Image 3D fingertip force
This research developed an external camera method for measuring fingertip forces
by imaging the fingernail and surrounding skin. A 3D model of the fingernail
surface and skin is obtained with a stereo camera and laser striping system.
Subsequent images from a single camera are registered to the 3D model by adding
fiducial markings to the fingernail. Calibration results with a force sensor show
that the measurement range depends on the region of the fingernail and skin. A
Bayesian inference mode is developed to predict fingertip force given
coloration changes.
(Publications: Haptics06,
ICRA06, TBME (in press)) Group Webpage
Observability indexes selection for robot calibration
This research relates 5 observability indexes for
robot calibration to the ”°alphabet optimalities”± from the
experimental design literature.
All observability indexes are proved to be equivalent when the
design is optimal after a perfect column scaling. It is shown that
when the goal is to minimize the variance of the parameters,
D-optimality is the best criterion. When the goal is to minimize
the uncertainty of the end-effector position, E-optimality is the
best criterion. It is proved that G-optimality is equivalent to
E-optimality for exact design.
(Publications: ICRA08-1, TOR (under review))
Active robot calibration
This research also developed a new updating algorithm to
reduce the complexity of computing an observability index for
kinematic calibration of robots. An active calibration algorithm
is developed to include an updating algorithm in the pose
selection process. Simulations on a 6-DOF PUMA robot with
27 unknown parameters shows that the proposed algorithm
performs more than 50,000 times better than exhaustive search
based on randomly generated designs.
(Publications: ICRA08-2)
Finger Force Direction Estimation with Computer Vision
This research develops a method of imaging the
coloration pattern in the fingernail and surrounding skin to
infer-fingertip force direction (4 major shear force directions
plus normal force) during planar contact. Nail images from 15
subjects were registered to reference images with RANSAC and
then warped to an atlas with elastic registration. Common linear
features corresponding to the force directions were obtained
using Linear Discriminant Analysis. Without any individual
calibration, the overall recognition accuracy on new images of
15 subjects was 90%. With individual training on distributions
of backwards shear and normal force directions, the overall
recognition accuracy on new images of 15 subjects was 94%.
The lowest imaging resolution without sacrificing classification
accuracy was found to be between 10-by-10 to 20-by-20.
(Publications: CVPR2007,
Haptics2007,
ICRA2007,
TRO (under revision))
