Add versioned AgiBot observed-replay physical assets
Browse filesPublish the independently qualified collision model for DexForce/EmbodiChain#1212 and #1214 through the project asset repository. Preserve upstream CC-BY-NC-SA-4.0 attribution, exact qualified inputs and native evidence. Existing assets are unchanged.
README.md
CHANGED
|
@@ -6,3 +6,8 @@ Resource-specific third-party terms are retained. In particular,
|
|
| 6 |
[`demo/real2sim_carton`](demo/real2sim_carton/README.md) contains DROID originals
|
| 7 |
and measured derivatives under CC BY 4.0; its per-file manifest and attribution
|
| 8 |
notice take precedence over this card's default license for those files.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 6 |
[`demo/real2sim_carton`](demo/real2sim_carton/README.md) contains DROID originals
|
| 7 |
and measured derivatives under CC BY 4.0; its per-file manifest and attribution
|
| 8 |
notice take precedence over this card's default license for those files.
|
| 9 |
+
|
| 10 |
+
The separately versioned [AgiBot observed-replay robot](robot_assets/AgibotG1ObservedReplayV1/README.md)
|
| 11 |
+
retains **CC-BY-NC-SA-4.0** for the robot resources and included AgiBot World
|
| 12 |
+
recording. These third-party terms take precedence over this card's default
|
| 13 |
+
Apache-2.0 license. Each asset includes attribution, source revisions and hashes.
|
robot_assets/AgibotG1ObservedReplayV1.zip
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bfedc336c02c61f4998da67b5f6df6e954725d2a023f8f104cc649451f3cf9f6
|
| 3 |
+
size 7387156
|
robot_assets/AgibotG1ObservedReplayV1/README.md
ADDED
|
@@ -0,0 +1,52 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# AgiBot G1 observed-replay physical model v1
|
| 2 |
+
|
| 3 |
+
This independently versioned URDF repairs the missing collision model in
|
| 4 |
+
EmbodiChain Real2Sim sample `03_agibot_toy` ([issue #1212](https://github.com/DexForce/EmbodiChain/issues/1212)).
|
| 5 |
+
The original official `AgibotG1` USD asset remains unchanged.
|
| 6 |
+
|
| 7 |
+
[Download the archive](../AgibotG1ObservedReplayV1.zip) · [Artifact hashes](artifact.json) · [SHA256SUMS](SHA256SUMS)
|
| 8 |
+
|
| 9 |
+
In EmbodiChain with [PR #1214](https://github.com/DexForce/EmbodiChain/pull/1214):
|
| 10 |
+
|
| 11 |
+
```bash
|
| 12 |
+
embodichain data download --name AgibotG1ObservedReplay
|
| 13 |
+
```
|
| 14 |
+
|
| 15 |
+
```python
|
| 16 |
+
from embodichain.data import get_data_path
|
| 17 |
+
urdf = get_data_path("AgibotG1ObservedReplay/robot_physical/robot.urdf")
|
| 18 |
+
```
|
| 19 |
+
|
| 20 |
+
The registered asset pins a distribution revision and verifies archive and
|
| 21 |
+
extracted-file digests in the versioned `AgibotG1ObservedReplayV1` cache.
|
| 22 |
+
For manual use, extract the flat ZIP into a new directory and load
|
| 23 |
+
`robot_physical/robot.urdf`; all mesh and texture paths are local. The included
|
| 24 |
+
README demonstrates direct `RobotCfg` construction; `robot.json` records the
|
| 25 |
+
observed joint/TCP/gripper metadata. No repair script or upstream asset download is
|
| 26 |
+
needed for ordinary use.
|
| 27 |
+
|
| 28 |
+
The derivative preserves the recorded model's original joints, inertials,
|
| 29 |
+
41 visuals, TCP definitions and gripper calibration. It adds 14 official arm
|
| 30 |
+
cylinders and 22 authored hand collision meshes. It includes the calibrated
|
| 31 |
+
trajectory used for qualification, but no videos or reconstructed scene.
|
| 32 |
+
|
| 33 |
+
Native CPU/default-backend qualification passed 11 checks over 1,920 steps at
|
| 34 |
+
600 Hz, including all 36 collision-bearing links, finite state, calibrated
|
| 35 |
+
opening/closing, four-pad contact and release. Exact input bindings, the native
|
| 36 |
+
report and its trace are included. This establishes model collision/contact
|
| 37 |
+
behavior, not toy-placement success, stable grasping or hardware accuracy.
|
| 38 |
+
Official arm cylinders and differences from USD mesh cooking are disclosed in
|
| 39 |
+
`robot_physical/collision_provenance.json`.
|
| 40 |
+
|
| 41 |
+
## Sources and license
|
| 42 |
+
|
| 43 |
+
Original prepared input: `DestoFaygo/embodichain-real2sim-inputs` at
|
| 44 |
+
`1b34442c9e3d3e6a3139d3df95d5cb0e61a93e1f`. The included AgiBot World recording
|
| 45 |
+
is from episode `619/839665.hdf5`; calibration and prepared joint arrays remain
|
| 46 |
+
unchanged. Official robot sources are AgiBot / agibot-world GenieSimAssets
|
| 47 |
+
G1_120s, pinned in the bundled provenance file.
|
| 48 |
+
|
| 49 |
+
The robot resources and included AgiBot World recording retain
|
| 50 |
+
**CC-BY-NC-SA-4.0 (non-commercial)**. Full upstream license text, attribution,
|
| 51 |
+
source hashes and modification provenance are bundled; the data repository's
|
| 52 |
+
default Apache-2.0 license does not replace those terms.
|
robot_assets/AgibotG1ObservedReplayV1/SHA256SUMS
ADDED
|
@@ -0,0 +1,2 @@
|
|
|
|
|
|
|
|
|
|
| 1 |
+
bfedc336c02c61f4998da67b5f6df6e954725d2a023f8f104cc649451f3cf9f6 AgibotG1ObservedReplayV1.zip
|
| 2 |
+
f7714cb068c07fdf3933b10c8759045d678e31c92ef6cb51b699f8df741036be artifact.json
|
robot_assets/AgibotG1ObservedReplayV1/artifact.json
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"source_package": {
|
| 3 |
+
"repository": "DestoFaygo/embodichain-real2sim-inputs",
|
| 4 |
+
"revision": "1b34442c9e3d3e6a3139d3df95d5cb0e61a93e1f",
|
| 5 |
+
"url": "https://huggingface.co/datasets/DestoFaygo/embodichain-real2sim-inputs/resolve/1b34442c9e3d3e6a3139d3df95d5cb0e61a93e1f/real2sim_inputs_5.zip",
|
| 6 |
+
"sha256": "e61038b8d49c91822439032f0343d280059d50f9945d5aab68edb1286ae6ffca",
|
| 7 |
+
"checksums_sha256": "f94951f0704bdf1095c581b4ca5637586f7c95c25a66a6c1759078dd86fbeca2"
|
| 8 |
+
},
|
| 9 |
+
"standalone_robot": {
|
| 10 |
+
"file": "AgibotG1ObservedReplayV1.zip",
|
| 11 |
+
"sha256": "bfedc336c02c61f4998da67b5f6df6e954725d2a023f8f104cc649451f3cf9f6",
|
| 12 |
+
"md5": "7aa86d274e394309c5f9955f5b1857ef",
|
| 13 |
+
"size": 7387156,
|
| 14 |
+
"archive_root": "",
|
| 15 |
+
"checksums_sha256": "e109121caa4ff8a211c472b7dadcae3d2516f18ca76b37e1a9863fd82cdc1185",
|
| 16 |
+
"asset_id": "AgibotG1ObservedReplay",
|
| 17 |
+
"repository": "DexForceAI/embodichain_data",
|
| 18 |
+
"path": "robot_assets/AgibotG1ObservedReplayV1.zip",
|
| 19 |
+
"urdf": "robot_physical/robot.urdf",
|
| 20 |
+
"license": "CC-BY-NC-SA-4.0"
|
| 21 |
+
},
|
| 22 |
+
"robot_geometry_sha256": "eb2df06b27924cb8c037992e44ec1b71cf6cd39f97fa1b2d84326d294f7fefec",
|
| 23 |
+
"native_qualification": {
|
| 24 |
+
"status": "passed",
|
| 25 |
+
"checks_passed": 11,
|
| 26 |
+
"steps": 1920,
|
| 27 |
+
"report_sha256": "329d1662e1da30cf6a39b743135774f5a76d8b7af1d80a5ddf6db07c6f452791",
|
| 28 |
+
"code_sha256": "2173888a3238ac4037087027c77a989854bae4eba1d112059fea3f0981459d3e",
|
| 29 |
+
"trace_sha256": "ca3b6e5bf9cadae9929cdca9fcb688068a30fdf0a73d182179cbbcb24a81e2b6",
|
| 30 |
+
"task_success_verified": false
|
| 31 |
+
},
|
| 32 |
+
"validation": {
|
| 33 |
+
"relocation": "passed",
|
| 34 |
+
"qualified_input_bindings_preserved": 66,
|
| 35 |
+
"robot_files_preserved": 66,
|
| 36 |
+
"checksummed_members": 73
|
| 37 |
+
}
|
| 38 |
+
}
|