File size: 4,357 Bytes
79b2ec3 13f8e1a 9a65400 e24376b 44d97e2 13f8e1a 44d97e2 e24376b a12bdf3 13f8e1a 79b2ec3 96f2ec4 79b2ec3 de1f20f 79b2ec3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 | ---
license: cc-by-4.0
language:
- en
---
<table>
<tr>
<th width="30%" style="font-size: 30px;">Task Overview</th>
<th width="70%" style="font-size: 30px;">Data Viewer</th>
</tr>
<tr>
<td style="vertical-align: middle; width:30%;">
<video width="100%" controls>
<source src="https://huggingface.co/datasets/ko-br/bulb_in_socket/resolve/main/videos/bulb_in_socket_video.mp4" type="video/mp4">
</video>
</td>
<td style="vertical-align: middle; width:70%;">
<video width="100%" controls>
<source src="https://huggingface.co/datasets/ko-br/bulb_in_socket/resolve/main/videos/data_viewer_one_episode.mp4" type="video/mp4">
</video>
</td>
</tr>
</table>
# About
The episodes in this dataset are pick and place sequences.
A light bulb is picked up from a random position on the wooden tabletop and placed in the socket, also on the wooden tabletop.
## Dataset summary
| Field | Value |
|---|---|
| Robot | UR10e (6-DoF) |
| Gripper | Robotiq Hand-E |
| Cameras | Intel Realsense 435 |
| Episodes | `23` |
| Frames | `15,548` |
| FPS | `30` |
| Tasks / language instructions | `pick up the bulb and place it in the socket` |
## Feature dictionary (all columns)
Shapes below match `meta/info.json`. LeRobot standard index columns (`timestamp`, `frame_index`, `episode_index`, `index`, `task_index`) are also present.
### Observations
| Feature | Dtype / shape | Names / semantics |
|---|---|---|
| `observation.state` | float32 `[6]` | Joint positions: Base, Shoulder, Elbow, Wrist 1, Wrist 2, Wrist 3 (rad) |
| `observation.ee_pose` | float32 `[6]` | TCP pose `[x, y, z, rx, ry, rz]` (m / UR rotation vector) |
| `observation.ee_force_torque` | float32 `[6]` | TCP wrench `[Fx, Fy, Fz, Tx, Ty, Tz]` |
| `observation.gripper_state` | float32 `[1]` | Measured aperture `POS/255` in `[0, 1]` (0 open, 1 closed) |
| `observation.gripper_current` | float32 `[1]` | Motor current (A) ≈ `COU * 0.01` |
| `observation.gripper_object` | float32 `[1]` | Hand-E `OBJ`: 0 moving, 1 outer stop, 2 inner stop, 3 at request |
| `observation.gripper_status` | float32 `[1]` | Hand-E `STA`: 0 reset, 1 activating, 3 active |
| `observation.gripper_fault` | float32 `[1]` | Hand-E `FLT` code (`0` = none) |
| `observation.gripper_position_request_echo` | float32 `[1]` | `PRE/255` last commanded position echo |
| `observation.gripper_speed_request_echo` | float32 `[1]` | `SPE/255` last commanded speed echo |
| `observation.gripper_force_request_echo` | float32 `[1]` | `FOR/255` last commanded force echo |
### Actions
| Feature | Dtype / shape | Names / semantics |
|---|---|---|
| `action` | float32 `[9]` | See action layout below |
| `action.ee_pose_next` | float32 `[6]` | Observed TCP at `t+1` (last frame repeats `t`); filled on assemble/export |
| `action.ee_pose_delta` | float32 `[6]` | Observed `ee_pose[t+1] - ee_pose[t]` (last frame zeros); filled on assemble/export |
**`action` layout (always 9-D):**
| Index | `joint_positions` names | `ee_pose_delta` names | Source |
|---|---|---|---|
| 0…5 | Base…Wrist 3 | x…rz | Sticky commanded `cmd_joint_positions` **or** `cmd_ee_pose_delta` (export-time choice) |
| 6 | `gripper` | `gripper` | Commanded `cmd_gripper` (`SET POS`), recorded at **send** time |
| 7 | `gripper_speed` | `gripper_speed` | Commanded `cmd_gripper_speed` (`SET SPE`), send time |
| 8 | `gripper_force` | `gripper_force` | Commanded `cmd_gripper_force` (`SET FOR`), send time |
Until the first grip command in an episode, indices 6–8 are **0**. They are **not** filled from observed gripper state.
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.
### Cameras (RGB + depth)
Present when the corresponding feed was recorded. Depth keys use LeRobot native `*_depth` video features.
| Feature | Modality |
|---|---|
| `observation.images.workspace_front` | Wrist Camera, RGB video |
| `observation.images.workspace_side` | Workspace Camera, RGB video |
| `observation.images.workspace_front_depth` | Wrist Camera, Depth video |
| `observation.images.workspace_side_depth` | Workspace Camera, Depth video |
## Maintainers
- Contact: [aya@ko-br.com](mailto:aya@ko-br.com), [info@ko-br.com](mailto:info@ko-br.com)
|