robot_localization setup [closed]

asked 2017-12-08 14:38:14 -0600

phil333 gravatar image

Hi,

I have some visual odometry, wheel odometry and an IMU, and I want to fuse all three.

Currently I am only looking at the visual odometry and the IMU, but I am having a hard time to understand the setup.

I have an odom frame and a base_link frame which is updated by the visual odometry node. Then when i add the robot_localization, i define the following:

    <param name="odom_frame" value="odom"/>
<param name="base_link_frame" value="base_link"/>
<param name="world_frame" value="odom"/>

When I then run the localization node, the base_link starts to twitch around, and get much less stable. What I do not really understand is which frame the localization node is supposed to update. I assumed that I would make a separate frame which would then have the fused localization.

Anyone recommend a step by step tutorial or a working example that I could look at?

edit retag flag offensive reopen merge delete

Closed for the following reason question is not relevant or outdated by Tom Moore
close date 2022-01-25 04:43:16.893043

Comments

A separate frame would not follow rep 105. When using robot_localization to update the tf odom->base_link make sure you have no other node that publishes this transformation.

Humpelstilzchen gravatar image Humpelstilzchen  ( 2017-12-11 00:37:04 -0600 )edit

Also see state_estimation_nodes#published-transforms for what frame is published.

Humpelstilzchen gravatar image Humpelstilzchen  ( 2017-12-11 00:39:19 -0600 )edit

thank you, I disabled everything that was publishing any tf. Now the localisation node does its job, but it is still very noisy. From what I understand, the robot_localisation node does already apply kalman filtering, so its a matter of configuration?

phil333 gravatar image phil333  ( 2017-12-11 02:46:57 -0600 )edit

Probably, mainly your covariances

Humpelstilzchen gravatar image Humpelstilzchen  ( 2017-12-11 05:39:06 -0600 )edit

Since none of my sensors produce covarience matrices, I had to add them myself. I guess, i need to set them to lower values. Is it worth using: initial_estimate_covariance?

phil333 gravatar image phil333  ( 2017-12-11 06:52:08 -0600 )edit

Please publish your entire EKF configuration, and a sample message for every sensor input.

Tom Moore gravatar image Tom Moore  ( 2018-01-26 02:56:01 -0600 )edit