 1307 ERB (Engineering Research Bldg)
Over the last several years, a number of sophisticated robot
hands have been developed for laboratory use. However, while
such hands approximate the mechanical complexity of human hands,
their application in manipulation tasks remains in a primitive
stage. Unlike human hands, they rely on minutely programmed
task descriptions that are time-consuming to generate and
susceptible to unanticipated changes in the task or the
immediate environment. This is largely because they cannot use
tactile information to detect and respond to changes.
We are investigating the performance of a sensing system based
on reflecting light off fluid-supported membrane fingers. Fluid
supported membranes, while superficially more complex than rigid
fingers, are actually simpler because the effects that need to
be modeled are object independent and surface independent. Most
important, for this talk, we open up a new sensing modality: how
to put 30,000 sensors in one milliliter! This would be tougher
but there is one such sensor in wide use: the CCD. We have
designed a prototype fingertip which uses a CCD sensor inside a
fluid supported membrane. A dot pattern has been drawn on the
inside of the membrane. Objects in contact with the membrane
influence the reflectance pattern in an observable way: from
such observations, we can reconstruct the shape of the membrane
(and hence the shape of objects in contact with the finger).