ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question
0

move_base's obstacles don't match laser scan output

asked 2011-08-25 16:48:25 -0600

PKG gravatar image

updated 2011-08-26 09:25:12 -0600

I have a (physical) robot with hokuyo laser mounted. I use rviz to see the raw scans. I also run move_base and subscribe to /move_base_node/local_costmap/obstacles on rviz to see the cells marked as obstacles by move_base.

One phenomenon I notice is that depending on the minimum obstacle height in costmap_common_params.yaml, move_base either sees no obstacles or sees obstacles in every cell. What I expect on rviz is that move_base marks a cell as an obstacle precisely when the laser scan indicates there's an obstacle on the cell. I have been trying to fine tune min_obstacle_height to this. Below 0.21 m, move_base marks every cell as an obstacle; above this, it sees no obstacle at all!

I don't know how to deal with this problem. My costmap_common_params file is:

map_type: costmap

transform_tolerance: 0.3

obstacle_range: 2.0

raytrace_range: 1.5

inflation_radius: 2.5 //has to fine tune the parameter

observation_sources: laser_scan_sensor

laser_scan_sensor: {sensor_frame: laser, data_type: LaserScan, topic: /scan, expected_update_rate: 0.2, observation_persistence: 2.0, marking: true, clearing: true, min_obstacle_height: 0.20, max_obstacle_height: 2.5}

Update after tfoote's comment

Screenshot of rviz with obst height = .21 m :http://tiny.cc/a96op

Screenshot of rviz with obst height = .20 m: http://tiny.cc/yv4k5

Please see the captions of the images for more detail.

edit retag flag offensive close merge delete

Comments

Some screen shots would be really useful to help debug.
tfoote gravatar image tfoote  ( 2011-08-25 19:48:33 -0600 )edit
How far above the ground is your laser? ie what is the height of a laser scan in the odom frame? Also, screenshots from a topdownortho view would probably be better for visualizing this just to eliminate any perspective weirdness (I can't exactly tell if all of the scan is inside light blue or not).
Eric Perko gravatar image Eric Perko  ( 2011-08-27 13:24:18 -0600 )edit

1 Answer

Sort by ยป oldest newest most voted
0

answered 2011-08-27 15:38:06 -0600

PKG gravatar image

I have fixed this - move_base was marking the vehicle's wheels and the laser mount point as obstacles. As a result, the vehicle wasn't moving properly. This wasn't obvious at all intially. After substantial tweaking of laser params, I was able to get move_base to move properly.

edit flag offensive delete link more

Comments

You may want to look into something like the http://www.ros.org/wiki/robot_self_filter package or otherwise filter out the points (maybe with a min_range tweak) that correspond to your base.
Eric Perko gravatar image Eric Perko  ( 2011-08-28 14:36:52 -0600 )edit

Question Tools

Stats

Asked: 2011-08-25 16:48:25 -0600

Seen: 1,156 times

Last updated: Aug 27 '11