google-deepmind / google-deepmind/mujoco
python bindings resource provider API
- Dominant language
- C++
- Stars
- 15.2k
- Forks
- 1.8k
- Avg merge
- 10d 16h
- Merged PRs (30d)
- 25
Description
## The problem
I require a way to build a highly cluttered mujoco model whose assets are loaded dynamically to memory in python. This follows the discussion in issue #1054.
## Suggested solution
The agreed solution in the original issue is to implement access to provide access to [mjpResourceProvider](https://mujoco.readthedocs.io/en/stable/programming/extension.html#resource-providers) through the python bindings.
## Considered alternatives
- loading with `mujoco.MjModel.from_xml_string(xml_string, assets_dict)` is limited to 2,000 assets in the dictionary via the `mjMAXVFS` constant.
- saving the entire model and assets to disk and reloading is highly inefficient.
- using RAMDisk solutions is not multi-platform friendly in python without using third-party file systems.
Contributor guide
Assessment
This issue has not been assessed yet.