improve odom by laserscan

asked 2014-03-12 08:07:47 -0600

BP

176 ●11 ●13 ●20

updated 2016-10-06 20:15:19 -0600

130s

10937 ●203 ●339 ●330 http://www.linkedin.co...

How can I improve odom of a robot by a laserscan- what package can i use?I don't have static, global map. I found laser_scan_matcher, but I don't know if it's still supported. Maybe i should play with dynamic costmap and amcl? or maybe there is sth better? Thanks in advance

edit retag flag offensive close merge delete

Comments

2

AMCL needs a static map to match against. I've not used laser_scan_matcher but it does look promising, and there's a hydro branch in the source repo. Is using a laser scanner your only option for improving odometry?

jbinney ( 2014-03-12 09:14:13 -0600 )edit

I think it is this option, but maybe i am wrong. What kind of improvement there exist (i know about IMU and this laser-scan-mach)

BP ( 2014-03-12 20:59:45 -0600 )edit

If you have a cmaera, visual odometry is pretty common (track visual features using cameras). Here's a ROS package that does visual odometry: http://wiki.ros.org/viso2_ros

jbinney ( 2014-03-13 20:45:50 -0600 )edit

add a comment

2

answered 2014-03-12 10:28:15 -0600

Stefan Kohlbrecher

24361 ●173 ●299 ●444 https://www.energy-rob...

You can also setup hector_mapping to provide odometry. In the "correct_odometry_publisher" branch there are some changes and this example launch file that has already been used for faking odometry on a robot to provide it to AMCL.

edit flag offensive delete link

Comments

This looks nice, but It provides odmo, not enchances one, right? I assume I'd have to make my own node to make this 2 odom sources "work together".

BP ( 2014-03-12 21:46:29 -0600 )edit

You can fuse two sources of odometry together with a Kalman filter, which if configured properly will improve your base odometry. robot_localization package provide a very nice and easy-to-integrate EKF implementation: http://wiki.ros.org/robot_localization