Robot localization does not recognize turns during SLAM and Navigation.

[BjorniFPV]

Robot localization does not recognize turns.

LeoOS v.2.4
LeoRover Version 1.8 upgraded to 1.9.

Following sensors/topics work correctly:
/wheel_odom
/imu/data
/scan

Using ros2 launch leo_nav slam_toolbox.launch.py for launching. No changes to the git code.
Rviz2 for Visualization, fixed frame set to map. Green arrow is /wheel_odom, red arrow is /merged_odom.

/merged_odom behaves weird. Forwards and backwards driving gets recognized correctly. But turns are not. The only turning that can be observed comes from IMU drift. /wheel_odom (green arrow) behaves correctly. The video should give a good overview (https://nc.uni-bremen.de/index.php/s/iLLZtPwNXbPGP6M).

I would appreciate some input to resolve that issue.
Best regards

frames.pdf

[jasiex01]

I have done some testing and /merged_odom works as intended during turning and IMU drift is being filtered out. My setup consisted of Leo Rover 1.9 with newly released LeoOS 2.5.0 and LeoCore firmware version 2.1.0. I also later tested it with LeoOS 2.4.0 and LeoCore firmware 2.0.0.

My suggestion would be to do a sanity check - turn off autonomous stack and observe /merged_odom topic. Also you can try updating the LeoOS and the firmware. As an added bonus, in the new firmware wheel control has been tuned to better support autonomous navigation.

Does your rover have LeoCore or Core2 as the controller?

Additionally, for better response times you can join our Discord community where you will be assigned a private channel for support.

[BjorniFPV]

I retested everything on LeoOS 2.5 with firmware 2.1.0 and observed the same problem.

The sanity check of /merged_odom while running SLAM (which, as far as I am aware, does not require the autonomous stack) fails. The topic does not respond to any wheel odometry changes. Instead, it only drifts slowly, which matches the drift that can also be observed in /imu/data.

The rover is equipped with the LeoCore Controller. One detail that may be relevant is that the rover was originally a v1.8 model and therefore uses the older mounting points. As a result, the LeoCore board is rotated by 180° compared to current models.

I also submitted an application to join the Discord server last week (29.05.2026).

Thank you for your help.

[jasiex01]

You could observer the /imu/data topic itself. On Leo Rover 1.9 the accelerometer readings (linear_acceleration field in IMU messgae). When stationary the x and y fields should be close to 0 while the z field should be affected by gravity. If this is not the case there might exist an issue with the IMU orientation. Please verify the IMU readings on /imu/data. You could also attach output of ros2 topic echo on the /imu/data topics so we could look into it.

[jasiex01]

You can also use service imu_filter/reset_calibration to reset the calibration of the IMU as there might have been an issue with the calibration.