Fix confusing ee_pos and ee_rot

[Dipan-Zhang]

Description

This PR fixes a bug in the dex_teleoperation data collection pipeline where ee_camera_pose was incorrectly assigned to both ee_pos and ee_rot, and the recording is missing the absolute end-effector pose.

Changes:

• Removed redundant ee_pos and ee_rot that falsely recording ee camera pose.
• Added new input ee_pose to record the absolute end-effector pose directly from observation.
• This fix enables accurate recording of absolute end-effector pose in cartesian space, enabling cartesian-based Imitation Learning.

Verification

• Hardware: Tested on Stretch Robot.
• Status: Validated that recorded data now reflects absolute poses. Training is currently in progress using the corrected dataset.

Checklist

• I have performed a self-review of my code
• If it is a core feature, I have added thorough tests
• I have added documentation for the changes
• I have updated the README file if necessary
• I have run on hardware if necessary

Additional context

Add any other context or information about the pull request here.

[CLAassistant]

All committers have signed the CLA.

[hello-robot-shehab]

Hi @Dipan-Zhang , thank you for your PR.

I believe you may be missing changes to your def add() method in https://github.com/hello-robot/stretch_ai/blob/5d963ab849f56ec42b9ff7575ec68cf0a6fbb7b1/src/stretch/utils/data_tools/record.py#L102 because it accepts head_rgb but not head_camera_rgb, and similarly for the other kwargs.

Thank you for taking a look at this.

Best,
Shehab

[Dipan-Zhang]

Hi Shehab,

Thanks for the heads up! I added the missing argument changes to the FileDataRecorder.add() function, and kept the arguments unchanged as much as possible.

Best,
Anran