Unable to detect obstacles in costmap [closed]

asked 2011-08-07 16:35:52 -0500

PKG gravatar image

updated 2011-08-08 07:00:54 -0500

I am running move_base , taking scan input from hokuyo laser, using laser_scan_matcher for localization, and using a blank costmap to start with. The global and local costmap params use rolling window instead of static map. My relevant launch files are dumped here :http://pastebin.com/8SArVL9p .

My problem is that I cannot see any obstacles in the costmap (I made logs in the costmap_2d.cpp & costmap_2d_ros.cpp to register obstacles). As such, move_base sends commands that go into obstacles, whenever I send it a goal. I suspect there is an issue with my costmap params, and would like to know which. I don't know if the initial, blank costmap I send causes a problem.

The robot controller computer on which I run all this has rviz segfaulting, and I'm unable to bring up rviz after exhaustive debugging ; but I was able to verify the presence of nearby obstacles by (temporarily) connecting the laser to a different computer that ran rviz.

Note:

(1) I have altered laser_scan_matcher to send a tf from /map -> /odom (this agrees very well with my previous tests) instead of to map -> baselink.

(2) I am also publishing a laser->base_link and a /odom->base_link transform separately.

edit retag flag offensive reopen merge delete

Closed for the following reason question is not relevant or outdated by tfoote
close date 2012-02-23 16:46:34

Comments

Are you sure that your logging is giving you the right data? It seems strange that you saw obstacles when you hooked the laser up to rviz, but not when you use your own logging.
eitan gravatar image eitan  ( 2011-08-08 05:46:48 -0500 )edit
Also, is your laser data really stamped in the "map" frame? That seems strange and likely means that you'll only accumulate scans near the origin of the map.
eitan gravatar image eitan  ( 2011-08-08 05:47:51 -0500 )edit