New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Domain Randomization Help #5
Comments
Hello! We did reply to the other message you've posted on dm-robotics. Here is a copy past of the message, if this does not work please feel free to ask more questions! My understanding is that the friction parameters you are setting are not having any effect in the simulation. The reason for this is that you are using doing this in the episode_initialiser where you should be doing it in the scene_initialiser (https://github.com/deepmind/dm_robotics/blob/main/py/moma/base_task.py#L58). The reasoning is the following, there is the physics model and the physics data. The model compiled from an mjcf and you cannot modify the model parameters after compilation. The friction being part of the model this explains why you are not seeing any changes.
|
@alaurens so i want to be certain , if my code for initializing friction is correct. Thanks for the sceneinit code too. do i need to set the friction on all geoms |
Hi I am trying to implement domain randomization.
Between scene_initialization and domain initialization where will be the best place to initialize the physics, mass as mentioined in the paper.
How does one really initialize the friction, Can you please examine my code.
How does one implement action delay, are the rewards and observations delayed too
@dataclasses.dataclass
class DomainRandomizer:
def init(self, basket, props, robot):
The text was updated successfully, but these errors were encountered: