Odom Publishing Dependency on High-Level Control in unitree_ros2

[CUGerM]

I've been working with unitree_ros2 for deploying a MPC framework on a Unitree robot (e.g., Go2). During my deployment, I observed a behavior that differs from this implementation I've seen.
In my setup with unitree_ros2, it seems that the odometry data published by unitree_ros2 is dependent on the high-level motion control being active. If I disable the high-level control, the odom data either stops publishing or becomes unreliable.
Is this the intended behavior for unitree_ros2? If so, is there a recommended way to obtain the robot's odometry (from IMU, leg encoders, etc.) via ROS2 without requiring the high-level motion control to be running?
Thank you for your time and help!

[marinarasauced]

@CUGerM I have encountered similar behavior with the Go2, and I believe it is the expected behavior since afaik the high-level mode publishes the odometry estimates as part of the SportModeState messages. Have you tried the go2_odometry package from Inria Paris ? I have seemingly had some success with it.

[CUGerM]

@marinarasauced
Thank you so much for your confirmation and recommendation!
I’m currently working on reproducing the legged_control_go2 project, but I found the built-in odometry module there isn’t performing reliably in my setup. That’s why I’ve been looking for alternative solutions to get stable odometry without relying on the high-level control.
I haven’t tried the go2_odometry package yet, but I’ll definitely dive into it and give it a shot. Your suggestion is really helpful for my current progress!