Difference between tf2 broadcaster and tf2 frame.

edit

If that is the case, which example should I use if I want to add an IMU frame?

Would it be the first example? Since I can subscribe to Imu/raw, and update the tf for each imu message?

Also, do you mind if you can give an real-life example of when should I use the second case? Thanks a lot!!!

The imu frame represents the position of the sensor on the robot: it shouldn't change with each message provided by the sensor, but only when calibrating it. Hence the "static tf publisher" would be more adequate. If you are using the data of the imu to compute the "odometry" of your robot, then the transformation from the /odom frame to your robot should be updated and published continuously.

I see, and in that case, I will just be publishing a odometry message (http://docs.ros.org/en/noetic/api/nav...), with header_id to be odometry, and child_id to be imu?

Also, will it be correct if I use the dynamic tf transform in situations such as imu is attached to a robotic arm? Or should I still just use a static transform, and rely on the odometry message as mentioned above?

Also, I wonder if you can give me an example on when should I use a dynamic tf transform instead of a static transform? Thank you very much!

You would use a dynamic tf transform to publish transformations that are not fixed, like for example the https://github.com/ros/joint_state_pu... does. The child Id of the odometry transform is usually a fixed frame on your robot, for example base_link. If you are using the information produced by the imu to compute the odometry, then it's a different story.