Unicycle-like Robots with Eye-in-Hand Monocular Cameras: From PBVS towards IBVS - Visual Servoing via Advanced Numerical Methods - page 345

Fig. 18.13 Experiment 2. Planned paths and trajectory of the tracked feature (up left). Initial
(bottom left), final (bottom right) and desired (up right) images taken from the vehicle. The
initial and desired positions of the features are plotted over the initial and final images taken
from the vehicle. The actual positions of the features are also shown in these images

The method accomplished the task of driving the tracked feature through the
planned path. It can be observed from the figures that the final positions of the
features are almost coincident with the desired positions. The final feature position
error is due to the tracking noise and to the estimation of the angle

(that deter-
mines the feature trajectories). Hence, the final position of the robot is close to the
desired one and the error is mostly due to a translation along the Z w axis. The ob-
served errors can be explained by image quantization deviations, presence of noisy
data, camera specifications, low number of features used to compute the planned
trajectory, estimation errors related to the SIFT recognition system, erroneous cam-
era calibration parameters and, finally, the fact that the robot control was based on
the tracking of just one feature in the image.
It is worthwhile to note that the previously presented PBVS VSLAM for servoing
can be compared to the IBVS controller with a VSLAM architecture. Indeed, due to
the very nature of the proposed solution, pure appearance-based visual maps can be
adopted, without considering any 3D spatial information ( i.e. , metric maps). Due to
space limits, the interested reader is referred to [15].

Δ

18.5 Conclusions

The problem of visually guide unicycle-like vehicles towards desired position in
space has been analyzed in this chapter. Solutions here discussed ranges from PBVS
to IBVS approaches applied “in the large”, which differs in which image data are
treated in the control procedure.
For position based, we have proposed a visual servoing scheme for a nonholo-
nomic vehicle in unknown indoor environments. The proposed approach is divided