v4hn
commented
6 years ago Note that even though this improves the floor detection, it is still not very good and adds artifacts to the floor. We might work on this in the future, but maybe it would be best to disable this completely (after fixing it as proposed int #27) to have a somewhat working move_base navigation.
k-okada
In our setup (an off-the-shelf PR2 with respect to these parts) the previous parameters were too tight: In many situations, especially when an obstacle split the ground-part of a single laser-scan into two segments, one of these segments is not picked up by the line-RANSAC with the old parameters. Thus it ends up as an obstacle obstructing the local costmap of the navigation. With indigo, nobody else noticed this up to now because of a broken filter-chain before that(see #27).