Hi, thank you for releasing the training and deployment code.
I am trying to understand how motion_anchor_pos_b is reproduced on the real robot without motion capture or global torso-position estimation. After comparing the paper, the training repository, and RoboNaldo_Deploy, I found what appears to be a significant difference in observation semantics.
Hi, thank you for releasing the training and deployment code.
I am trying to understand how
motion_anchor_pos_bis reproduced on the real robot without motion capture or global torso-position estimation. After comparing the paper, the training repository, andRoboNaldo_Deploy, I found what appears to be a significant difference in observation semantics.