Update dataset README with full docs

README.md

@@ -1,3 +1,180 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+task_categories:
+  - robotics
+tags:
+  - robotics
+  - manipulation
+  - imitation-learning
+  - elevator-button-press
+  - isaac-sim
+  - isaac-lab
+  - omy
+pretty_name: ev Button Push (OMY call-button, Isaac Lab)
+size_categories:
+  - 10B<n<100B
+---
+
+# evButtonPush
+
+Simulated demonstrations of a **ROBOTIS OMY** 6-DoF arm pressing an elevator
+call-button, collected in **NVIDIA Isaac Sim / Isaac Lab** with a scripted
+IK teacher.
+
+Two versions are provided:
+
+| Folder | Teacher success criterion | Key properties |
+|---|---|---|
+| **`v01/`** | Distance-based (tip within 5 cm of button) | 46 demos; mixed press-target height; no elevator interaction |
+| **`v02/`** | **Physical contact force ≥ 0.5 N** | 45 demos; uniform Z+0.11 m target; elevator door FSM; **`call_button_lit` LED observation** |
+
+For imitation learning, **v02 is the recommended training set**. v01 is a
+baseline / ablation reference from the same robot and cameras.
+
+---
+
+## Task
+
+- Task ID (Isaac Lab): `RobotisLab-CallButton-Right-OMY-v0`
+- Robot: ROBOTIS **OMY** (6-DoF arm + 2-finger gripper)
+- Pedestal: 0.908 m, gripper always closed (stiffness 2 × 10⁶, effort 5000)
+- Elevator USD: random choice from 100 variants (`elevator_setup_new`)
+- Robot initial pose randomization (per-seed):
+  - standoff distance to button: 0.55 ~ 0.65 m
+  - lateral offset: −0.05 ~ 0.05 m
+  - yaw: 0°
+
+---
+
+## Cameras
+
+All three cameras record **RGB, 1920 × 1200, Pinhole**:
+- horizontal aperture 20.955 mm
+- vertical aperture 13.097 mm (derived from 16:10 aspect)
+- clipping (0.01, 100) m
+
+| Name | Prim path | Frame | Position (m) | Quat (w, x, y, z) | FOV (H, V, D) | focal |
+|---|---|---|---|---|---|---|
+| `cam_wrist` | `Robot/OMY/link6/cam_wrist` | link6 (local, moves with arm) | (0.000, −0.080, 0.070) | (0.0018, −0.0018, 0.7071, 0.7071) | 83.21° / 58.06° / 92.64° | 11.8 mm |
+| `cam_top` | `Robot/OMY/world/cam_top` | robot base | (0.000233, 0.1499, 0.43953) | (0.5144, 0.4146, −0.5064, −0.5542) | 92.67° / 66.44° / 102.03° | 10.0 mm |
+| `cam_belly` | `Robot/OMY/world/cam_belly` | robot base | (0.08705, 0.1499, 0.1075) | (−0.4901, −0.5097, 0.5097, 0.4901) | 92.67° / 66.44° / 102.03° | 10.0 mm |
+
+The `world` prim inside the OMY USD refers to the robot's base frame (link0),
+**not** the simulation world. To obtain world-frame poses at runtime:
+
+```
+world_pose(cam_top)    = robot.root_pose_w  ⊙  offset_cam_top
+world_pose(cam_wrist)  = robot.root_pose_w  ⊙  FK(link0→link6)  ⊙  offset_cam_wrist
+```
+
+---
+
+## v02 — `v02/` (recommended)
+
+- **45 demos**, seeds `0` – `44`, all pressed = True
+- ~35 s per episode, ~1050 frames at 30 fps
+- Elevator **door FSM** progresses on press:
+  `state 0 (idle) → 1 (pressed / LED on) → 2 (door opening, progress 0→1) → 3 (open)`
+- Average peak gripper tip contact force: 60 – 80 N
+- Teacher CLI:
+  ```
+  --approach-steps 100   --press-steps 400
+  --stall-steps 20       --hold-steps 30
+  --return-steps 300     --home-idle-steps 150
+  --press-depth 0.035    --contact-force-threshold 0.5
+  ```
+- Press target Z offset: `+0.11 m` relative to `button_base` (fixed inside teacher).
+
+### HDF5 layout per episode
+
+```
+data/demo_0/
+├── actions              (T, 7)    float32   arm 6 + gripper 1
+├── processed_actions    (T, 10)   float32   full joint targets
+├── obs/
+│   ├── cam_wrist        (T, 1200, 1920, 3)   uint8
+│   ├── cam_top          (T, 1200, 1920, 3)   uint8
+│   ├── cam_belly        (T, 1200, 1920, 3)   uint8
+│   ├── joint_pos                (T, 10)     float32
+│   ├── joint_vel                (T, 10)     float32
+│   ├── call_button_pos          (T, 3)      float32   button world pos
+│   ├── rel_ee_call_button_distance (T, 3)   float32   tip-button vec
+│   └── call_button_lit          (T, 1)      bool      ⭐ LED on/off
+├── states/articulation/robot/...
+├── states/rigid_object/pedestal/...
+└── initial_state/...
+```
+
+`call_button_lit` is specific to v02 and lets a policy learn an explicit
+"button is now pressed → door is opening → retract" cue.
+
+---
+
+## v01 — `v01/`
+
+- **46 demos**, seeds span `0`–`60` with gaps (see `v01/metadata.csv`)
+- Older teacher that judged success purely by tip-to-button distance
+  (threshold 0.050 m) — no physical contact check
+- **Press-target Z offset is NOT uniform** across demos:
+  - `z = 0.00 m`: seeds 0–27 (teacher's default height)
+  - `z = +0.01 m`: seeds 50–54
+  - `z = +0.04 m`: seeds 55–60
+  - `z = +0.11 m`: seeds 23–32 (matched to v02 for a cleaner subset)
+- No `call_button_lit` observation, no elevator door motion.
+
+The height heterogeneity is intentional — v01 is useful for studying how a
+student policy reacts to inconsistent teacher behavior, but for a clean
+baseline a user may want to filter to a single `z_offset_m` group
+(metadata is in `v01/metadata.csv`).
+
+---
+
+## Per-version files
+
+Each version directory contains:
+
+- `demo_XX.hdf5` — one episode per file (see layout above)
+- `demo_XX.log` — scripted-teacher stdout with `[SETUP]`, `[PHASE]`,
+  `[APPROACH DONE]`, `[PRESSED]`, `[RESULT]` markers
+- `summary.log` — one line per demo
+- `metadata.csv` — `seed, z_offset_m, size_bytes, pressed, min_distance_m, teacher_version`
+- `README.md` — version-specific notes
+
+---
+
+## Reproducibility
+
+- Isaac Sim **5.1**, Isaac Lab (`robotis_lab` package)
+- DataCrunch Cloud, NVIDIA RTX PRO 6000 Blackwell
+- `v02` additionally depends on the `frontier_simulation` package
+  (`isaac_sim.assets.elevator.ELEVATOR_CFG`) for the door FSM.
+
+Scripted teacher: `diffik_teacher_jointpos.py` (ports an open-loop IK
+press policy into Isaac Lab's env loop and records to HDF5 via
+`ActionStateRecorderManagerCfg`).
+
+---
+
+## Suggested usage
+
+Imitation learning (ACT, Diffusion Policy, BC) with `qpos_dim = 8`:
+
+```python
+qpos = concat(joint_pos[:6], gripper_closed_flag, call_button_lit)  # 8
+action = [arm_delta[:6], gripper_cmd]                                 # 7
+images = {cam_wrist, cam_top, cam_belly}                              # 1920×1200 (resize as needed)
+```
+
+For efficiency, training pipelines commonly resize to 240 × 320 before
+consumption. We retain the original 1920 × 1200 to keep the data
+lossless.
+
+---
+
+## Authors / Contact
+
+- Collected by: **Hwang Yeeun**
+- Organization: **ROBOTIS** (AI team)
+- Simulator: NVIDIA Isaac Sim 5.1 + Isaac Lab
+
+For issues or pull requests, please open them on the dataset repo.