14. Real Asset

In this tutorial, we will show you how to use real asset generated from EmbodiedGen in MetaSim.

Common Usage

python get_started/14_real_assets.py  --sim <simulator>

In headless mode:

python get_started/14_real_assets.py  --sim isaacsim --headless
python get_started/14_real_assets.py  --sim isaacgym --headless
MUJOCO_GL=egl python get_started/14_real_assets.py  --sim mujoco --headless
python get_started/14_real_assets.py  --sim genesis --headless
python get_started/14_real_assets.py  --sim sapien3 --headless
python get_started/14_real_assets.py  --sim pybullet --headless

You will get the following image:

Isaac Lab	Isaac Gym	Mujoco

Genesis	Sapien	PyBullet

Asset Converter

Use EmbodiedGen generated assets with correct physical collisions and consistent visual effects in MetaSim. (isaacsim, mujoco, genesis, pybullet, isaacgym, sapien). Example in generation/tests/test_asset_converter.py.

Simulator	Conversion Class
isaacsim	MeshtoUSDConverter
mujoco	MeshtoMJCFConverter
genesis / sapien / isaacgym / pybullet	EmbodiedGen generated .urdf can be used directly

from huggingface_hub import snapshot_download
from generation.asset_converter import AssetConverterFactory, AssetType

data_dir = "roboverse_data/assets/EmbodiedGenData"
snapshot_download(
    repo_id="HorizonRobotics/EmbodiedGenData",
    repo_type="dataset",
    local_dir=data_dir,
    allow_patterns="demo_assets/*",
    local_dir_use_symlinks=False,
)

target_asset_type = AssetType.MJCF # or AssetType.USD

urdf_paths = [
    f"{data_dir}/demo_assets/remote_control/result/remote_control.urdf",
]

if target_asset_type == AssetType.MJCF:
    output_files = [
        f"{data_dir}/demo_assets/remote_control/mjcf/remote_control.mjcf",
    ]
    asset_converter = AssetConverterFactory.create(
        target_type=AssetType.MJCF,
        source_type=AssetType.URDF,
    )
elif target_asset_type == AssetType.USD:
    output_files = [
        f"{data_dir}/demo_assets/remote_control/usd/remote_control.usd",
    ]
    asset_converter = AssetConverterFactory.create(
        target_type=AssetType.USD,
        source_type=AssetType.MESH,
    )

with asset_converter:
    for urdf_path, output_file in zip(urdf_paths, output_files):
        asset_converter.convert(urdf_path, output_file)