amcl on quadrotor

asked 2012-01-28 23:29:17 -0500

vivek
53 ●6 ●6 ●12

updated 2014-01-28 17:11:13 -0500

ngrennan
1 ●1 ●1

hey im trying to use AMCL on an asctec pelican, im using odometry from the imu which is not very good. i am able to make a map from hector mapping without odometry, but amcl dosent work , my tf is map -> odom -> base_link -> base_laser , im using a hokuyo laser range finder this is the error im getting TF_OLD_DATA ignoring data from the past for frame /base_link at time 0 according to authority /odometry_publisher Possible reasons are listed at plus Message from [/hokuyo_node] has a non-fully-qualified frame_id [laser_link]. Resolved locally to [/laser_link]. This is will likely not work in multi-robot systems. This message will only print once. can someone please tell me what im missing ?

edit retag flag offensive close merge delete

Comments

Hi ! Do you have any code to share about this project? I'm trying to develop something similar.. Thank you !

polde ( 2017-12-28 07:38:23 -0500 )edit

add a comment

0

answered 2012-01-30 01:24:48 -0500

raahlb
317 ●6 ●13 ●21

updated 2012-01-30 01:26:13 -0500

Hi,

If I've understood things correctly the tf name 'base_laser' and the frame ID in the laser nodes scan messages, 'laser_link', should be identical. Otherwise AMCL won't know what position the scan was recorded from. Though, if that was the problem I think you should have seen an error message... But better fix it.

The message about the frame ID not being qualified can be ignored. It just means that if you have several robots, the system won't be able to tell apart their scans because the frame ID doesn't have a "/<robot name="">/" prefix.

About the TD_OLD_DATA, are you playing back data using rosbag? Have you run "rosparam set use_sim_time true" before starting any nodes? Do you use the --clock option when running "rosbag play"? Is the warning message repeated, or only displayed at the beginning?

edit flag offensive delete link

Comments

Just in case you're interested in solving the non-fully qualified frame ID warning....take a look at this: https://code.ros.org/trac/ros-pkg/ticket/5511

kedarm ( 2013-10-03 06:52:10 -0500 )edit

add a comment

-1

answered 2012-01-30 04:33:27 -0500

vivek
53 ●6 ●6 ●12

yea those problems have been sorted out , now everything runs , im getting a good costmap and my local and global planners work but amcl dosent work , is there any way to get AMCL working without odometry, right now the odom frame is only using data from an IMU which is very very bad.

edit flag offensive delete link

Comments

My guess would be that constant zero odometry would be better than the integrated IMU odometry, which I'd guess to drift meters away within seconds or minutes at the latest. In case AMCL doesn't work let me know if you want a localization only version of hector_mapping.

Stefan Kohlbrecher ( 2012-01-30 04:37:13 -0500 )edit

please ask this as a new question, and post here what you did to fix your previous problem. This is a Stack Overflow style QA site, not a forum with threads.

tfoote ( 2012-01-30 05:45:45 -0500 )edit

Please include new information in a comment to an answer or by editing your question to add the new info. ROS Answers is not a forum/mailing list format and answers may be reordered by votes in the future, so replies-as-answers won't make sense later. Thanks for helping keep ROS Answers organized

mjcarroll ( 2012-02-02 09:16:03 -0500 )edit

add a comment

-2

answered 2012-01-30 19:46:27 -0500

vivek
53 ●6 ●6 ●12

@stefan - thanks for the info but amcl dosent work with the constant zero odometry too , but im trying to use laser scan matcher to provide odometry. im already implementing an optical flow algorithm for localization and stability, i need amcl just to cross check with the map as i need precise co-ordinates and optical flow does have error which can pile up as time goes on. If u do have something that can help me with map reading can u please help me , just in case amcl dosent work with laser scan matcher

@tfoote - thanks for letting me know , im new to this forum , and wont make such a mistake again

edit flag offensive delete link

Comments

Actually AMCL only does anything after the robot has moved a certain distance or turned a certain amount. You need an at least semi-accurate odometry source for it to do anything. I guess you could just try generating random movements to force it to update...

raahlb ( 2012-01-30 20:24:48 -0500 )edit

Shouldn't one be able to disable this behavior by setting update_min_d and update_min_a both to zero? This together with zero mean odometry and increased noise for all motion model parameters should be the best starting point for the given scenario IMHO, at least in theory.

Stefan Kohlbrecher ( 2012-01-30 23:22:41 -0500 )edit

Hm, true, definitely worth testing, with zero odometry. I was kind of assuming that it would not work because it's not intended to be used like that. . Hm, though isn't the noise proportional to the distance traveled, so if you go nowhere you'll have no new particles?

raahlb ( 2012-01-31 01:34:16 -0500 )edit

Yes this could indeed be the case. So it might be necessary to make modifications to the motion model inside AMCL for the described application.

Stefan Kohlbrecher ( 2012-01-31 09:33:09 -0500 )edit

Please include new information in a comment to an answer or by editing your question to add the new info. ROS Answers is not a forum/mailing list format and answers may be reordered by votes in the future, so replies-as-answers won't make sense later. Thanks for helping keep ROS Answers organized

mjcarroll ( 2012-02-02 09:15:58 -0500 )edit

add a comment

amcl on quadrotor

Comments

3 Answers

Comments

Comments

Comments

Question Tools

Stats

Related questions

amcl on quadrotor edit

Comments

3 Answers

Comments

Comments

Comments

Question Tools

Stats

Related questions

amcl on quadrotor