Why are the action spaces infinite?

Omniverse Isaac Sim
,
1

Please provide all relevant details below before submitting your post. This will help the community provide more accurate and timely assistance. After submitting, you can check the appropriate boxes. Remember, you can always edit your post later to include additional information if needed.

Isaac Sim Version

4.2.0
4.1.0
4.0.0
2023.1.1
2023.1.0-hotfix.1
Other (please specify):

Isaac Lab Version (if applicable)

1.2
1.1
1.0
Other (please specify):

Operating System

Ubuntu 22.04
Ubuntu 20.04
Windows 11
Windows 10
Other (please specify):

GPU Information

  • Model: NA
  • Driver Version: NA

Topic Description

Detailed Description

I want to train an RL using the built-in Isaac-Velocity-Flat-Unitree-Go2-v0, but I’m not sure how to interpret the action space. In particular, the action space is in the range [-inf, inf]. Why is the action space infinite if the actuators on the robot have finite capacity? Is clipping applied somewhere? What if I want to use a bounded representation for my policy, such as “squashed Gaussian?” Any guidance would be appreciated!

Steps to Reproduce

  1. Create the environment, e.g., env = gymnasium.make("Isaac-Velocity-Flat-Unitree-Go2-v0")
  2. inspect env.action_space

Error Messages

NA

Screenshots or Videos

NA

Additional Information

What I’ve Tried

NA

Related Issues

NA

Additional Context

NA

3

Thank you for your interest in Isaac Lab. To ensure efficient support and collaboration, please submit your topic to its GitHub repo following the instructions provided on Isaac Lab’s Contributing Guidelines regarding discussions, submitting issues, feature requests, and contributing to the project.

We appreciate your understanding and look forward to assisting you.