Visual Integration of Tactile Information in Telerobotics
Roger A. Browse and Marcia L. McDonald
Abstract
We have carried out three experiments which explore the effectiveness of various visual
depictions of tactile data for the task in which subjects view a robot gripping an object and
attempt to predict if the robot will be able to pick the object up. The display types tested range
from simple total closing force values through to graphic three-dimensional contour plots,
overlaid with force-torque vectors, with the subject having dynamic viewpoint control over the
display.
Our results suggest that an operator's utilization of such tactile displays depends on the
effectiveness of the model of the gripping configuration which can be obtained through visual
images of the scene. In the case of simple planar contact between gripper and object, a total force
value produced maximum performance enhancement. In more complicated, or awkward non-planar gripping,
the total force value was of no use, but spatially distributed force representations
could still yield enhanced performance.
Further experimentation was carried out with complex displays which subjects reported as
depicting the contact forces better. It was found that these displays actually reduced performance,
drawing into question the telepresence assumption that more realistic information will lead to
better telerobotic operator performance.