scale_action in evaluate_ood_detectors.py

[purentap]

Hi, thank you for making this work publicly available.
While attempting to reproduce the methodology in a different environment, I observed that the actions in my setup are normalized to the range [-1, 1], whereas in your implementation they appear as upscaled values. Could you clarify the purpose of the use of scale_action in evaluate_ood_detectors.py?

[agiachris]

Thank you for the question! The scale_action parameter could be disregarded in the implementation. This parameter is only used within the context of policy's that predict end-effector velocities, i.e., to convert (scale) end-effector velocities to end-effector delta positions based on a known control frequency.