aya-s commited on
Commit
77d7aba
·
verified ·
1 Parent(s): 35c8ca0

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +103 -0
README.md ADDED
@@ -0,0 +1,103 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ language:
4
+ - en
5
+ ---
6
+
7
+ <table>
8
+ <tr>
9
+ <th width="30%" style="font-size: 30px;">Task Overview</th>
10
+ <th width="70%" style="font-size: 30px;">Data Viewer</th>
11
+ </tr>
12
+ <tr>
13
+ <td style="vertical-align: middle; width:30%;">
14
+ <video width="100%" controls>
15
+ <source src="https://huggingface.co/datasets/ko-br/bulb_in_socket/resolve/main/videos/bulb_in_socket_video.mp4" type="video/mp4">
16
+ </video>
17
+ </td>
18
+ <td style="vertical-align: middle; width:70%;">
19
+ <video width="100%" controls>
20
+ <source src="https://huggingface.co/datasets/ko-br/bulb_in_socket/resolve/main/videos/data_viewer_one_episode.mp4" type="video/mp4">
21
+ </video>
22
+ </td>
23
+ </tr>
24
+ </table>
25
+
26
+ # About
27
+
28
+ The episodes in this dataset are pick and place sequences.
29
+ A light bulb is picked up from a random position on the wooden tabletop and placed in the socket, also on the wooden tabletop.
30
+
31
+
32
+
33
+ ## Dataset summary
34
+
35
+ | Field | Value |
36
+ |---|---|
37
+ | Robot | UR10e (6-DoF) |
38
+ | Gripper | Robotiq Hand-E |
39
+ | Cameras | Intel Realsense 435 |
40
+ | Episodes | `23` |
41
+ | Frames | `15,548` |
42
+ | FPS | `30` |
43
+ | Tasks / language instructions | `pick up the bulb and place it in the socket` |
44
+
45
+
46
+
47
+ ## Feature dictionary (all columns)
48
+
49
+ Shapes below match `meta/info.json`. LeRobot standard index columns (`timestamp`, `frame_index`, `episode_index`, `index`, `task_index`) are also present.
50
+
51
+ ### Observations
52
+
53
+ | Feature | Dtype / shape | Names / semantics |
54
+ |---|---|---|
55
+ | `observation.state` | float32 `[6]` | Joint positions: Base, Shoulder, Elbow, Wrist 1, Wrist 2, Wrist 3 (rad) |
56
+ | `observation.ee_pose` | float32 `[6]` | TCP pose `[x, y, z, rx, ry, rz]` (m / UR rotation vector) |
57
+ | `observation.ee_force_torque` | float32 `[6]` | TCP wrench `[Fx, Fy, Fz, Tx, Ty, Tz]` |
58
+ | `observation.gripper_state` | float32 `[1]` | Measured aperture `POS/255` in `[0, 1]` (0 open, 1 closed) |
59
+ | `observation.gripper_current` | float32 `[1]` | Motor current (A) ≈ `COU * 0.01` |
60
+ | `observation.gripper_object` | float32 `[1]` | Hand-E `OBJ`: 0 moving, 1 outer stop, 2 inner stop, 3 at request |
61
+ | `observation.gripper_status` | float32 `[1]` | Hand-E `STA`: 0 reset, 1 activating, 3 active |
62
+ | `observation.gripper_fault` | float32 `[1]` | Hand-E `FLT` code (`0` = none) |
63
+ | `observation.gripper_position_request_echo` | float32 `[1]` | `PRE/255` last commanded position echo |
64
+ | `observation.gripper_speed_request_echo` | float32 `[1]` | `SPE/255` last commanded speed echo |
65
+ | `observation.gripper_force_request_echo` | float32 `[1]` | `FOR/255` last commanded force echo |
66
+
67
+ ### Actions
68
+
69
+ | Feature | Dtype / shape | Names / semantics |
70
+ |---|---|---|
71
+ | `action` | float32 `[9]` | See action layout below |
72
+ | `action.ee_pose_next` | float32 `[6]` | Observed TCP at `t+1` (last frame repeats `t`); filled on assemble/export |
73
+ | `action.ee_pose_delta` | float32 `[6]` | Observed `ee_pose[t+1] - ee_pose[t]` (last frame zeros); filled on assemble/export |
74
+
75
+ **`action` layout (always 9-D):**
76
+
77
+ | Index | `joint_positions` names | `ee_pose_delta` names | Source |
78
+ |---|---|---|---|
79
+ | 0…5 | Base…Wrist 3 | x…rz | Sticky commanded `cmd_joint_positions` **or** `cmd_ee_pose_delta` (export-time choice) |
80
+ | 6 | `gripper` | `gripper` | Commanded `cmd_gripper` (`SET POS`), recorded at **send** time |
81
+ | 7 | `gripper_speed` | `gripper_speed` | Commanded `cmd_gripper_speed` (`SET SPE`), send time |
82
+ | 8 | `gripper_force` | `gripper_force` | Commanded `cmd_gripper_force` (`SET FOR`), send time |
83
+
84
+ Until the first grip command in an episode, indices 6–8 are **0**. They are **not** filled from observed gripper state.
85
+
86
+ Look-ahead features (`action.ee_pose_*`) are derived from observed TCP and are the intended arm supervision for freemotion / freedrive demos when commanded arm channels are empty.
87
+
88
+ ### Cameras (RGB + depth)
89
+
90
+ Present when the corresponding feed was recorded. Depth keys use LeRobot native `*_depth` video features.
91
+
92
+ | Feature | Modality |
93
+ |---|---|
94
+ | `observation.images.workspace_front` | Wrist Camera, RGB video |
95
+ | `observation.images.workspace_side` | Workspace Camera, RGB video |
96
+ | `observation.images.workspace_front_depth` | Wrist Camera, Depth video |
97
+ | `observation.images.workspace_side_depth` | Workspace Camera, Depth video |
98
+
99
+
100
+ ## Maintainers
101
+
102
+ - Contact: [aya@ko-br.com](mailto:aya@ko-br.com), [info@ko-br.com](mailto:info@ko-br.com)
103
+