odometry not suitable for ROS navigation stack

[moritzboeker]

Hi everyone,

we wanted to use the ROS navigation stack for the FRE.
I did a quick check if the odometry is suitable for the navigation stack according to:
Chapter 1.2 Odometry in the ROS Navigation Tuning Guide
http://wiki.ros.org/navigation/Tutorials/Navigation%20Tuning%20Guide

check rotational odometry

• run simulation: robot is face row entries
• run rviz with robot visualization
• • fixed frame to odom
• • show laserscan
• • change decay time of laserscan to 20s
• • perform in-place rotation
• ideally the laserscan should stay roughly at the same place even though the robot is rotating
• but the result looks like this: the rows are also rotating https://ibb.co/8cHRxf1

check translational odometry

• run simulation: robot is facing the field from the side
• run rviz with robot visualization
• • fixed frame to odom
• • show laserscan
• • change decay time of laserscan to 20s
• • move robot slowly in x-direction towards the field
• ideally the laserscan of the side of the rows should stay roughly at the same place even though the robot is moving
• but the result looks like this: the rows are also translating https://ibb.co/C0fnXnJ

Thus it seems that the ROS navigation stack cannot be used :-(

Has anyone found a solution to this problem?

Sincerely,

Moritz

[moritzboeker]

I have seen that there is the topic /odometry/filtered which fuses together the odometry and imu data by the ekf_localization node. But this ekf_localization already gives this information forth to /tf in form of the link between /odom and /base_link
But still the odometry is not stable :-(