search

strands-project / scitos_2d_navigation

The scitos_2d_navigation stack holds common configuration options for running the 2D navigation stack on a Scitos A5 robot.
2 stars 15 forks source link

Moving over thresholds #17

Open hawesie opened 11 years ago

hawesie commented 11 years ago

At TUW there are problems with door thresholds. We need to address these. Can someone outline the problems again? @ToMadoRe @arbeitor @mzillich

ToMadoRe commented 11 years ago

I tried to take a picture from the thresholds that we are struggling with. Especially the second case is really risky for the robot as it becomes very unstable and seems to lose balance completely. That's why we avoid going over that threshold at the moment but might consider some floor modifications in the future.

For the first threshold, it is rather tiny, it creates navigation problems because the robot turns to the side a bit and therefore gets wrong odometry data. This messes up the path planner in a tiny room like ours and eventually the robot gets stuck (as there are obstacles either in the current laser scan or predicted from a misaligned map).

For the marathon, we are going to use the robot in an area without these nasty thresholds. Still, we would need to address this issue to get it working in the lab as well.

img_1020 img_1021

hawesie commented 11 years ago

I wonder if the force parameter from https://github.com/strands-project/strands_management/issues/22 might improve threshold performance?

ToMadoRe commented 11 years ago

we have moved our robot to a newer building without such thresholds. But we will investigate that after the marathon at the latest.