Biomedical Engineering Reference
In-Depth Information
Fig. 4.7 For evaluation, we
mount the tracking system to
a Kuka robot, positioned
roughly 2 m from the Adept
robot, to move the tracking
system by fixed distances to
perform, and evaluate the
calibration accuracy
4.2.6.1 Data Acquisition for Evaluation of Marker Calibration
To estimate the translational error of the marker calibration (cf.
Sect. 4.2.2
), we
mount the marker to the third robot link onto five different positions. For each
marker position, we perform the marker calibration and estimate the distance of
the marker to the axis of rotation in joint 3. Therewith, we estimate the transla-
tional error of the marker calibration for the x- and y-axis, as explained in
Sect. 4.2.4
.
Furthermore, to check the stability of the marker calibration, we perform seven
different marker calibrations (marker at the same position at third link) and
compare the transforms found among one another, as introduced in
Sect. 4.2.5
.
Factors that might influence the stability are the used robot positions and noise in
the measurements.
For each marker calibration, we employ 500 randomly chosen robot positions
within a sphere of 200 mm radius with respect to the initial robot position. The
initial robot position is chosen such that the marker is directly facing the tracking
system and the required robot movements are inside the robot's workspace
(cf. Fig.
4.7
).
4.2.6.2 Data Acquisition for Evaluation of Online Calibration
For evaluation of the online calibration method, we systematically analyze its
accuracy compared to hand-eye calibration, as this is the currently applied cali-
bration method. First, we test the performance in a world calibration setup, in
which a ground truth as a reference is known. Second, we investigate the variance
of the online calibration when using different positions in the robot workspace.
This is a measure for showing the stability of the calibration during application as
the robot arm might operate in the full workspace. And third, we evaluate the
impact of the different calibration methods on the robotized TMS system's overall
Search WWH ::
Custom Search