Inconsistent gripper range in actions/states (expected [0,1])

#2
by shaneH2000 - opened

Hi, thanks for your great work! I’m trying to obtain a normalized gripper state in the range [0, 1], but I’m seeing inconsistent gripper scales across tasks.

In particular, for alpha_bot_2_pass_the_sandbag, I observed that the gripper values in actions / states are not within [0, 1]. I also checked gripper_open_scale_state, but it does not match the expected convention (open = 1, close = 0).

Could you clarify which field should be treated as the canonical normalized gripper state, and what is the recommended way to map the gripper signal to [0, 1] for this task (and across tasks)?

Sign up or log in to comment