ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
1

Unable to clear obstacles marking of costmap_2d

asked 2011-05-25 15:23:33 -0600

TayXR gravatar image

updated 2011-05-26 20:24:02 -0600

Hi everyone,

I encounter a problem while trying to use a tilting laser for navigation. I manage to generate a point cloud with the help of the laser assembler and filter out the ground with the ground removal. Rviz is used for visualization.

I tried to use the point cloud for marking of obstacles and the tilted laser scan for clearing obstacles. The marking of obstacles was alright, but when the obstacle was being removed, there was still some traces of marking not removed.

Aftering a few testing, i realise that obstacles that are higher or lower than the plane that the laser mounted on will leave traces of marking. Can anyone tell me why is this happening? Thanks in advance.

Best Regards,

TayXR

edit retag flag offensive close merge delete

2 Answers

Sort by ยป oldest newest most voted
2

answered 2011-05-27 07:05:32 -0600

eitan gravatar image

The costmap raytraces to clear out obstacles. This means that for obstacles to be cleared out of the map the space they occupy needs to be seen through by a new observation. If you're using the 2D grid representation for the costmap, this just means seeing through any part of the column of space in the world above the occupied cell. If you're using the 3D voxel grid representation of the costmap, this means seeing through a 3D grid cell that contains the old obstacle. If your tilting laser doesn't see through enough of these 3D cells, meaning the scan isn't dense enough, its possible to leave some obstacles hanging.

edit flag offensive delete link more

Comments

Hi eitan, thanks for replying to my question. I was using voxel map_type previously which caused this problem. When I changed it to the costmap map_type, all the markings were able to be cleared without leaving any traces. So its might be what u had said, that my scan was not dense enough to see through the 3D cells.
TayXR gravatar image TayXR  ( 2011-05-29 13:55:38 -0600 )edit
I am using hokuyo utm-30lx as my laser sensor. May I know what can i do to increase the density of my laser scan. Does slowing down the tilting speed of the laser sensor helps to increase it?
TayXR gravatar image TayXR  ( 2011-05-29 13:56:27 -0600 )edit
Yes, slowing down the speed at which the laser tilts should give you a denser aggregated scan. There will, of course, be other trade offs with this, like the time it takes to see an obstacle of a certain height with the laser.
eitan gravatar image eitan  ( 2011-05-31 05:03:04 -0600 )edit
0

answered 2011-05-28 22:30:40 -0600

manabu gravatar image

If point clouds and tilt range data have enough noise, some obstacles are not removed. I think the "observation_persistence" parameter in costmap_2d is related to this problem.

edit flag offensive delete link more

Comments

Hi manabu, thanks for replying. Can i ask are there any ways to check if my laser data have any noise, do you think that its possible to use rviz to visualize those noise data? Also, i have tried using 0.0 value for the observation_persistence of both the point cloud and the tilt laser scan observation source but its still cant clear the obstacles.
TayXR gravatar image TayXR  ( 2011-05-29 14:06:39 -0600 )edit
Hi TayXR, To check the scan data in rviz, scan a flat plane and select the data in rviz. Then, choose XYZ type to the point color type. Set the min and max value near the distance from sensor to target plane. you will see the noize point as a different color.
manabu gravatar image manabu  ( 2011-05-31 04:31:35 -0600 )edit

Question Tools

1 follower

Stats

Asked: 2011-05-25 15:23:33 -0600

Seen: 2,600 times

Last updated: May 28 '11