Restore dataset card with images and add V1 Force Reduction results

README.md

@@ -9,70 +9,184 @@ tags:
   - insertion
   - lerobot
   - isaac-lab
+  - forge
   - franka
   - simulation
   - benchmark
+  - physical-ai
+  - wrench
 language:
   - en
 size_categories:
   - 1K<n<10K
 task_categories:
   - robotics
+dataset_info:
+  features:
+    - name: observation.state
+      dtype: float32
+      shape:
+        - 24
+    - name: observation.wrench
+      dtype: float32
+      shape:
+        - 6
+    - name: action
+      dtype: float32
+      shape:
+        - 7
+    - name: timestamp
+      dtype: float64
+    - name: frame_index
+      dtype: int64
+    - name: episode_index
+      dtype: int64
+    - name: index
+      dtype: int64
+    - name: task_index
+      dtype: int64
+  splits:
+    - name: train
+      num_examples: 330929
+configs:
+  - config_name: default
+    data_files:
+      - split: train
+        path: "data/**/*.parquet"
 ---
 
 # EXOKERN ContactBench v0 — Peg Insertion with Force/Torque
 
 <p align="center">
-  <b>The first publicly available insertion dataset with calibrated force/torque annotations.</b>
+  <strong>The first publicly available insertion dataset with calibrated 6-axis force/torque annotations.</strong><br>
+  <em>Part of the <a href="https://huggingface.co/EXOKERN">ContactBench</a> collection by EXOKERN — The Data Engine for Physical AI</em>
 </p>
 
-## What is this?
+<p align="center">
+  <img src="force_profile_sample.png" alt="Force profile during peg insertion episode" width="720">
+</p>
+
+---
+
+## Why This Dataset Exists
+
+Over 95% of existing robotics manipulation datasets contain **no force/torque data**. Yet contact-rich tasks — insertion, threading, snap-fit assembly — fundamentally depend on haptic feedback for reliable execution. Vision alone cannot distinguish a jammed peg from a seated one.
 
-This dataset contains **2,221 simulated peg-in-hole insertion episodes** with full **6-axis force/torque (wrench) data** at every timestep. It is generated in NVIDIA Isaac Lab using a Franka Emika Panda robot arm.
+This dataset provides **2,221 peg-in-hole insertion episodes** with full **6-axis wrench data at every timestep**, generated using the [FORGE](https://arxiv.org/abs/2408.04587) (Force-Guided Exploration) framework in NVIDIA Isaac Lab. Every frame captures what the robot *feels*, not just what it sees.
 
-**Why this matters:** Over 95% of existing robotics datasets lack force/torque data. Yet contact-rich manipulation tasks — insertion, assembly, cable routing — fundamentally require haptic feedback for reliable execution. This dataset fills that gap.
+---
 
-## Dataset Statistics
+## Dataset Overview
 
 | Metric | Value |
-|--------|-------|
+|---|---|
 | Episodes | 2,221 |
 | Total Frames | 330,929 |
-| Avg Episode Length | 149 steps |
-| FPS | 20 Hz |
-| Format | LeRobot v3.0 (Parquet) |
+| Avg Episode Length | ~149 steps |
+| Control Frequency | 20 Hz |
+| Format | [LeRobot v3.0](https://github.com/huggingface/lerobot) (Parquet) |
 | Robot | Franka Emika Panda (7-DOF) |
-| Simulator | NVIDIA Isaac Lab 2.3.x |
-| Task | Peg-in-Hole Insertion (FORGE) |
+| Simulator | NVIDIA Isaac Lab 2.3.x + PhysX GPU |
+| Task | `Isaac-Forge-PegInsert-Direct-v0` |
 | Size | ~75 MB |
+| License | CC-BY-NC 4.0 |
+
+---
 
 ## Features
 
-Each frame contains:
+Each frame contains the following tensors:
 
 | Feature | Shape | Description |
-|---------|-------|-------------|
-| `observation.state` | (24,) | Flat observation tensor (joint pos/vel, ee pose, object pose) |
-| **`observation.wrench`** | **(6,)** | **Force/Torque [Fx,Fy,Fz,Mx,My,Mz] in N/Nm** |
-| `action` | (7,) | Joint position targets |
-| `task` | string | Task description |
+|---|---|---|
+| `observation.state` | `(24,)` | Flattened observation vector — see [State Tensor Semantics](#state-tensor-semantics) below |
+| **`observation.wrench`** | **`(6,)`** | **6-axis force/torque: [Fx, Fy, Fz, Mx, My, Mz] — see [Wrench Specification](#wrench-specification)** |
+| `action` | `(7,)` | Delta end-effector pose command [dx, dy, dz, dRx, dRy, dRz] + success prediction |
+| `timestamp` | scalar | Wall-clock time within episode (s) |
+| `frame_index` | int | Frame position within episode |
+| `episode_index` | int | Episode identifier |
+
+---
+
+## State Tensor Semantics
 
-> **Note:** Isaac Lab Forge PegInsert uses `collapse_obs_dict` — all observations are compressed into a flat 24-tensor. Force/Torque is extracted separately via `env.unwrapped.force_sensor_smooth`.
+The `observation.state` tensor is a flat 24-element vector produced by FORGE's `collapse_obs_dict()`. It concatenates the following quantities in order:
+
+| Index | Field | Dims | Unit | Description |
+|---|---|---|---|---|
+| 0–2 | `fingertip_pos` | 3 | m | End-effector (fingertip) position in world frame |
+| 3–5 | `fingertip_pos_rel_fixed` | 3 | m | EE position relative to the socket (fixed part) |
+| 6–9 | `fingertip_quat` | 4 | — | EE orientation as quaternion [w, x, y, z] |
+| 10–12 | `ee_linvel` | 3 | m/s | End-effector linear velocity |
+| 13–15 | `ee_angvel` | 3 | rad/s | End-effector angular velocity |
+| 16–21 | `force_sensor_smooth` | 6 | N, N·m | Smoothed 6-axis wrench (same data as `observation.wrench`) |
+| 22 | `force_threshold` | 1 | N | Maximum allowable contact force (FORGE parameter) |
+| 23 | `ema_factor` | 1 | — | Exponential moving average smoothing coefficient |
+
+> **Implementation reference:** These fields correspond to `OBS_DIM_CFG` in [`factory_env_cfg.py`](https://github.com/isaac-sim/IsaacLab/blob/main/source/isaaclab_tasks/isaaclab_tasks/direct/factory/factory_env_cfg.py) (indices 0–15) extended by the FORGE force-sensing additions (indices 16–23). See [Noseworthy et al., 2024](https://arxiv.org/abs/2408.04587) §III for details.
+
+---
+
+## Wrench Specification
+
+| Property | Value |
+|---|---|
+| **Coordinate frame** | **End-effector body frame** (Franka `panda_hand` link) |
+| **Convention** | [Fx, Fy, Fz, Mx, My, Mz] |
+| **Force unit** | Newtons (N) |
+| **Torque unit** | Newton-meters (N·m) |
+| **Source** | `env.unwrapped.force_sensor_smooth` — Isaac Lab contact sensor with EMA smoothing |
+| **Typical force range** | ±50 N per axis |
+| **Typical torque range** | ±10 N·m per axis |
+
+The wrench is reported at the **end-effector body frame** attached to the Franka gripper link. Forces are measured via Isaac Lab's `get_link_incoming_joint_force()` method applied to the wrist joint, then smoothed with an exponential moving average (EMA factor configurable, default 0.2).
+
+> **Note:** `observation.wrench` contains the same data as `observation.state[16:22]`. The wrench is stored as a dedicated column for direct access without index arithmetic.
+
+---
 
 ## Quick Start
 
 ```python
 from lerobot.datasets.lerobot_dataset import LeRobotDataset
 
-dataset = LeRobotDataset("exokern/contactbench-forge-peginsert-v0")
+# Load dataset
+dataset = LeRobotDataset("EXOKERN/contactbench-forge-peginsert-v0")
 print(f"Episodes: {dataset.num_episodes}, Frames: {len(dataset)}")
 
-# Access a frame
+# Access a single frame
 frame = dataset[0]
-wrench = frame["observation.wrench"]  # Force/Torque tensor (6,)
-print(f"Force: {wrench[:3]} N, Torque: {wrench[3:]} Nm")
+state  = frame["observation.state"]   # (24,) — full observation
+wrench = frame["observation.wrench"]  # (6,)  — force/torque
+
+# Decompose wrench
+force  = wrench[:3]  # [Fx, Fy, Fz] in N
+torque = wrench[3:]  # [Mx, My, Mz] in N·m
+print(f"Force: {force} N")
+print(f"Torque: {torque} N·m")
+
+# Decompose state vector
+ee_pos       = state[0:3]    # fingertip position (m)
+ee_pos_rel   = state[3:6]    # position relative to socket (m)
+ee_quat      = state[6:10]   # orientation quaternion
+ee_linvel    = state[10:13]  # linear velocity (m/s)
+ee_angvel    = state[13:16]  # angular velocity (rad/s)
+wrench_state = state[16:22]  # force/torque (N, N·m) — same as observation.wrench
+f_threshold  = state[22]     # force threshold (N)
+ema          = state[23]     # EMA smoothing factor
+```
+
+### Load with pure HuggingFace Datasets
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("EXOKERN/contactbench-forge-peginsert-v0", split="train")
+print(ds[0].keys())
 ```
 
+---
+
 ## Experimental Results: The Value of Force/Torque
 
 We trained a Behavior Cloning (BC) policy on this dataset to ablate the impact of Force/Torque data on the Peg-In-Hole task. Both policies achieve a 100% insertion success rate, but the difference in physical execution is massive:
@@ -84,57 +198,81 @@ We trained a Behavior Cloning (BC) policy on this dataset to ablate the impact o
 
 **Conclusion:** Both policies solve the geometric task. But the F/T-aware policy performs the insertion softly like an expert, reducing contact forces by **98.4%**. In industrial applications, this is the difference between a successful assembly and a damaged part.
 
+---
+
 ## Data Collection
 
-- **Policy:** rl_games PPO, trained ~200 epochs (reward ~352)
-- **Environment:** `Isaac-Forge-PegInsert-Direct-v0` (NVIDIA Isaac Lab 2.3.x)
-- **Collection:** Single-environment rollout (`num_envs=1`), deterministic policy
-- **Force/Torque Source:** `env.unwrapped.force_sensor_smooth` (6-axis, smoothed)
-- **Episode Horizon:** Fixed 149 steps (no early termination)
+| Parameter | Value |
+|---|---|
+| **RL Algorithm** | rl_games PPO (~200 epochs, reward ~352) |
+| **Environment** | `Isaac-Forge-PegInsert-Direct-v0` |
+| **Collection mode** | Single-env rollout (`num_envs=1`), deterministic policy |
+| **Episode horizon** | Fixed 149 steps (no early termination during collection) |
+| **Sensor bandwidth** | 20 Hz (matched to control frequency) |
+| **Contact dynamics** | PhysX GPU solver, Isaac Lab default parameters |
+| **Domain randomization** | FORGE defaults (controller gains, friction, mass, dead-zone) |
+
+---
+
+## Intended Use
+
+**Research applications:**
+- Training and benchmarking force-aware manipulation policies
+- Sim-to-real transfer studies for contact-rich assembly
+- Hybrid vision + force/torque policy architectures
+- Evaluating the effect of F/T data on manipulation performance
 
-## Data Quality
+**Not intended for:**
+- Direct deployment on physical robots without sim-to-real calibration
+- Safety-critical applications without additional validation
 
-- **Sensor Bandwidth:** 20 Hz control frequency
-- **Force Range:** Typical Fx,Fy,Fz: [-50, 50] N; Mx,My,Mz: [-10, 10] Nm
-- **Noise Model:** Isaac Lab default contact dynamics (PhysX GPU)
-- **Domain Randomization:** Default FORGE PegInsert parameters
+---
 
 ## Reproduction
 
-```bash
-# 1. Train policy (~20-30 min on RTX 4090)
-./isaaclab.sh -p scripts/reinforcement_learning/rl_games/train.py \
-    --task Isaac-Forge-PegInsert-Direct-v0 --headless --num_envs 512
+Methodology documentation available upon request.
+Visit [exokern.com](https://exokern.com) for details.
 
-# 2. Collect data (~2 episodes/min with num_envs=1)
-./isaaclab.sh -p /workspace/play_with_logging.py \
-    --task Isaac-Forge-PegInsert-Direct-v0 --headless --num_envs 1 \
-    --checkpoint /path/to/Forge.pth
+---
 
-# 3. Build LeRobot dataset (in conda env, NOT Isaac Sim Python!)
-python build_lerobot_dataset.py
-```
+## Related Work
+
+This dataset builds on the following research:
+
+- **FORGE** — Noseworthy et al., "[Force-Guided Exploration for Robust Contact-Rich Manipulation under Uncertainty](https://arxiv.org/abs/2408.04587)" (ISRR 2024)
+- **Factory** — Narang et al., "[Factory: Fast Contact for Robotic Assembly](https://arxiv.org/abs/2205.03532)" (RSS 2022)
+- **IndustReal** — Tang et al., "[IndustReal: Transferring Contact-Rich Assembly Tasks from Simulation to Reality](https://arxiv.org/abs/2305.17110)" (RSS 2023)
+- **LeRobot** — Cadene et al., "[LeRobot: State-of-the-art Machine Learning for Real-World Robotics](https://github.com/huggingface/lerobot)" (2024)
+
+---
 
 ## License
 
-CC-BY-NC 4.0 — Free for research and non-commercial use.
-Commercial licensing available from EXOKERN.
+**CC-BY-NC 4.0** — Free for research and non-commercial use.
+
+Commercial licensing and custom Contact Skill Packs available from EXOKERN.
+Visit [exokern.com](https://exokern.com) for enterprise inquiries.
+
+---
 
 ## About EXOKERN
 
 **EXOKERN — The Data Engine for Physical AI**
 
-We produce industrially calibrated force/torque manipulation data for enterprise robotics.
-Contact-rich. Sensor-annotated. Industrially validated.
+We produce industrially calibrated force/torque manipulation data for enterprise robotics, humanoid manufacturers, and research institutions. Contact-rich. Sensor-annotated. Industrially validated.
+
+🌐 [exokern.com](https://exokern.com) · 🤗 [huggingface.co/EXOKERN](https://huggingface.co/EXOKERN)
+
+---
 
 ## Citation
 
 ```bibtex
-@dataset{exokern_contactbench_v0_2026,
-  title={EXOKERN ContactBench v0: Peg Insertion with Force/Torque},
-  author={EXOKERN},
-  year={2026},
-  publisher={Hugging Face},
-  url={https://huggingface.co/datasets/exokern/contactbench-forge-peginsert-v0}
+@dataset{exokern_contactbench_peginsert_v0,
+  title   = {ContactBench v0: Peg Insertion with Force/Torque},
+  author  = {{EXOKERN}},
+  year    = {2026},
+  url     = {https://huggingface.co/datasets/EXOKERN/contactbench-forge-peginsert-v0},
+  note    = {2,221 episodes, 330K frames, 6-axis F/T, LeRobot v3.0 format}
 }
 ```