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

Tuning AMCL's diff-corrected and omni-corrected odom models

asked 2016-02-29 10:11:39 -0600

spmaniato gravatar image

updated 2016-04-10 12:48:28 -0600

I just read the discussion on amcl's incorrect odometry models. See:

I was wondering if anyone has figured out how to properly tune the parameters odom_alpha1 through odom_alpha5 (relative to the "incorrect" models). The consensus seems to be that the old defaults (0.2) don't work properly with the corrected models. I tried a few things, such as lowering them by an order of magnitude, but didn't get decent results.

PS. To be more specific, here's my experience with a made-up omni-drive robot in Stage (drive "omni") and the omni-corrected odom model. If I tell the DWAPlanner to treat the robot as nonholonomic (holonomic_robot: false), then AMCL does OK. If I let it command lateral velocities (holonomic_robot: true and same y velocity params as x params), all hell breaks loose. Interestingly, it's mainly the position that's off, not the orientation.

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted

answered 2016-04-10 12:48:15 -0600

spmaniato gravatar image

updated 2016-04-10 12:48:47 -0600

I found some numbers in this test config: amcl/test/small_loop_crazy_driving_prg_corrected.xml

They seem to be in the right ballpark for my omnidirectional platform.

<param name="odom_alpha1" value="0.005"/>
<param name="odom_alpha2" value="0.005"/>
<param name="odom_alpha3" value="0.010"/>
<param name="odom_alpha4" value="0.005"/>
<param name="odom_alpha5" value="0.003"/>
edit flag offensive delete link more

Question Tools



Asked: 2016-02-29 10:11:39 -0600

Seen: 11,914 times

Last updated: Apr 10 '16