Information Technology Reference
In-Depth Information
worthwhile to note that such a reconstruction does not need to be exact, since a
scaled one is sufficient [30].
Finally, the robot's proportional control laws
u
=(
ν
,
ω
) are obtained using Lya-
punov theory. For each image trajectory component a different Lyapunov function
is chosen in order to minimize the feature tracking error.
18.4.2
Experimental Results
Experimental results for the IBVS and optimal path framework are now presented.
The experimental setup comprises of a Quickcam Ultravision camera, whose res-
olution is 320x240 pixels, mounted over the front-part of a K-team Koala vehicle.
The ERSP vision library ([17]) is used to perform SIFT recognition. The controller
bandwidth is almost 7 Hz.
In the first experiment, the optimal path
SL
reported in Figure 18.9 is accom-
plished. The path is composed of the following maneuvers: a counter-clockwise
rotation on the spot (the image conic
I
x
i
→
I
x
1
), a forward motion (the straight line
I
x
1
→
I
x
2
), and a clockwise rotation on the spot (the image conic
I
x
2
→
I
x
d
). Figure
18.12 reports the related experimental result.
In the second experiment, the optimal path
SL
T
1
P
reported in Figure 18.10
is accomplished. The path is composed of the following sequence of maneuvers: a
counter-clockwise rotation on the spot (the image conic
I
x
i
→
−
I
x
1
), a forward motion
(the straight line
I
x
1
→
I
x
b
), a logarithmic spiral (the straight line
I
y
2
→
I
y
3
)anda
clockwise rotation on the spot (the image conic
I
x
b
→
I
x
d
). Figure 18.13 reports the
related experimental result.
Fig. 18.12
Experiment 1. Planned paths for all features and the trajectory of the tracked
feature (up left). Initial (bottom left), final (bottom right) and desired (up right) images taken
from the vehicle. The planned paths and also the actual position of the features are plotted
over the initial and final images
Search WWH ::
Custom Search