Improving Area Center Robot Navigation Using a Novel Range Scan Segmentation Method - Foundations on Natural and Artificial Computation - page 241

Fig. 5. Dynamic split points. In a) robot is slightly closer to the segment end marked
with the arrow than to the other segment end. Then when area center becomes inac-
cessible inside the obstacle, the marked end is selected as split point and right sector is
selected due to goal direction, this would lead to a dangerous situation b) and c), like
in fig. 4, with the basic split point strategy. It would be better to change split point
to the other end when the other segment end is closer to robot than split point, d), so
area center path is far away enough from obstacle. The selected sector has dark blue
color and the unselected one has a lighter color.
In order to get a safer and more robust navigation, specially in crowded or
dynamic environments, we gave a step forwards enhancing split points behavior
must be improved. Until new split points were static, but to response quickly
to environment changes it would be better if they were dynamic. Even if the
split point was well located at first time it is possible that a new protuberance
appears in robot path without making area center inaccessible. In this case
no split process is triggered, a basic solution for this problem is to change the
position of a split point. So, if some segment end, located between split point and
the bisector of current sector plus an offset, is closer to the robot, such segment
end becomes the split point. This situation is depicted in fig. 5, and similar to
one described in fig. 4, selected sector contains big protuberances. Changing split
point to a new corner avoids the problem, see fig. 5 d).