imu "buffer length" is zero

edit

the imu frame comes from mpu9250 IMU "it is the real imu" not from robot_state_publisher I cant understand you about static transform? and what is deference between my tf_treelink text and others project in this links link textlink textlink text

Does the IMU ever change position RELATIVE to the base_link? If it does not, then that means the transformation between the imu_link and base_link is "static". From your screen shots it looks like the transform is being produced by a tf2_rosstatic_transform_publisher, and your other partners are using either tf or some dynamic tf publisher.

Start a roscore, open a new terminal and run rosrun tf2_ros static_transform_publisher 0 0 0 0 0 0 1 map imu . Then in another terminal run rosrun tf view_frames. Inspect the pdf output of that. Kill the rocore and open a new terminal and run roscore again, and in another terminal run rosrun tf static_transform_publisher 0 0 0 0 0 0 map imu 100. Then in another terminal run rosrun tf view_frames. Inspect the pdf output of that and compare to the previous case. You will see that tf requires you to specify a frequency (100 in this case) and that tf2_ros does not specify a frequency.

Thanks a lot for your reply.

The tf_tree shows buffer zero because I activate the imu_link and base_link transformation in the urdf file.

When I deactivate "comment" this transformation from urdf file. It is works and the imu buffer is 0.697.

It should not matter where you activate it. It matters WHAT you activate. I imagine in one case it is using a different package than in the other (tf or tf2_ros) but I am not familiar enough with URDF files to layout why this does or does not happen. Again, because the transform is static I think "buffer length=0" and "oldest transform=0" (behavior from tf2_ros) is actually a more desirable notation, and works better with ROS timing.