uooon commited on
Commit
0330296
·
1 Parent(s): eb3fac2

Add reference demos

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. LIBERO-GeoMotion/task_000/steps.txt +31 -0
  2. LIBERO-GeoMotion/task_000/trajectory.json +31 -0
  3. LIBERO-GeoMotion/task_000/trajectory.jsonl +0 -0
  4. LIBERO-GeoMotion/task_000/videos/agentview.mp4 +3 -0
  5. LIBERO-GeoMotion/task_000/videos/videos.json +19 -0
  6. LIBERO-GeoMotion/task_000/videos/wrist.mp4 +3 -0
  7. LIBERO-GeoMotion/task_004/steps.txt +22 -0
  8. LIBERO-GeoMotion/task_004/trajectory.json +31 -0
  9. LIBERO-GeoMotion/task_004/trajectory.jsonl +0 -0
  10. LIBERO-GeoMotion/task_004/videos/agentview.mp4 +3 -0
  11. LIBERO-GeoMotion/task_004/videos/videos.json +19 -0
  12. LIBERO-GeoMotion/task_004/videos/wrist.mp4 +3 -0
  13. LIBERO-GeoMotion/task_008/steps.txt +26 -0
  14. LIBERO-GeoMotion/task_008/trajectory.json +31 -0
  15. LIBERO-GeoMotion/task_008/trajectory.jsonl +0 -0
  16. LIBERO-GeoMotion/task_008/videos/agentview.mp4 +3 -0
  17. LIBERO-GeoMotion/task_008/videos/videos.json +19 -0
  18. LIBERO-GeoMotion/task_008/videos/wrist.mp4 +3 -0
  19. LIBERO-GeoMotion/task_026/steps.txt +20 -0
  20. LIBERO-GeoMotion/task_026/trajectory.json +31 -0
  21. LIBERO-GeoMotion/task_026/trajectory.jsonl +0 -0
  22. LIBERO-GeoMotion/task_026/videos/agentview.mp4 +3 -0
  23. LIBERO-GeoMotion/task_026/videos/videos.json +19 -0
  24. LIBERO-GeoMotion/task_026/videos/wrist.mp4 +3 -0
  25. LIBERO-GeoMotion/task_028/steps.txt +61 -0
  26. LIBERO-GeoMotion/task_028/trajectory.json +31 -0
  27. LIBERO-GeoMotion/task_028/trajectory.jsonl +0 -0
  28. LIBERO-GeoMotion/task_028/videos/agentview.mp4 +3 -0
  29. LIBERO-GeoMotion/task_028/videos/videos.json +19 -0
  30. LIBERO-GeoMotion/task_028/videos/wrist.mp4 +3 -0
  31. LIBERO-GeoMotion/task_034/steps.txt +57 -0
  32. LIBERO-GeoMotion/task_034/trajectory.json +31 -0
  33. LIBERO-GeoMotion/task_034/trajectory.jsonl +0 -0
  34. LIBERO-GeoMotion/task_034/videos/agentview.mp4 +3 -0
  35. LIBERO-GeoMotion/task_034/videos/videos.json +19 -0
  36. LIBERO-GeoMotion/task_034/videos/wrist.mp4 +3 -0
  37. README.md +177 -0
  38. README.zh-CN.md +177 -0
  39. RoboDojo/arrange_largest_number/steps.txt +5 -0
  40. RoboDojo/arrange_largest_number/trajectory.json +37 -0
  41. RoboDojo/arrange_largest_number/trajectory.jsonl +0 -0
  42. RoboDojo/arrange_largest_number/videos/head.mp4 +3 -0
  43. RoboDojo/arrange_largest_number/videos/left_wrist.mp4 +3 -0
  44. RoboDojo/arrange_largest_number/videos/right_wrist.mp4 +3 -0
  45. RoboDojo/arrange_largest_number/videos/videos.json +8 -0
  46. RoboDojo/build_tower/steps.txt +7 -0
  47. RoboDojo/build_tower/trajectory.json +37 -0
  48. RoboDojo/build_tower/trajectory.jsonl +0 -0
  49. RoboDojo/build_tower/videos/head.mp4 +3 -0
  50. RoboDojo/build_tower/videos/left_wrist.mp4 +3 -0
LIBERO-GeoMotion/task_000/steps.txt ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ## Expert Procedure
2
+
3
+ Grasp and lift the blue bar. Reorient it, set it down, and regrasp it near one end. Move it through the narrow opening, release it on the green pad, and retreat.
4
+
5
+ ## Action Details
6
+
7
+ | Steps (0-based) | Time (s) | Action | End-effector XYZ at segment end (m) | Gripper command |
8
+ | --- | --- | --- | --- | --- |
9
+ | 0–76 | 0.00–3.85 | Grasp and lift | -0.1615, 0.1083, 1.1204 | Open |
10
+ | 77–107 | 3.85–5.40 | Grasp and lift | -0.1619, 0.1081, 1.0286 | Open |
11
+ | 108–137 | 5.40–6.90 | Grasp and lift | -0.1621, 0.1083, 1.0260 | Close |
12
+ | 138–170 | 6.90–8.55 | Grasp and lift | -0.1629, 0.1078, 1.1175 | Close |
13
+ | 171–247 | 8.55–12.40 | Set down / regrasp / pass through opening | -0.1510, 0.1194, 1.2271 | Close |
14
+ | 248–308 | 12.40–15.45 | Set down / regrasp / pass through opening | -0.1510, 0.1197, 1.0342 | Close |
15
+ | 309–343 | 15.45–17.20 | Set down / regrasp / pass through opening | -0.1507, 0.1200, 1.0318 | Open |
16
+ | 344–401 | 17.20–20.10 | Set down / regrasp / pass through opening | -0.1508, 0.1199, 1.2054 | Open |
17
+ | 402–651 | 20.10–32.60 | Set down / regrasp / pass through opening | -0.3007, 0.1240, 1.2384 | Open |
18
+ | 652–730 | 32.60–36.55 | Set down / regrasp / pass through opening | -0.3059, 0.1201, 1.0302 | Open |
19
+ | 731–759 | 36.55–38.00 | Set down / regrasp / pass through opening | -0.2382, 0.1196, 1.0274 | Open |
20
+ | 760–799 | 38.00–40.00 | Set down / regrasp / pass through opening | -0.2358, 0.1195, 1.0274 | Close |
21
+ | 800–852 | 40.00–42.65 | Set down / regrasp / pass through opening | -0.2358, 0.1194, 1.1664 | Close |
22
+ | 853–865 | 42.65–43.30 | Set down / regrasp / pass through opening | -0.2055, 0.1189, 1.1570 | Close |
23
+ | 866–918 | 43.30–45.95 | Set down / regrasp / pass through opening | -0.0576, 0.1194, 1.1532 | Close |
24
+ | 919–984 | 45.95–49.25 | Set down / regrasp / pass through opening | 0.1331, 0.1389, 1.1534 | Close |
25
+ | 985–1028 | 49.25–51.45 | Set down / regrasp / pass through opening | 0.1341, 0.1422, 1.0470 | Close |
26
+ | 1029–1029 | 51.45–51.50 | Set down / regrasp / pass through opening | 0.1342, 0.1421, 1.0471 | Open |
27
+ | 1030–1209 | 51.50–60.50 | Set down / regrasp / pass through opening | 0.0941, 0.1426, 1.0611 | Open |
28
+ | 1210–1409 | 60.50–70.50 | Set down / regrasp / pass through opening | 0.0953, 0.1431, 1.0594 | Open |
29
+ | 1410–1659 | 70.50–83.00 | Set down / regrasp / pass through opening | 0.0936, 0.1421, 1.0632 | Open |
30
+ | 1660–1909 | 83.00–95.50 | Set down / regrasp / pass through opening | 0.0460, 0.1189, 1.1787 | Open |
31
+ | 1910–1949 | 95.50–97.50 | Wait for settling after release | 0.0459, 0.1188, 1.1786 | Open |
LIBERO-GeoMotion/task_000/trajectory.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "libero-codex-trajectory-v1",
3
+ "frames": 1950,
4
+ "fps": 20,
5
+ "pose": {
6
+ "reference": "robot0_grip_site",
7
+ "coordinate_frame": "environment",
8
+ "position_unit": "metre",
9
+ "quaternion_order": "wxyz"
10
+ },
11
+ "action": {
12
+ "meaning": "current frame to next frame; last frame is null",
13
+ "dpos": "p[t+1] - p[t], metres in environment axes",
14
+ "dquat": "q[t+1] * inverse(q[t]), shortest relative wxyz rotation, left multiplication",
15
+ "gripper": "actions[t,6]: -1=open, +1=close; null for gaps or unknown commands",
16
+ "execution": "reference only; execution parameters are chosen by the consumer"
17
+ },
18
+ "sampling": {
19
+ "video": "frame_aligned_pre_action",
20
+ "nominal_sample_interval_seconds": 0.05,
21
+ "sampling_gaps": [],
22
+ "unknown_gripper_action_frames": []
23
+ },
24
+ "files": {
25
+ "trajectory": "trajectory.jsonl",
26
+ "videos": {
27
+ "agentview": "videos/agentview.mp4",
28
+ "robot0_eye_in_hand": "videos/wrist.mp4"
29
+ }
30
+ }
31
+ }
LIBERO-GeoMotion/task_000/trajectory.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
LIBERO-GeoMotion/task_000/videos/agentview.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:743c4f1e7aa96489197819479632086cd98b6166d29d710bee20bea318091385
3
+ size 7126750
LIBERO-GeoMotion/task_000/videos/videos.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "sim-reference-videos-v1",
3
+ "videos": {
4
+ "agentview": "agentview.mp4",
5
+ "robot0_eye_in_hand": "wrist.mp4"
6
+ },
7
+ "rendering": {
8
+ "camera_names": [
9
+ "agentview",
10
+ "robot0_eye_in_hand"
11
+ ],
12
+ "camera_resolution": [
13
+ 256,
14
+ 256
15
+ ],
16
+ "image_origin": "top-left",
17
+ "method": "saved_state_render"
18
+ }
19
+ }
LIBERO-GeoMotion/task_000/videos/wrist.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52d6a040c37312a0ae7f8a69da19b2180a26c5da3d6d8e9dc8311f35c6fbd326
3
+ size 22593115
LIBERO-GeoMotion/task_004/steps.txt ADDED
@@ -0,0 +1,22 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ## Expert Procedure
2
+
3
+ Align the gripper with the book and lift it. Rotate it upright, align it with the organizer slot, then lower it, release it, and retreat.
4
+
5
+ ## Action Details
6
+
7
+ | Steps (0-based) | Time (s) | Action | End-effector XYZ at segment end (m) | Gripper command |
8
+ | --- | --- | --- | --- | --- |
9
+ | 0–69 | 0.00–3.50 | Grasp and lift | -0.1933, -0.2043, 1.1204 | Open |
10
+ | 70–136 | 3.50–6.85 | Grasp and lift | -0.1932, -0.2060, 0.9154 | Open |
11
+ | 137–166 | 6.85–8.35 | Grasp and lift | -0.1930, -0.2070, 0.9137 | Close |
12
+ | 167–236 | 8.35–11.85 | Grasp and lift | -0.1940, -0.2056, 1.1172 | Close |
13
+ | 237–354 | 11.85–17.75 | Rotate upright and insert | 0.0197, 0.0644, 1.1801 | Close |
14
+ | 355–374 | 17.75–18.75 | Align using feedback | 0.0212, 0.0660, 1.1186 | Close |
15
+ | 375–389 | 18.75–19.50 | Align using feedback | 0.0248, 0.0634, 1.1005 | Close |
16
+ | 390–404 | 19.50–20.25 | Align using feedback | 0.0240, 0.0644, 1.1010 | Close |
17
+ | 405–424 | 20.25–21.25 | Align using feedback | 0.0213, 0.0663, 1.0392 | Close |
18
+ | 425–439 | 21.25–22.00 | Align using feedback | 0.0246, 0.0633, 1.0044 | Close |
19
+ | 440–454 | 22.00–22.75 | Align using feedback | 0.0242, 0.0642, 1.0050 | Close |
20
+ | 455–489 | 22.75–24.50 | Rotate upright and insert | 0.0239, 0.0644, 1.0050 | Open |
21
+ | 490–548 | 24.50–27.45 | Rotate upright and insert | 0.0241, 0.0644, 1.1828 | Open |
22
+ | 549–588 | 27.45–29.45 | Wait for settling after release | 0.0239, 0.0646, 1.1828 | Open |
LIBERO-GeoMotion/task_004/trajectory.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "libero-codex-trajectory-v1",
3
+ "frames": 589,
4
+ "fps": 20,
5
+ "pose": {
6
+ "reference": "robot0_grip_site",
7
+ "coordinate_frame": "environment",
8
+ "position_unit": "metre",
9
+ "quaternion_order": "wxyz"
10
+ },
11
+ "action": {
12
+ "meaning": "current frame to next frame; last frame is null",
13
+ "dpos": "p[t+1] - p[t], metres in environment axes",
14
+ "dquat": "q[t+1] * inverse(q[t]), shortest relative wxyz rotation, left multiplication",
15
+ "gripper": "actions[t,6]: -1=open, +1=close; null for gaps or unknown commands",
16
+ "execution": "reference only; execution parameters are chosen by the consumer"
17
+ },
18
+ "sampling": {
19
+ "video": "frame_aligned_pre_action",
20
+ "nominal_sample_interval_seconds": 0.05,
21
+ "sampling_gaps": [],
22
+ "unknown_gripper_action_frames": []
23
+ },
24
+ "files": {
25
+ "trajectory": "trajectory.jsonl",
26
+ "videos": {
27
+ "agentview": "videos/agentview.mp4",
28
+ "robot0_eye_in_hand": "videos/wrist.mp4"
29
+ }
30
+ }
31
+ }
LIBERO-GeoMotion/task_004/trajectory.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
LIBERO-GeoMotion/task_004/videos/agentview.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6cac85a0d26ddf3c2a98e203cce97f1b7b0089ca9e79a856376787dd4bf1d7bb
3
+ size 5109743
LIBERO-GeoMotion/task_004/videos/videos.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "sim-reference-videos-v1",
3
+ "videos": {
4
+ "agentview": "agentview.mp4",
5
+ "robot0_eye_in_hand": "wrist.mp4"
6
+ },
7
+ "rendering": {
8
+ "camera_names": [
9
+ "agentview",
10
+ "robot0_eye_in_hand"
11
+ ],
12
+ "camera_resolution": [
13
+ 256,
14
+ 256
15
+ ],
16
+ "image_origin": "top-left",
17
+ "method": "saved_state_render"
18
+ }
19
+ }
LIBERO-GeoMotion/task_004/videos/wrist.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:470a3a05e273b131855431ef1efc5b05ce30bf7fdcdc3919912c1342b6df01b4
3
+ size 11580697
LIBERO-GeoMotion/task_008/steps.txt ADDED
@@ -0,0 +1,26 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ## Expert Procedure
2
+
3
+ Lift the upright bottle and lay it down in a clear area. Regrasp it with the cap facing the rack opening and the body parallel to the long sides. Place it in the rack, release it, and retreat.
4
+
5
+ ## Action Details
6
+
7
+ | Steps (0-based) | Time (s) | Action | End-effector XYZ at segment end (m) | Gripper command |
8
+ | --- | --- | --- | --- | --- |
9
+ | 0–83 | 0.00–4.20 | Grasp and lift | -0.1470, -0.2161, 1.1207 | Open |
10
+ | 84–113 | 4.20–5.70 | Grasp and lift | -0.1478, -0.2161, 1.0339 | Open |
11
+ | 114–143 | 5.70–7.20 | Grasp and lift | -0.1480, -0.2161, 1.0319 | Close |
12
+ | 144–175 | 7.20–8.80 | Grasp and lift | -0.1490, -0.2156, 1.1177 | Close |
13
+ | 176–425 | 8.80–21.30 | Lay down and orient bottle | -0.2923, -0.1429, 1.0177 | Close |
14
+ | 426–465 | 21.30–23.30 | Lay down and orient bottle | -0.2920, -0.1418, 1.0177 | Open |
15
+ | 466–665 | 23.30–33.30 | Lay down and orient bottle | -0.3212, -0.0598, 1.0370 | Open |
16
+ | 666–705 | 33.30–35.30 | Wait for settling after release | -0.3203, -0.0578, 1.0363 | Open |
17
+ | 706–826 | 35.30–41.35 | Lay down and orient bottle | -0.1807, -0.1212, 1.2976 | Open |
18
+ | 827–893 | 41.35–44.70 | Grasp and lift | -0.2492, -0.1231, 1.1217 | Open |
19
+ | 894–955 | 44.70–47.80 | Grasp and lift | -0.2496, -0.1230, 0.9305 | Open |
20
+ | 956–985 | 47.80–49.30 | Grasp and lift | -0.2509, -0.1234, 0.9272 | Close |
21
+ | 986–1053 | 49.30–52.70 | Grasp and lift | -0.2509, -0.1232, 1.1176 | Close |
22
+ | 1054–1175 | 52.70–58.80 | Place and release | -0.0033, 0.1180, 1.1684 | Close |
23
+ | 1176–1250 | 58.80–62.55 | Place and release | 0.0004, 0.1193, 0.9410 | Close |
24
+ | 1251–1280 | 62.55–64.05 | Place and release | -0.0012, 0.1192, 0.9390 | Open |
25
+ | 1281–1356 | 64.05–67.85 | Place and release | -0.0019, 0.1194, 1.1681 | Open |
26
+ | 1357–1396 | 67.85–69.85 | Wait for settling after release | -0.0015, 0.1195, 1.1684 | Open |
LIBERO-GeoMotion/task_008/trajectory.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "libero-codex-trajectory-v1",
3
+ "frames": 1397,
4
+ "fps": 20,
5
+ "pose": {
6
+ "reference": "robot0_grip_site",
7
+ "coordinate_frame": "environment",
8
+ "position_unit": "metre",
9
+ "quaternion_order": "wxyz"
10
+ },
11
+ "action": {
12
+ "meaning": "current frame to next frame; last frame is null",
13
+ "dpos": "p[t+1] - p[t], metres in environment axes",
14
+ "dquat": "q[t+1] * inverse(q[t]), shortest relative wxyz rotation, left multiplication",
15
+ "gripper": "actions[t,6]: -1=open, +1=close; null for gaps or unknown commands",
16
+ "execution": "reference only; execution parameters are chosen by the consumer"
17
+ },
18
+ "sampling": {
19
+ "video": "frame_aligned_pre_action",
20
+ "nominal_sample_interval_seconds": 0.05,
21
+ "sampling_gaps": [],
22
+ "unknown_gripper_action_frames": []
23
+ },
24
+ "files": {
25
+ "trajectory": "trajectory.jsonl",
26
+ "videos": {
27
+ "agentview": "videos/agentview.mp4",
28
+ "robot0_eye_in_hand": "videos/wrist.mp4"
29
+ }
30
+ }
31
+ }
LIBERO-GeoMotion/task_008/trajectory.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
LIBERO-GeoMotion/task_008/videos/agentview.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8ddb1cfffa8904ce6a62f8079a3e211a1489c223f22928c45d4eb119aafa6aa
3
+ size 7936166
LIBERO-GeoMotion/task_008/videos/videos.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "sim-reference-videos-v1",
3
+ "videos": {
4
+ "agentview": "agentview.mp4",
5
+ "robot0_eye_in_hand": "wrist.mp4"
6
+ },
7
+ "rendering": {
8
+ "camera_names": [
9
+ "agentview",
10
+ "robot0_eye_in_hand"
11
+ ],
12
+ "camera_resolution": [
13
+ 256,
14
+ 256
15
+ ],
16
+ "image_origin": "top-left",
17
+ "method": "saved_state_render"
18
+ }
19
+ }
LIBERO-GeoMotion/task_008/videos/wrist.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b267ade440dedb03eda3ccf3dd97178c774858c95a5ef7c45c998323e038c0d
3
+ size 25935683
LIBERO-GeoMotion/task_026/steps.txt ADDED
@@ -0,0 +1,20 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ## Expert Procedure
2
+
3
+ Grasp the mango half and carry it around the vase and cup without disturbing them. Release it on the green pad and retreat.
4
+
5
+ ## Action Details
6
+
7
+ | Steps (0-based) | Time (s) | Action | End-effector XYZ at segment end (m) | Gripper command |
8
+ | --- | --- | --- | --- | --- |
9
+ | 0–75 | 0.00–3.80 | Grasp and lift | -0.1370, -0.1864, 1.1208 | Open |
10
+ | 76–138 | 3.80–6.95 | Grasp and lift | -0.1381, -0.1859, 0.9289 | Open |
11
+ | 139–168 | 6.95–8.45 | Grasp and lift | -0.1384, -0.1862, 0.9266 | Close |
12
+ | 169–236 | 8.45–11.85 | Grasp and lift | -0.1389, -0.1856, 1.1179 | Close |
13
+ | 237–294 | 11.85–14.75 | Carry around obstacles | -0.2690, -0.0819, 1.1717 | Close |
14
+ | 295–400 | 14.75–20.05 | Carry around obstacles | -0.2709, 0.2383, 1.1728 | Close |
15
+ | 401–650 | 20.05–32.55 | Carry around obstacles | 0.0949, 0.2783, 1.1676 | Close |
16
+ | 651–710 | 32.55–35.55 | Place and release | 0.1023, 0.2001, 1.1717 | Close |
17
+ | 711–783 | 35.55–39.20 | Place and release | 0.0978, 0.1995, 0.9352 | Close |
18
+ | 784–813 | 39.20–40.70 | Place and release | 0.1011, 0.2002, 0.9380 | Open |
19
+ | 814–894 | 40.70–44.75 | Place and release | 0.0982, 0.2001, 1.1678 | Open |
20
+ | 895–934 | 44.75–46.75 | Wait for settling after release | 0.0988, 0.2002, 1.1683 | Open |
LIBERO-GeoMotion/task_026/trajectory.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "libero-codex-trajectory-v1",
3
+ "frames": 935,
4
+ "fps": 20,
5
+ "pose": {
6
+ "reference": "robot0_grip_site",
7
+ "coordinate_frame": "environment",
8
+ "position_unit": "metre",
9
+ "quaternion_order": "wxyz"
10
+ },
11
+ "action": {
12
+ "meaning": "current frame to next frame; last frame is null",
13
+ "dpos": "p[t+1] - p[t], metres in environment axes",
14
+ "dquat": "q[t+1] * inverse(q[t]), shortest relative wxyz rotation, left multiplication",
15
+ "gripper": "actions[t,6]: -1=open, +1=close; null for gaps or unknown commands",
16
+ "execution": "reference only; execution parameters are chosen by the consumer"
17
+ },
18
+ "sampling": {
19
+ "video": "frame_aligned_pre_action",
20
+ "nominal_sample_interval_seconds": 0.05,
21
+ "sampling_gaps": [],
22
+ "unknown_gripper_action_frames": []
23
+ },
24
+ "files": {
25
+ "trajectory": "trajectory.jsonl",
26
+ "videos": {
27
+ "agentview": "videos/agentview.mp4",
28
+ "robot0_eye_in_hand": "videos/wrist.mp4"
29
+ }
30
+ }
31
+ }
LIBERO-GeoMotion/task_026/trajectory.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
LIBERO-GeoMotion/task_026/videos/agentview.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4995063116c08353476a26bb481e6b10d3bae039b40b74058e5532063be996cc
3
+ size 6769610
LIBERO-GeoMotion/task_026/videos/videos.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "sim-reference-videos-v1",
3
+ "videos": {
4
+ "agentview": "agentview.mp4",
5
+ "robot0_eye_in_hand": "wrist.mp4"
6
+ },
7
+ "rendering": {
8
+ "camera_names": [
9
+ "agentview",
10
+ "robot0_eye_in_hand"
11
+ ],
12
+ "camera_resolution": [
13
+ 256,
14
+ 256
15
+ ],
16
+ "image_origin": "top-left",
17
+ "method": "saved_state_render"
18
+ }
19
+ }
LIBERO-GeoMotion/task_026/videos/wrist.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54724f1f3bfba67c58a8e2234e048ecd9a4fb0ca56afb8951ad1eb596d196d44
3
+ size 16395946
LIBERO-GeoMotion/task_028/steps.txt ADDED
@@ -0,0 +1,61 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ## Expert Procedure
2
+
3
+ Open the top drawer if needed. Grasp and reorient the cream cheese package, align it with the organizer slot, and place it inside. Release it, retreat, close the drawer, and let it settle.
4
+
5
+ ## Action Details
6
+
7
+ | Steps (0-based) | Time (s) | Action | End-effector XYZ at segment end (m) | Gripper command |
8
+ | --- | --- | --- | --- | --- |
9
+ | 0–106 | 0.00–5.35 | Operate drawer and store object | -0.2237, 0.1523, 1.2036 | Open |
10
+ | 107–152 | 5.35–7.65 | Operate drawer and store object | -0.1458, 0.1530, 1.0859 | Open |
11
+ | 153–174 | 7.65–8.75 | Operate drawer and store object | -0.0859, 0.1528, 1.0838 | Open |
12
+ | 175–214 | 8.75–10.75 | Operate drawer and store object | -0.0836, 0.1527, 1.0844 | Close |
13
+ | 215–266 | 10.75–13.35 | Operate drawer and store object | -0.2363, 0.1525, 1.0849 | Close |
14
+ | 267–301 | 13.35–15.10 | Operate drawer and store object | -0.2364, 0.1526, 1.0849 | Open |
15
+ | 302–461 | 15.10–23.10 | Operate drawer and store object | -0.3283, 0.1700, 1.2274 | Open |
16
+ | 462–570 | 23.10–28.55 | Operate drawer and store object | -0.2001, -0.1189, 1.3199 | Open |
17
+ | 571–646 | 28.55–32.35 | Grasp and lift | -0.1882, -0.2252, 1.1221 | Open |
18
+ | 647–806 | 32.35–40.35 | Grasp and lift | -0.1852, -0.2279, 0.9490 | Open |
19
+ | 807–841 | 40.35–42.10 | Grasp and lift | -0.1849, -0.2279, 0.9490 | Close |
20
+ | 842–900 | 42.10–45.05 | Grasp and lift | -0.1889, -0.2256, 1.1176 | Close |
21
+ | 901–981 | 45.05–49.10 | Operate drawer and store object | -0.1896, -0.2100, 1.0180 | Close |
22
+ | 982–1021 | 49.10–51.10 | Operate drawer and store object | -0.1891, -0.2089, 1.0194 | Open |
23
+ | 1022–1074 | 51.10–53.75 | Operate drawer and store object | -0.1892, -0.2087, 1.1774 | Open |
24
+ | 1075–1177 | 53.75–58.90 | Operate drawer and store object | -0.2002, -0.1190, 1.3197 | Open |
25
+ | 1178–1254 | 58.90–62.75 | Grasp and lift | -0.1873, -0.2036, 1.1203 | Open |
26
+ | 1255–1321 | 62.75–66.10 | Grasp and lift | -0.1879, -0.2031, 0.9143 | Open |
27
+ | 1322–1351 | 66.10–67.60 | Grasp and lift | -0.1878, -0.2033, 0.9117 | Close |
28
+ | 1352–1422 | 67.60–71.15 | Grasp and lift | -0.1883, -0.2025, 1.1173 | Close |
29
+ | 1423–1551 | 71.15–77.60 | Operate drawer and store object | -0.1141, 0.1451, 1.2494 | Close |
30
+ | 1552–1571 | 77.60–78.60 | Align using feedback | -0.1118, 0.1461, 1.1903 | Close |
31
+ | 1572–1591 | 78.60–79.60 | Align using feedback | -0.1139, 0.1456, 1.1245 | Close |
32
+ | 1592–1606 | 79.60–80.35 | Align using feedback | -0.1126, 0.1462, 1.1003 | Close |
33
+ | 1607–1621 | 80.35–81.10 | Align using feedback | -0.1123, 0.1464, 1.1003 | Close |
34
+ | 1622–1636 | 81.10–81.85 | Align using feedback | -0.1123, 0.1466, 1.1003 | Close |
35
+ | 1637–1651 | 81.85–82.60 | Align using feedback | -0.1123, 0.1467, 1.1003 | Close |
36
+ | 1652–1666 | 82.60–83.35 | Align using feedback | -0.1123, 0.1469, 1.1003 | Close |
37
+ | 1667–1681 | 83.35–84.10 | Align using feedback | -0.1123, 0.1470, 1.1003 | Close |
38
+ | 1682–1696 | 84.10–84.85 | Align using feedback | -0.1123, 0.1472, 1.1003 | Close |
39
+ | 1697–1711 | 84.85–85.60 | Align using feedback | -0.1124, 0.1473, 1.1003 | Close |
40
+ | 1712–1726 | 85.60–86.35 | Align using feedback | -0.1124, 0.1474, 1.1003 | Close |
41
+ | 1727–1741 | 86.35–87.10 | Align using feedback | -0.1124, 0.1475, 1.1003 | Close |
42
+ | 1742–1756 | 87.10–87.85 | Align using feedback | -0.1124, 0.1476, 1.1003 | Close |
43
+ | 1757–1771 | 87.85–88.60 | Align using feedback | -0.1124, 0.1477, 1.1003 | Close |
44
+ | 1772–1786 | 88.60–89.35 | Align using feedback | -0.1124, 0.1478, 1.1003 | Close |
45
+ | 1787–1801 | 89.35–90.10 | Align using feedback | -0.1124, 0.1479, 1.1003 | Close |
46
+ | 1802–1816 | 90.10–90.85 | Align using feedback | -0.1124, 0.1479, 1.1003 | Close |
47
+ | 1817–1831 | 90.85–91.60 | Align using feedback | -0.1125, 0.1480, 1.1003 | Close |
48
+ | 1832–1846 | 91.60–92.35 | Align using feedback | -0.1125, 0.1481, 1.1003 | Close |
49
+ | 1847–1861 | 92.35–93.10 | Align using feedback | -0.1125, 0.1482, 1.1003 | Close |
50
+ | 1862–1876 | 93.10–93.85 | Align using feedback | -0.1125, 0.1482, 1.1003 | Close |
51
+ | 1877–1891 | 93.85–94.60 | Align using feedback | -0.1125, 0.1482, 1.1003 | Close |
52
+ | 1892–1906 | 94.60–95.35 | Align using feedback | -0.1125, 0.1483, 1.1003 | Close |
53
+ | 1907–1921 | 95.35–96.10 | Align using feedback | -0.1125, 0.1483, 1.1003 | Close |
54
+ | 1922–1956 | 96.10–97.85 | Operate drawer and store object | -0.1125, 0.1482, 1.1003 | Open |
55
+ | 1957–2019 | 97.85–101.00 | Operate drawer and store object | -0.1129, 0.1483, 1.2878 | Open |
56
+ | 2020–2075 | 101.00–103.80 | Operate drawer and store object | -0.2710, 0.1523, 1.2638 | Close |
57
+ | 2076–2138 | 103.80–106.95 | Operate drawer and store object | -0.2720, 0.1527, 1.0957 | Close |
58
+ | 2139–2588 | 106.95–129.45 | Operate drawer and store object | -0.0852, 0.1500, 1.0673 | Close |
59
+ | 2589–2635 | 129.45–131.80 | Operate drawer and store object | -0.2125, 0.1501, 1.0882 | Close |
60
+ | 2636–2722 | 131.80–136.15 | Operate drawer and store object | -0.2132, 0.1501, 1.2857 | Open |
61
+ | 2723–2762 | 136.15–138.15 | Wait for settling after release | -0.2131, 0.1500, 1.2859 | Open |
LIBERO-GeoMotion/task_028/trajectory.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "libero-codex-trajectory-v1",
3
+ "frames": 2763,
4
+ "fps": 20,
5
+ "pose": {
6
+ "reference": "robot0_grip_site",
7
+ "coordinate_frame": "environment",
8
+ "position_unit": "metre",
9
+ "quaternion_order": "wxyz"
10
+ },
11
+ "action": {
12
+ "meaning": "current frame to next frame; last frame is null",
13
+ "dpos": "p[t+1] - p[t], metres in environment axes",
14
+ "dquat": "q[t+1] * inverse(q[t]), shortest relative wxyz rotation, left multiplication",
15
+ "gripper": "actions[t,6]: -1=open, +1=close; null for gaps or unknown commands",
16
+ "execution": "reference only; execution parameters are chosen by the consumer"
17
+ },
18
+ "sampling": {
19
+ "video": "frame_aligned_pre_action",
20
+ "nominal_sample_interval_seconds": 0.05,
21
+ "sampling_gaps": [],
22
+ "unknown_gripper_action_frames": []
23
+ },
24
+ "files": {
25
+ "trajectory": "trajectory.jsonl",
26
+ "videos": {
27
+ "agentview": "videos/agentview.mp4",
28
+ "robot0_eye_in_hand": "videos/wrist.mp4"
29
+ }
30
+ }
31
+ }
LIBERO-GeoMotion/task_028/trajectory.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
LIBERO-GeoMotion/task_028/videos/agentview.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bf84dcdc3e5f2d8ad498032ac2e204190675ce27eb9328cdf165772291a05f0b
3
+ size 11992187
LIBERO-GeoMotion/task_028/videos/videos.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "sim-reference-videos-v1",
3
+ "videos": {
4
+ "agentview": "agentview.mp4",
5
+ "robot0_eye_in_hand": "wrist.mp4"
6
+ },
7
+ "rendering": {
8
+ "camera_names": [
9
+ "agentview",
10
+ "robot0_eye_in_hand"
11
+ ],
12
+ "camera_resolution": [
13
+ 256,
14
+ 256
15
+ ],
16
+ "image_origin": "top-left",
17
+ "method": "saved_state_render"
18
+ }
19
+ }
LIBERO-GeoMotion/task_028/videos/wrist.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8feb7a60fb37ac8201b3f76179d932afa4dc729dbe266eafe7b01111ec0c7f72
3
+ size 42649890
LIBERO-GeoMotion/task_034/steps.txt ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ## Expert Procedure
2
+
3
+ Grasp and lift each mug in turn. Place the yellow and white mug in the cabinet with its handle facing the back, then place the white mug beside it with its handle facing the opening. Release each mug, retreat, and close the door.
4
+
5
+ ## Action Details
6
+
7
+ | Steps (0-based) | Time (s) | Action | End-effector XYZ at segment end (m) | Gripper command |
8
+ | --- | --- | --- | --- | --- |
9
+ | 0–74 | 0.00–3.75 | Orient and store mug | -0.2004, -0.1386, 1.3482 | Open |
10
+ | 75–175 | 3.75–8.80 | Grasp and lift | -0.2480, -0.3441, 1.1215 | Open |
11
+ | 176–225 | 8.80–11.30 | Grasp and lift | -0.2483, -0.3459, 0.9706 | Open |
12
+ | 226–255 | 11.30–12.80 | Grasp and lift | -0.2480, -0.3462, 0.9690 | Close |
13
+ | 256–309 | 12.80–15.50 | Grasp and lift | -0.2488, -0.3451, 1.1177 | Close |
14
+ | 310–394 | 15.50–19.75 | Orient and store mug | -0.2494, -0.3447, 1.3551 | Close |
15
+ | 395–554 | 19.75–27.75 | Place and release | -0.0044, 0.0516, 1.1684 | Close |
16
+ | 555–610 | 27.75–30.55 | Place and release | -0.0012, 0.0554, 0.9891 | Close |
17
+ | 611–640 | 30.55–32.05 | Place and release | -0.0008, 0.0553, 0.9891 | Open |
18
+ | 641–700 | 32.05–35.05 | Place and release | -0.0004, 0.0561, 1.1679 | Open |
19
+ | 701–808 | 35.05–40.45 | Orient and store mug | -0.1985, -0.1384, 1.3487 | Open |
20
+ | 809–921 | 40.45–46.10 | Grasp and lift | 0.0393, -0.2527, 1.1202 | Open |
21
+ | 922–974 | 46.10–48.75 | Grasp and lift | 0.0419, -0.2537, 0.9677 | Open |
22
+ | 975–1004 | 48.75–50.25 | Grasp and lift | 0.0420, -0.2537, 0.9657 | Close |
23
+ | 1005–1061 | 50.25–53.10 | Grasp and lift | 0.0408, -0.2536, 1.1177 | Close |
24
+ | 1062–1281 | 53.10–64.10 | Orient and store mug | 0.0378, -0.2528, 1.3527 | Close |
25
+ | 1282–1441 | 64.10–72.10 | Place and release | 0.0109, 0.1691, 1.1853 | Close |
26
+ | 1442–1501 | 72.10–75.10 | Place and release | -0.0021, 0.2025, 0.9895 | Close |
27
+ | 1502–1531 | 75.10–76.60 | Place and release | -0.0016, 0.2030, 0.9888 | Open |
28
+ | 1532–1591 | 76.60–79.60 | Place and release | -0.0016, 0.2039, 1.1677 | Open |
29
+ | 1592–1751 | 79.60–87.60 | Close cabinet door | -0.3865, -0.0866, 1.2191 | Close |
30
+ | 1752–1911 | 87.60–95.60 | Close cabinet door | -0.3751, -0.1326, 1.0866 | Close |
31
+ | 1912–1936 | 95.60–96.85 | Close cabinet door | -0.3749, -0.1298, 1.0869 | Close |
32
+ | 1937–1961 | 96.85–98.10 | Close cabinet door | -0.3757, -0.1241, 1.0870 | Close |
33
+ | 1962–1986 | 98.10–99.35 | Close cabinet door | -0.3788, -0.1100, 1.0869 | Close |
34
+ | 1987–2011 | 99.35–100.60 | Close cabinet door | -0.3821, -0.0887, 1.0868 | Close |
35
+ | 2012–2036 | 100.60–101.85 | Close cabinet door | -0.3837, -0.0604, 1.0866 | Close |
36
+ | 2037–2061 | 101.85–103.10 | Close cabinet door | -0.3806, -0.0254, 1.0866 | Close |
37
+ | 2062–2086 | 103.10–104.35 | Close cabinet door | -0.3697, 0.0160, 1.0868 | Close |
38
+ | 2087–2111 | 104.35–105.60 | Close cabinet door | -0.3542, 0.0510, 1.0871 | Close |
39
+ | 2112–2136 | 105.60–106.85 | Close cabinet door | -0.3393, 0.0746, 1.0873 | Close |
40
+ | 2137–2161 | 106.85–108.10 | Close cabinet door | -0.3268, 0.0906, 1.0873 | Close |
41
+ | 2162–2186 | 108.10–109.35 | Close cabinet door | -0.3160, 0.1026, 1.0873 | Close |
42
+ | 2187–2211 | 109.35–110.60 | Close cabinet door | -0.3056, 0.1124, 1.0873 | Close |
43
+ | 2212–2236 | 110.60–111.85 | Close cabinet door | -0.2949, 0.1212, 1.0873 | Close |
44
+ | 2237–2261 | 111.85–113.10 | Close cabinet door | -0.2839, 0.1292, 1.0873 | Close |
45
+ | 2262–2286 | 113.10–114.35 | Close cabinet door | -0.2724, 0.1364, 1.0872 | Close |
46
+ | 2287–2311 | 114.35–115.60 | Close cabinet door | -0.2606, 0.1428, 1.0872 | Close |
47
+ | 2312–2336 | 115.60–116.85 | Close cabinet door | -0.2484, 0.1484, 1.0871 | Close |
48
+ | 2337–2361 | 116.85–118.10 | Close cabinet door | -0.2360, 0.1532, 1.0871 | Close |
49
+ | 2362–2386 | 118.10–119.35 | Close cabinet door | -0.2199, 0.1576, 1.0871 | Close |
50
+ | 2387–2411 | 119.35–120.60 | Close cabinet door | -0.2040, 0.1610, 1.0871 | Close |
51
+ | 2412–2436 | 120.60–121.85 | Close cabinet door | -0.1892, 0.1631, 1.0871 | Close |
52
+ | 2437–2461 | 121.85–123.10 | Close cabinet door | -0.1868, 0.1640, 1.0868 | Close |
53
+ | 2462–2486 | 123.10–124.35 | Close cabinet door | -0.1858, 0.1649, 1.0868 | Close |
54
+ | 2487–2511 | 124.35–125.60 | Close cabinet door | -0.1831, 0.1658, 1.0869 | Close |
55
+ | 2512–2546 | 125.60–127.35 | Close cabinet door | -0.2812, 0.1657, 1.0868 | Close |
56
+ | 2547–2613 | 127.35–130.70 | Close cabinet door | -0.2816, 0.1659, 1.2850 | Open |
57
+ | 2614–2653 | 130.70–132.70 | Wait for settling after release | -0.2813, 0.1659, 1.2853 | Open |
LIBERO-GeoMotion/task_034/trajectory.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "libero-codex-trajectory-v1",
3
+ "frames": 2654,
4
+ "fps": 20,
5
+ "pose": {
6
+ "reference": "robot0_grip_site",
7
+ "coordinate_frame": "environment",
8
+ "position_unit": "metre",
9
+ "quaternion_order": "wxyz"
10
+ },
11
+ "action": {
12
+ "meaning": "current frame to next frame; last frame is null",
13
+ "dpos": "p[t+1] - p[t], metres in environment axes",
14
+ "dquat": "q[t+1] * inverse(q[t]), shortest relative wxyz rotation, left multiplication",
15
+ "gripper": "actions[t,6]: -1=open, +1=close; null for gaps or unknown commands",
16
+ "execution": "reference only; execution parameters are chosen by the consumer"
17
+ },
18
+ "sampling": {
19
+ "video": "frame_aligned_pre_action",
20
+ "nominal_sample_interval_seconds": 0.05,
21
+ "sampling_gaps": [],
22
+ "unknown_gripper_action_frames": []
23
+ },
24
+ "files": {
25
+ "trajectory": "trajectory.jsonl",
26
+ "videos": {
27
+ "agentview": "videos/agentview.mp4",
28
+ "robot0_eye_in_hand": "videos/wrist.mp4"
29
+ }
30
+ }
31
+ }
LIBERO-GeoMotion/task_034/trajectory.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
LIBERO-GeoMotion/task_034/videos/agentview.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bb661949b3b26a9047e3dd5e3838a2f299384d9ec85dc743a87fe9c4e16902bc
3
+ size 14786759
LIBERO-GeoMotion/task_034/videos/videos.json ADDED
@@ -0,0 +1,19 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "sim-reference-videos-v1",
3
+ "videos": {
4
+ "agentview": "agentview.mp4",
5
+ "robot0_eye_in_hand": "wrist.mp4"
6
+ },
7
+ "rendering": {
8
+ "camera_names": [
9
+ "agentview",
10
+ "robot0_eye_in_hand"
11
+ ],
12
+ "camera_resolution": [
13
+ 256,
14
+ 256
15
+ ],
16
+ "image_origin": "top-left",
17
+ "method": "saved_state_render"
18
+ }
19
+ }
LIBERO-GeoMotion/task_034/videos/wrist.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c732ccdcb91b838760103c2b0903a3d5dd349c6153a6128dc607a16de7b1846
3
+ size 48310866
README.md ADDED
@@ -0,0 +1,177 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <h1 align="center">🎬 GPT-GEAR Reference Demonstrations</h1>
2
+
3
+ <p align="center"><strong>🌐 English</strong> | 🌐 <a href="README.zh-CN.md">简体中文</a></p>
4
+
5
+ <p align="center">
6
+ <strong>Explore robot learning from demonstrations with GPT-GEAR</strong><br>
7
+ <sub>91 demonstrations · 3 simulation environments · Video, text, and action trajectories</sub>
8
+ </p>
9
+
10
+ <p align="center">
11
+ 🤖 <a href="https://github.com/6UOOON9/GPT-GEAR">GPT-GEAR</a> &nbsp;·&nbsp;
12
+ 🧭 <a href="#overview">Overview</a> &nbsp;·&nbsp;
13
+ 📦 <a href="#contents">Contents</a> &nbsp;·&nbsp;
14
+ 🚀 <a href="#quick-start">Quick Start</a> &nbsp;·&nbsp;
15
+ 🗺️ <a href="#task-mapping">Task Mapping</a> &nbsp;·&nbsp;
16
+ 💬 <a href="#community">Community</a>
17
+ </p>
18
+
19
+ ---
20
+
21
+ <a id="overview"></a>
22
+
23
+ ## 🧭 Welcome
24
+
25
+ Welcome to the **GPT-GEAR reference demonstration collection**! This package brings together multi-view videos, textual steps, and action trajectories for **RoboDojo, RoboTwin, and LIBERO-GeoMotion**, making it easier to explore how vision-language models use examples to guide robot actions.
26
+
27
+ In GPT-GEAR, demonstrations become part of the model's context: the model can inspect an example, relate it to the current scene, and choose its next action using live observations and feedback. This is **embodied in-context learning (ICL)**. The trajectories are references for the model to read, rather than commands automatically replayed on the robot.
28
+
29
+ Start by watching a demonstration or reading its steps, then try different input combinations in your own evaluations. We welcome reproduction notes, questions, and ideas from the community!
30
+
31
+ <a id="contents"></a>
32
+
33
+ ## 📦 What's included
34
+
35
+ | Environment | Demonstrations | How they are used in GPT-GEAR |
36
+ |---|---:|---|
37
+ | [RoboDojo](RoboDojo/) | 35 | 34 same-name task references in the evaluation queue, plus `spell_RoboDojo` |
38
+ | [RoboTwin](RoboTwin/) | 50 | One per task, shared across `clean` / `random` configurations and evaluation seeds |
39
+ | [LIBERO-GeoMotion](LIBERO-GeoMotion/) | 6 | Shared within task families, covering 24 evaluation scenes |
40
+ | **Total** | **91** | **Video, text, and trajectory files for every demonstration** |
41
+
42
+ Each task directory follows the same layout:
43
+
44
+ ```text
45
+ reference_demo/
46
+ ├── README.md
47
+ ├── README.zh-CN.md
48
+ ├── RoboDojo/<task>/
49
+ ├── RoboTwin/<task>/
50
+ └── LIBERO-GeoMotion/task_<id>/
51
+ ├── steps.txt
52
+ ├── trajectory.json
53
+ ├── trajectory.jsonl
54
+ └── videos/
55
+ ├── videos.json
56
+ └── *.mp4
57
+ ```
58
+
59
+ | File | What it contains | GPT-GEAR reference field |
60
+ |---|---|---|
61
+ | `steps.txt` | Textual steps and demonstration notes | `reference.text` |
62
+ | `trajectory.json` | Trajectory format, metadata, and file references | `reference.trajectory` |
63
+ | `trajectory.jsonl` | Detailed trajectory records, read through the descriptor | Referenced by `trajectory.json` |
64
+ | `videos/` | Camera-view MP4 files and the `videos.json` camera mapping | `reference.video` |
65
+
66
+ Keep each task directory intact so the descriptor can resolve its trajectory and videos. Directory names and task IDs are case-sensitive.
67
+
68
+ RoboDojo and RoboTwin use English numbered steps. LIBERO-GeoMotion retains the original Chinese Markdown expert notes, including recorded coordinates and replay results. Those notes also link to upstream CSV, HDF5, NPY, and combined videos that are not included in this package; the files distributed here are listed above.
69
+
70
+ <a id="quick-start"></a>
71
+
72
+ ## 🚀 Quick start
73
+
74
+ ### 1. Explore a demonstration
75
+
76
+ Try [RoboTwin: adjust a bottle](RoboTwin/adjust_bottle/steps.txt), [RoboDojo: fasten screws](RoboDojo/fasten_screws/steps.txt), or [LIBERO-GeoMotion: move a blue bar onto a green pad](LIBERO-GeoMotion/task_000/steps.txt). Open the corresponding `videos/` directory to view the camera recordings.
77
+
78
+ ### 2. Check the reference files
79
+
80
+ Install the tools following the [GPT-GEAR installation guide](https://github.com/6UOOON9/GPT-GEAR/blob/main/docs/installation.md). With FFmpeg and FFprobe available, run this example from the GPT-GEAR repository root. Replace `/path/to/reference_demo` with the absolute path to this package:
81
+
82
+ ```bash
83
+ gpt-gear reference prepare --benchmark robotwin \
84
+ --reference-video /path/to/reference_demo/RoboTwin/adjust_bottle/videos \
85
+ --reference-text /path/to/reference_demo/RoboTwin/adjust_bottle/steps.txt \
86
+ --trajectory /path/to/reference_demo/RoboTwin/adjust_bottle/trajectory.json \
87
+ --reference-mode sequence --reference-fps 1 \
88
+ --output outputs/robotwin_reference_check/reference
89
+ ```
90
+
91
+ Use a new output directory. This command validates the files and builds a reference cache without starting a simulator or calling a model.
92
+
93
+ ### 3. Add demonstrations to evaluation
94
+
95
+ The GPT-GEAR queue files `configs/queues/robodojo.json`, `configs/queues/robotwin.json`, and `configs/queues/libero_geomotion.json` already contain the task mappings below. Replace their environment paths and `/path/to/reference_demo` placeholders with your local paths before running.
96
+
97
+ **For RoboDojo and LIBERO-GeoMotion with video, text, and full trajectories, increase the context window to `432000`.** Their queue files already set top-level `"codex_context_window": 432000`; retain it when copying a configuration. For a standalone `gpt-gear run`, add `--codex-context-window 432000` explicitly, using a model that supports this window. The default `272000` can trigger context compaction before the full reference is inspected. The offline check above does not need this option.
98
+
99
+ #### Run one RoboDojo task
100
+
101
+ Try `fasten_screws` with video, text, and trajectory references. Prepare the RoboDojo environment and model authentication following the GPT-GEAR guide, then activate the tool environment in the **GPT-GEAR repository root**. Replace every `/path/to/...` placeholder with your local path:
102
+
103
+ ```bash
104
+ gpt-gear run fasten_screws outputs/robodojo_reference \
105
+ --benchmark robodojo \
106
+ --env-root /path/to/RoboDojo \
107
+ --sim-python /path/to/robodojo/bin/python \
108
+ --seed 0 --layout-id 9 \
109
+ --codex-context-window 432000 \
110
+ --reference-video /path/to/reference_demo/RoboDojo/fasten_screws/videos \
111
+ --trajectory /path/to/reference_demo/RoboDojo/fasten_screws/trajectory.json \
112
+ --reference-text /path/to/reference_demo/RoboDojo/fasten_screws/steps.txt \
113
+ --reference-mode sequence --reference-fps 1
114
+ ```
115
+
116
+ This command starts the simulator and calls the model. Use a new output directory; do not create it beforehand. The example uses seed `0` and layout `9`, matching the RoboDojo evaluation queue. After the run, inspect the result with:
117
+
118
+ ```bash
119
+ gpt-gear summarize outputs/robodojo_reference
120
+ ```
121
+
122
+ #### Add references to a queue
123
+
124
+ For a custom queue, add a `reference` object to the matching episode. For example, the `adjust_bottle` episode uses:
125
+
126
+ ```json
127
+ {
128
+ "reference": {
129
+ "video": "/path/to/reference_demo/RoboTwin/adjust_bottle/videos",
130
+ "text": "/path/to/reference_demo/RoboTwin/adjust_bottle/steps.txt",
131
+ "trajectory": "/path/to/reference_demo/RoboTwin/adjust_bottle/trajectory.json",
132
+ "mode": "sequence",
133
+ "fps": 1
134
+ }
135
+ }
136
+ ```
137
+
138
+ This is an episode fragment; see the [queue guide](https://github.com/6UOOON9/GPT-GEAR/blob/main/docs/queues.md) for complete configurations. Use the original files here, even if you have already prepared a cache. Absolute paths make configurations easier to move; relative paths resolve from the JSON configuration's directory.
139
+
140
+ You can supply text, video, or both, and optionally add trajectories alongside text or video. For input combinations and command-line examples, see the [embodied ICL guide](https://github.com/6UOOON9/GPT-GEAR/blob/main/docs/reference-protocol.md).
141
+
142
+ <a id="task-mapping"></a>
143
+
144
+ ## 🗺️ Task mapping
145
+
146
+ ### RoboDojo
147
+
148
+ Task IDs match `RoboDojo/<task>/` directory names. The GPT-GEAR queue contains 42 tasks and uses 34 same-name demonstrations. Its eight tasks without references are `align_blocks`, `classify_objects_by_language`, `general_pickup`, `pick_from_conveyor_by_image`, `pour_by_language`, `solve_equation`, `stack_blocks_by_language`, and `store_tools_in_toolbox`.
149
+
150
+ The additional [spell_RoboDojo demonstration](RoboDojo/spell_RoboDojo/) is included for exploration and is not part of that queue.
151
+
152
+ ### RoboTwin
153
+
154
+ All 50 tasks in the GPT-GEAR queue have a same-name demonstration at `RoboTwin/<task>/`. Each task uses the same demonstration for both `clean` and `random` configurations and all five evaluation seeds (`0–4`).
155
+
156
+ ### LIBERO-GeoMotion
157
+
158
+ One demonstration is shared across each listed task family. Scene IDs use the `libero_rp/` prefix, and ranges below are inclusive.
159
+
160
+ | Evaluation scenes | Task family | Demonstration directory |
161
+ |---|---|---|
162
+ | `libero_rp/0`–`libero_rp/3` | Blue bar onto green pad | [task_000](LIBERO-GeoMotion/task_000/) |
163
+ | `libero_rp/4`–`libero_rp/7` | Book upright in organizer | [task_004](LIBERO-GeoMotion/task_004/) |
164
+ | `libero_rp/8`–`libero_rp/11` | Bottle oriented in rack | [task_008](LIBERO-GeoMotion/task_008/) |
165
+ | `libero_rp/24`–`libero_rp/27` | Mango transfer around protected objects | [task_026](LIBERO-GeoMotion/task_026/) |
166
+ | `libero_rp/28`–`libero_rp/33` | Cream cheese into drawer, then close | [task_028](LIBERO-GeoMotion/task_028/) |
167
+ | `libero_rp/34`–`libero_rp/35` | Two mugs into cabinet, then close | [task_034](LIBERO-GeoMotion/task_034/) |
168
+
169
+ Scenes `libero_rp/12`–`libero_rp/23` have no demonstrations in this package; their queue entries omit reference inputs.
170
+
171
+ <a id="community"></a>
172
+
173
+ ## 💬 Join the community
174
+
175
+ We welcome feedback on the demonstrations, documentation improvements, and experiments with different reference inputs. Share your findings through [GPT-GEAR Issues](https://github.com/6UOOON9/GPT-GEAR/issues). For data or mapping questions, include the environment, task ID, relevant file path, and what you observed so others can reproduce it.
176
+
177
+ For contributions to the framework, see the [contributing guide](https://github.com/6UOOON9/GPT-GEAR/blob/main/CONTRIBUTING.md). Thank you to the RoboDojo, RoboTwin, and LIBERO-GeoMotion communities for their simulation environments and work on robotic manipulation!
README.zh-CN.md ADDED
@@ -0,0 +1,177 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <h1 align="center">🎬 GPT-GEAR 参考示范集</h1>
2
+
3
+ <p align="center">🌐 <a href="README.md">English</a> | <strong>🌐 简体中文</strong></p>
4
+
5
+ <p align="center">
6
+ <strong>与 GPT-GEAR 一起探索机器人如何从示范中学习</strong><br>
7
+ <sub>91 份示范 · 3 个仿真环境 · 视频、文字与动作轨迹</sub>
8
+ </p>
9
+
10
+ <p align="center">
11
+ 🤖 <a href="https://github.com/6UOOON9/GPT-GEAR">GPT-GEAR</a> &nbsp;·&nbsp;
12
+ 🧭 <a href="#overview">项目介绍</a> &nbsp;·&nbsp;
13
+ 📦 <a href="#contents">内容一览</a> &nbsp;·&nbsp;
14
+ 🚀 <a href="#quick-start">快速开始</a> &nbsp;·&nbsp;
15
+ 🗺️ <a href="#task-mapping">任务映射</a> &nbsp;·&nbsp;
16
+ 💬 <a href="#community">参与社区</a>
17
+ </p>
18
+
19
+ ---
20
+
21
+ <a id="overview"></a>
22
+
23
+ ## 🧭 欢迎使用
24
+
25
+ 欢迎来到 **GPT-GEAR 参考示范集**!这里汇集了 **RoboDojo、RoboTwin 和 LIBERO-GeoMotion** 的多视角视频、文字步骤与动作轨迹,方便大家探索视觉语言模型如何借助示范来指导机器人的动作。
26
+
27
+ 在 GPT-GEAR 中,示范会进入模型的上下文:模型可以查看示例、联系当前场景,再结合实时观测和反馈选择下一步动作。这就是**具身上下文学习(Embodied In-Context Learning,ICL)**。轨迹供模型阅读和参考,不会作为指令自动在当前机器人上回放。
28
+
29
+ 你可以先观看一段示范、读一读操作步骤,再把不同输入组合接入自己的评测。欢迎分享复现记录、使用中的问题,以及新的实验想法!
30
+
31
+ <a id="contents"></a>
32
+
33
+ ## 📦 内容一览
34
+
35
+ | 仿真环境 | 示范数量 | 在 GPT-GEAR 中的使用方式 |
36
+ |---|---:|---|
37
+ | [RoboDojo](RoboDojo/) | 35 | 评测队列使用 34 份同名任务示范,另附 `spell_RoboDojo` |
38
+ | [RoboTwin](RoboTwin/) | 50 | 每个任务一份,在 `clean` / `random` 配置及各评测 seed 间共用 |
39
+ | [LIBERO-GeoMotion](LIBERO-GeoMotion/) | 6 | 同一任务族共用一份,覆盖 24 个评测场景 |
40
+ | **合计** | **91** | **每份示范均包含视频、文字与轨迹文件** |
41
+
42
+ 每个任务目录采用统一结构:
43
+
44
+ ```text
45
+ reference_demo/
46
+ ├── README.md
47
+ ├── README.zh-CN.md
48
+ ├── RoboDojo/<task>/
49
+ ├── RoboTwin/<task>/
50
+ └── LIBERO-GeoMotion/task_<id>/
51
+ ├── steps.txt
52
+ ├── trajectory.json
53
+ ├── trajectory.jsonl
54
+ └── videos/
55
+ ├── videos.json
56
+ └── *.mp4
57
+ ```
58
+
59
+ | 文件 | 内容 | GPT-GEAR 参考字段 |
60
+ |---|---|---|
61
+ | `steps.txt` | 文字步骤与示范说明 | `reference.text` |
62
+ | `trajectory.json` | 轨迹格式、元信息及文件引用 | `reference.trajectory` |
63
+ | `trajectory.jsonl` | 逐条轨迹记录,通过描述文件读取 | 由 `trajectory.json` 引用 |
64
+ | `videos/` | 各相机视角的 MP4 视频与 `videos.json` 视角映射 | `reference.video` |
65
+
66
+ 请保留每个任务目录的完整结构,以便描述文件找到对应的轨迹和视频。目录名和任务 ID 区分大小写。
67
+
68
+ RoboDojo 和 RoboTwin 的文字示范采用英文编号步骤。LIBERO-GeoMotion 保留原始中文 Markdown 专家记录,包括实测坐标和回放结果。这些记录还链接了上游 CSV、HDF5、NPY 及合并视频,它们不包含在本示范包中;本包提供的文件见上表。
69
+
70
+ <a id="quick-start"></a>
71
+
72
+ ## 🚀 快速开始
73
+
74
+ ### 1. 浏览一份示范
75
+
76
+ 可以从 [RoboTwin:调整瓶子](RoboTwin/adjust_bottle/steps.txt)、[RoboDojo:拧紧螺丝](RoboDojo/fasten_screws/steps.txt)或 [LIBERO-GeoMotion:把蓝色长条放到绿色垫子上](LIBERO-GeoMotion/task_000/steps.txt)开始。打开对应任务的 `videos/` 目录即可观看各视角录像。
77
+
78
+ ### 2. 检查参考文件
79
+
80
+ 按照 [GPT-GEAR 安装指南](https://github.com/6UOOON9/GPT-GEAR/blob/main/docs/installation.zh-CN.md)安装工具,并确保 FFmpeg 和 FFprobe 可用。在 GPT-GEAR 仓库根目录运行下面的示例,将 `/path/to/reference_demo` 替换为本示范包的绝对路径:
81
+
82
+ ```bash
83
+ gpt-gear reference prepare --benchmark robotwin \
84
+ --reference-video /path/to/reference_demo/RoboTwin/adjust_bottle/videos \
85
+ --reference-text /path/to/reference_demo/RoboTwin/adjust_bottle/steps.txt \
86
+ --trajectory /path/to/reference_demo/RoboTwin/adjust_bottle/trajectory.json \
87
+ --reference-mode sequence --reference-fps 1 \
88
+ --output outputs/robotwin_reference_check/reference
89
+ ```
90
+
91
+ 请使用尚不存在的输出目录。该命令会校验文件并生成参考缓存,不启动仿真器,也不调用模型。
92
+
93
+ ### 3. 将示范接入评测
94
+
95
+ GPT-GEAR 的 `configs/queues/robodojo.json`、`configs/queues/robotwin.json` 和 `configs/queues/libero_geomotion.json` 已包含下文的任务映射。运行前,将其中的环境路径及 `/path/to/reference_demo` 占位符替换为本机路径。
96
+
97
+ **RoboDojo 和 LIBERO-GeoMotion 同时使用视频、文字与完整轨迹时,请将上下文窗口加大到 `432000`。** 两份队列已在 JSON 顶层设置 `"codex_context_window": 432000`,复制配置时请保留。���独执行 `gpt-gear run` 时,需要显式添加 `--codex-context-window 432000`,并使用支持该窗口大小的模型。默认值 `272000` 可能在模型完整阅读示范前触发上下文压缩。上面的离线校验命令无需添加此参数。
98
+
99
+ #### 运行一个 RoboDojo 任务
100
+
101
+ 可以先用 `fasten_screws`(拧紧螺丝)尝试视频、文字与轨迹三种示范输入。按 GPT-GEAR 指南准备好 RoboDojo 环境并完成模型认证,在 **GPT-GEAR 仓库根目录**激活工具环境,将所有 `/path/to/...` 占位符替换为本机路径:
102
+
103
+ ```bash
104
+ gpt-gear run fasten_screws outputs/robodojo_reference \
105
+ --benchmark robodojo \
106
+ --env-root /path/to/RoboDojo \
107
+ --sim-python /path/to/robodojo/bin/python \
108
+ --seed 0 --layout-id 9 \
109
+ --codex-context-window 432000 \
110
+ --reference-video /path/to/reference_demo/RoboDojo/fasten_screws/videos \
111
+ --trajectory /path/to/reference_demo/RoboDojo/fasten_screws/trajectory.json \
112
+ --reference-text /path/to/reference_demo/RoboDojo/fasten_screws/steps.txt \
113
+ --reference-mode sequence --reference-fps 1
114
+ ```
115
+
116
+ 该命令会启动仿真器并调用模型。输出目录必须尚不存在,无需提前创建。示例使用 seed `0`、layout `9`,与 RoboDojo 评测队列一致。运行结束后,可查看结果:
117
+
118
+ ```bash
119
+ gpt-gear summarize outputs/robodojo_reference
120
+ ```
121
+
122
+ #### 在队列中加入示范
123
+
124
+ 自定义队列时,在对应 episode 中加入 `reference` 对象。例如,`adjust_bottle` 的 episode 使用:
125
+
126
+ ```json
127
+ {
128
+ "reference": {
129
+ "video": "/path/to/reference_demo/RoboTwin/adjust_bottle/videos",
130
+ "text": "/path/to/reference_demo/RoboTwin/adjust_bottle/steps.txt",
131
+ "trajectory": "/path/to/reference_demo/RoboTwin/adjust_bottle/trajectory.json",
132
+ "mode": "sequence",
133
+ "fps": 1
134
+ }
135
+ }
136
+ ```
137
+
138
+ 上面是 episode 的配置片段;完整配置请参考[队列指南](https://github.com/6UOOON9/GPT-GEAR/blob/main/docs/queues.zh-CN.md)。即使已经生成缓存,这里仍应填写原始文件路径。建议使用绝对路径,方便移动配置;相对路径以 JSON 配置文件所在目录为基准解析。
139
+
140
+ 可以提供文字、视频或两者组合,也可以在文字或视频之外加入轨迹。输入组合和命令行用法见[具身上下文学习指南](https://github.com/6UOOON9/GPT-GEAR/blob/main/docs/reference-protocol.zh-CN.md)。
141
+
142
+ <a id="task-mapping"></a>
143
+
144
+ ## 🗺️ 任务映射
145
+
146
+ ### RoboDojo
147
+
148
+ 任务 ID 与 `RoboDojo/<task>/` 的目录名一致。GPT-GEAR 队列包含 42 个任务,其中 34 个使用同名示范。其余 8 个任务不使用参考输入:`align_blocks`、`classify_objects_by_language`、`general_pickup`、`pick_from_conveyor_by_image`、`pour_by_language`、`solve_equation`、`stack_blocks_by_language` 和 `store_tools_in_toolbox`。
149
+
150
+ 本包还提供了 [spell_RoboDojo 示范](RoboDojo/spell_RoboDojo/),可供探索,但它不在该评测队列中。
151
+
152
+ ### RoboTwin
153
+
154
+ GPT-GEAR 队列中的 50 个任务均在 `RoboTwin/<task>/` 下有同名示范。每个任务在 `clean` 和 `random` 两种配置及全部 5 个评测 seed(`0–4`)中共用同一份示范。
155
+
156
+ ### LIBERO-GeoMotion
157
+
158
+ 下列每个任务族共用一份示范。场景 ID 使用 `libero_rp/` 前缀,表中的范围均包含两端。
159
+
160
+ | 评测场景 | 任务族 | 示范目录 |
161
+ |---|---|---|
162
+ | `libero_rp/0`–`libero_rp/3` | 把蓝色长条放到绿色垫子上 | [task_000](LIBERO-GeoMotion/task_000/) |
163
+ | `libero_rp/4`–`libero_rp/7` | 将书本竖直放入收纳架 | [task_004](LIBERO-GeoMotion/task_004/) |
164
+ | `libero_rp/8`–`libero_rp/11` | 调整瓶子朝向并放入架子 | [task_008](LIBERO-GeoMotion/task_008/) |
165
+ | `libero_rp/24`–`libero_rp/27` | 移动芒果并避让需保护的物体 | [task_026](LIBERO-GeoMotion/task_026/) |
166
+ | `libero_rp/28`–`libero_rp/33` | 将奶油奶酪放入抽屉并关上抽屉 | [task_028](LIBERO-GeoMotion/task_028/) |
167
+ | `libero_rp/34`–`libero_rp/35` | 将两个杯子放入柜子并关门 | [task_034](LIBERO-GeoMotion/task_034/) |
168
+
169
+ 本包未提供 `libero_rp/12`–`libero_rp/23` 的示范,队列中的对应条目不设置参考输入。
170
+
171
+ <a id="community"></a>
172
+
173
+ ## 💬 参与社区
174
+
175
+ 欢迎反馈示范内容、改进文档,也欢迎分享不同参考输入组合下的实验结果。你可以通过 [GPT-GEAR Issues](https://github.com/6UOOON9/GPT-GEAR/issues)参与讨论。反馈数据或任务映射问题时,请附上环境名称、任务 ID、相关文件路径和实际观察,方便大家复现。
176
+
177
+ 如果希望参与框架开发,请参考[贡献指南](https://github.com/6UOOON9/GPT-GEAR/blob/main/CONTRIBUTING.zh-CN.md)。感谢 RoboDojo、RoboTwin 和 LIBERO-GeoMotion 社区提供的仿真环境,以及在机器人操作研究中的投入!
RoboDojo/arrange_largest_number/steps.txt ADDED
@@ -0,0 +1,5 @@
 
 
 
 
 
 
1
+ 1. Identify the four red number pieces and the four circular pads; the final visible order is 6, 5, 3, 1 from left to right in the head view.
2
+ 2. Pick up the 6 from the left portion of the table, lift it onto the leftmost pad, and release it; check that it remains on the pad as the gripper withdraws.
3
+ 3. Pick up the 5 and place it on the second pad, leaving the first number undisturbed.
4
+ 4. Pick up the 3 with the other gripper and place it on the third pad; check its orientation and separation from the neighboring number.
5
+ 5. Move the 1 onto the rightmost pad, release it, and retract both arms. Check that all four pieces remain visible on their pads in the demonstrated order.
RoboDojo/arrange_largest_number/trajectory.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "robodojo-codex-trajectory-v1",
3
+ "frames": 374,
4
+ "fps": 25,
5
+ "pose": {
6
+ "reference": {
7
+ "left": "left_ee",
8
+ "right": "right_ee"
9
+ },
10
+ "coordinate_frame": "environment",
11
+ "position_unit": "metre",
12
+ "quaternion_order": "wxyz"
13
+ },
14
+ "action": {
15
+ "meaning": "current frame to next frame; last frame is null",
16
+ "dpos": "p[t+1] - p[t], metres in environment axes",
17
+ "dquat": "q[t+1] * inverse(q[t]), shortest relative wxyz rotation, left multiplication",
18
+ "gripper": "per-arm target: 0=closed, 1=open, [0,1]; null=unknown, not hold",
19
+ "execution": "reference only; execution parameters are chosen by the consumer"
20
+ },
21
+ "sampling": {
22
+ "nominal_sample_interval_seconds": 0.04,
23
+ "sampling_gaps": [],
24
+ "unknown_gripper_action_frames": {
25
+ "left": [],
26
+ "right": []
27
+ }
28
+ },
29
+ "files": {
30
+ "trajectory": "trajectory.jsonl",
31
+ "videos": {
32
+ "head": "videos/head.mp4",
33
+ "left_wrist": "videos/left_wrist.mp4",
34
+ "right_wrist": "videos/right_wrist.mp4"
35
+ }
36
+ }
37
+ }
RoboDojo/arrange_largest_number/trajectory.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
RoboDojo/arrange_largest_number/videos/head.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41d71b437c41a653b68cb23f1621ffd5e70bfea86909cf2f8ee2c468754f4f70
3
+ size 2057531
RoboDojo/arrange_largest_number/videos/left_wrist.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dcd821a1b4cc231bcd584fc79d1ae4ead983d3f255b07fe9836cbcc15a203a7e
3
+ size 2291821
RoboDojo/arrange_largest_number/videos/right_wrist.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d52da153a342cddaab978c4eecef6a8bac6eefbab2277cd0a7e82bc845590412
3
+ size 2370884
RoboDojo/arrange_largest_number/videos/videos.json ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "robodojo-reference-videos-v1",
3
+ "videos": {
4
+ "head": "head.mp4",
5
+ "left_wrist": "left_wrist.mp4",
6
+ "right_wrist": "right_wrist.mp4"
7
+ }
8
+ }
RoboDojo/build_tower/steps.txt ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ 1. Locate the four white support blocks, two flat boards, small wooden top block, and green roof-shaped piece.
2
+ 2. Place one white support on the left and another on the right of the intended base, leaving a gap for the long board to span.
3
+ 3. Grasp the long board at both ends, lift it level, and lower it across the two supports. Release and check that the board remains supported.
4
+ 4. Place the remaining two white blocks on top of the first board, one toward each side; avoid displacing the lower supports.
5
+ 5. Grasp the second board with both grippers and place it across the upper pair of blocks. Withdraw and check that the two-level structure remains upright.
6
+ 6. Pick up the small wooden block and place it near the center of the upper board, then place the green roof-shaped piece on top of it.
7
+ 7. Retract both arms and inspect the completed visible tower for shifted supports or falling parts.
RoboDojo/build_tower/trajectory.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "robodojo-codex-trajectory-v1",
3
+ "frames": 710,
4
+ "fps": 25,
5
+ "pose": {
6
+ "reference": {
7
+ "left": "left_ee",
8
+ "right": "right_ee"
9
+ },
10
+ "coordinate_frame": "environment",
11
+ "position_unit": "metre",
12
+ "quaternion_order": "wxyz"
13
+ },
14
+ "action": {
15
+ "meaning": "current frame to next frame; last frame is null",
16
+ "dpos": "p[t+1] - p[t], metres in environment axes",
17
+ "dquat": "q[t+1] * inverse(q[t]), shortest relative wxyz rotation, left multiplication",
18
+ "gripper": "per-arm target: 0=closed, 1=open, [0,1]; null=unknown, not hold",
19
+ "execution": "reference only; execution parameters are chosen by the consumer"
20
+ },
21
+ "sampling": {
22
+ "nominal_sample_interval_seconds": 0.04,
23
+ "sampling_gaps": [],
24
+ "unknown_gripper_action_frames": {
25
+ "left": [],
26
+ "right": []
27
+ }
28
+ },
29
+ "files": {
30
+ "trajectory": "trajectory.jsonl",
31
+ "videos": {
32
+ "head": "videos/head.mp4",
33
+ "left_wrist": "videos/left_wrist.mp4",
34
+ "right_wrist": "videos/right_wrist.mp4"
35
+ }
36
+ }
37
+ }
RoboDojo/build_tower/trajectory.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
RoboDojo/build_tower/videos/head.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d1ed3ada31a0915b0ce4263fc508a843f4ff384f04e403eb79a91902603f92b0
3
+ size 4190197
RoboDojo/build_tower/videos/left_wrist.mp4 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cfe4e45950c05489dc42ecb4e5fa3834a6611624f21b29de1fd165fb0cf82a20
3
+ size 4685008