A question about the approach

[George-Chia]

Hello!
Thanks for your wonderful work.
You predicted the next states of humans with the relational graph containing the robot.
However, you also assumed that the robot is invisible to humans. This means that the next states of humans have nothing to do with the robot.
So I wonder that whether the robot is a useless node, or even a noise node, for the state prediction in the graph.
Looking forward to your reply...

[ChanganVR]

Hi, so there are two experiment settings with the robot. One is being invisible and one is being visible. When the robot is invisible, the graph features learned for humans has nothing to do with the robot (the robot node should not be supplied to human nodes). And when the robot is visible, the human nodes learn more about robot related interaction features.