False positives with laser scan
Hi, I'm trying to work on an obstacle avoidance algorithm using Husky gazebo where I try to reach the desired goal. The algorithm seems to work well but the issue is that the laser scan range data (the data at /scan topic) sometimes outputs really low values even when there's no object in sight or the object is far. An example screenshot is here. The laser scan output for that state is here. I initially thought that I could simply ignore those values if there were less than, say, 20 of them. But as you can see, the laser scan data for the "front" region seems to have 60 values (which is not a small number) around 0.17 (this value is above range_min=0.1) even though the closest object is much farther than that. I'm really not sure why this is happening because the other range values seem to be accurate. If anyone has any ideas, that would be really helpful!
Asked by suohmikoto1234 on 2021-12-14 18:19:31 UTC
Answers
If you are using ray sensor, check for any collisions that may be close to your sensor (you can show them by clicking View->Collisions on the top of the GUI. If it's gpu_ray, then it's detecting visuals, so some visuals in your sensor or robot may get in the way.
You can use the visualize tag to show the scan points in the GUI. That should help you understand more where the close points were from.
I'm not sure about the behavior of the the ray sensor, but for gpu_ray, the scans see through the faces closer than the minimum range, and hit the next faces behind instead.
Asked by Veerachart on 2021-12-20 03:45:54 UTC
Comments
Did you check your real time update rate and step size in world physics. I had this problem when using non standard real time update rates and step sizes. I havent found a solutioin but using the standard values.
Asked by joshuacv on 2022-04-01 00:51:02 UTC