Information Technology Reference
In-Depth Information
Target object
Image space
s
d
s
i
Projection
Camera space
Image plane
γ
i
γ
d
Camera frame
Local tracking
Joint space
Robot arm
q
d
q
i
Fig. 11.1
Our proposed planner [32] alternatively explores the camera space and joint space
for feasible robot/camera paths to obtain feasible feature trajectories in image space
to transform camera paths to waypoints on the image feature trajectories. We also
assume that obstacles in the workspace are known
apriori
.
To be successful, global path-planning approaches require a complete and (rela-
tively) precise knowledge of the environment, camera calibration, and object model.
These requirements can be limiting in many visual servoing scenarios. The need for
such exact knowledge can be relaxed by accounting for modeling and calibration
uncertainties at planning stage. In the following section we discuss the effects of
uncertainties in visual servoing, especially in tracking planned trajectories at the ex-
ecution time, and report on a few recent works on path-planning under uncertainties
for visual servoing.
11.4
Path-planning under Uncertainty for Visual Servoing
Planned paths need to be executed and the robot may not exactly follow the planned
path due to the uncertainties and in fact, in some cases, the followed paths in the
workspace and in the image space can be quite different from the planned ones
thereby resulting in violation of some of the constraints even if they have been ful-
filled at the planning stage. The influence of errors in intrinsic and extrinsic camera
parameters on the performance of visual servoing scheme has been examined in
[20]. In [37] the propagation of image error through pose estimation and visual ser-
voing control law has been analyzed.
A number of researchers have proposed local and real-time techniques to ac-
count for likely deviations from the planned path at the execution (tracking) stage.
For example, these could be locally taken care of using potential field type tech-
niques, however as mentioned before, such techniques are prone to failure due to
Search WWH ::
Custom Search