Ndolphin/SoftManipulator_sim2real

1---2language:3- en4tags:5- robotics6- soft-robotics7- sim2real8- physics-simulation9- neural-networks10- pneumatic-actuation11- motion-capture12- surrogate-modeling13- fem-simulation14- sofa-framework15size_categories:16- 100K<n<1M17task_categories:18- tabular-regression19- time-series-forecasting20task_ids:21- tabular-single-column-regression22- univariate-time-series-forecasting23pretty_name: "Soft Manipulator Sim2Real Dataset"24configs:25- config_name: default26  data_files:27  - "*.csv"28dataset_info:29  features:30  - name: P131    dtype: float6432    description: "Pneumatic pressure for cavity 1 (Pa)"33  - name: P234    dtype: float6435    description: "Pneumatic pressure for cavity 2 (Pa)"  36  - name: P337    dtype: float6438    description: "Pneumatic pressure for cavity 3 (Pa)"39  - name: thetaX40    dtype: float6441    description: "Joint angle X-axis (radians)"42  - name: thetaY43    dtype: float6444    description: "Joint angle Y-axis (radians)"45  - name: d46    dtype: float6447    description: "Linear displacement (mm)"48  - name: TCP_X49    dtype: float6450    description: "Tool center point X position (mm)"51  - name: TCP_Y52    dtype: float6453    description: "Tool center point Y position (mm)"54  - name: TCP_Z55    dtype: float6456    description: "Tool center point Z position (mm)"57  splits:58  - name: train59    num_bytes: 16700000060    num_examples: 20000061  download_size: 16700000062  dataset_size: 16700000063license: mit64paperswithcode_id: null65---66 67 68# SoftManipulator Sim2Real Dataset69 70This dataset accompanies the research paper "Bridging High-Fidelity Simulations and Physics-Based Learning Using A Surrogate Model for Soft Robot Control" published in Advanced Intelligent Systems, 2025.71 72## 📋 Dataset Overview73 74This dataset contains experimental and simulation data for a 3-actuator pneumatic soft manipulator, designed to enable sim-to-real transfer learning and surrogate model development. The data includes motion capture recordings, pressure mappings, SOFA FEM simulation outputs, and surrogate model training datasets.75 76## 🎯 Dataset Purpose77 78- **Sim2Real Research**: Bridge the gap between SOFA simulations and real hardware79- **Surrogate Model Training**: Train neural networks for fast dynamics prediction80- **Model Calibration**: Calibrate FEM parameters using real-world data81- **Workspace Analysis**: Understand the robot's range of motion and capabilities82- **Validation**: Compare simulation outputs with experimental ground truth83 84## 📊 Dataset Files85 86| File | Size | Samples | Description | Usage |87|------|------|---------|-------------|--------|88| `ForwardDynamics_Pybullet_joint_to_pos.csv` | ~66MB | 100,000+ | PyBullet forward dynamics: joint commands → TCP positions | Surrogate model training |89| `MotionCaptureData_ROM.csv` | ~15MB | 10,000+ | Real robot motion capture trajectories | Ground truth validation |90| `PressureThetaMappingData.csv` | ~2MB | 5,000+ | Pressure inputs → joint angle outputs | Actuation mapping |91| `Pressure_vs_TCP.csv` | ~8MB | 8,000+ | Pressure commands → tool center point positions | Control modeling |92| `RealPressure_vs_SOFAPressure.csv` | ~3MB | 3,000+ | Hardware vs simulation pressure comparison | Model calibration |93| `SOFA_snapshot_data.csv` | ~45MB | 50,000+ | FEM nodal displacements from SOFA simulations | Physics validation |94| `SurrogateModel_ROM.csv` | ~12MB | 15,000+ | Reduced-order model training data | Fast inference |95| `SurrogateModel_withTooltip_ROM.csv` | ~18MB | 20,000+ | ROM data with tooltip contact forces | Contact modeling |96 97## 🔧 Data Collection Setup98 99### Hardware Configuration100- **Robot**: 3-cavity pneumatic soft manipulator (silicone, ~150mm length)101- **Actuation**: Pneumatic pressure control (-20 kPa to +35 kPa per cavity)102- **Sensing**: 6-DOF motion capture system (OptiTrack), pressure sensors103- **Materials**: Ecoflex 00-30 silicone with embedded pneumatic chambers104 105### Simulation Environment106- **SOFA Framework**: v22.12 with SoftRobots plugin107- **FEM Model**: TetrahedronFEMForceField with NeoHookean material108- **Material Properties**: Young's modulus 3-6 kPa, Poisson ratio 0.41109- **PyBullet**: v3.2.5 for surrogate model validation110 111## 📈 Data Schema112 113### Joint Space Data114- `thetaX`, `thetaY`: Joint angles (radians, -π/4 to π/4)115- `d`: Linear displacement (mm, 0 to 50)116 117### Pressure Commands118- `P1`, `P2`, `P3`: Cavity pressures (Pa, -20000 to 35000)119 120### Cartesian Space121- `TCP_X`, `TCP_Y`, `TCP_Z`: Tool center point position (mm)122- `Normal_X`, `Normal_Y`, `Normal_Z`: End-effector orientation123 124### Forces125- `Fx`, `Fy`, `Fz`: External forces (N, contact/manipulation tasks)126 127### Temporal Information128- `Time`: Timestamp (seconds)129- `Episode`: Experiment episode number130 131## 🚀 Usage Examples132 133### Loading Data in Python134```python135import pandas as pd136from datasets import load_dataset137 138# Load from HuggingFace139dataset = load_dataset("Ndolphin/SoftManipulator_sim2real")140 141# Or load locally142df = pd.read_csv("ForwardDynamics_Pybullet_joint_to_pos.csv")143print(f"Dataset shape: {df.shape}")144print(f"Columns: {df.columns.tolist()}")145```146 147### Training a Surrogate Model148```python149# Pressure to joint angle mapping150X = df[['P1', 'P2', 'P3']].values  # Pressure inputs151y = df[['thetaX', 'thetaY', 'd']].values  # Joint outputs152 153from sklearn.model_selection import train_test_split154X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)155 156# Train your neural network model157```158 159### Motion Analysis160```python161# Analyze workspace coverage162import matplotlib.pyplot as plt163 164tcp_data = df[['TCP_X', 'TCP_Y', 'TCP_Z']]165fig = plt.figure()166ax = fig.add_subplot(111, projection='3d')167ax.scatter(tcp_data['TCP_X'], tcp_data['TCP_Y'], tcp_data['TCP_Z'])168ax.set_title('Robot Workspace')169```170 171## 📝 Data Quality & Preprocessing172 173### Quality Assurance174- **Filtering**: Outliers removed using 3-sigma rule175- **Smoothing**: Savitzky-Golay filter applied to motion capture data176- **Synchronization**: All sensors synchronized to 100Hz sampling rate177- **Validation**: Cross-validated against multiple experimental runs178 179### Recommended Preprocessing180```python181from sklearn.preprocessing import StandardScaler182 183# Normalize features for neural network training184scaler = StandardScaler()185X_normalized = scaler.fit_transform(X)186 187# Save scaler for inference188import joblib189joblib.dump(scaler, 'scaler.pkl')190```191 192## 🎓 Citation193 194If you use this dataset in your research, please cite:195 196```bibtex197@article{hong2025bridging,198  title={Bridging High-Fidelity Simulations and Physics-Based Learning Using A Surrogate Model for Soft Robot Control},199  author={Hong, T. and Lee, J. and Song, B.-H. and Park, Y.-L.},200  journal={Advanced Intelligent Systems},201  year={2025},202  publisher={Wiley}203}204```205 206## 📄 License207 208This dataset is released under the MIT License. See LICENSE file for details.209 210## 🤝 Contact211 212For questions about the dataset or research:213- **Authors**: T. Hong, J. Lee, B.-H. Song, Y.-L. Park214- **Institution**: [Your Institution]215- **Email**: [Contact Email]216- **Paper**: [ArXiv/DOI Link when available]217 218## 🔍 Related Resources219 220- **Code Repository**: https://github.com/ndolphin-github/Sim2Real_framework_SoftRobot221- **SOFA Simulations**: Included in the repository222- **Pre-trained Models**: Available in the code repository223- **Demo Videos**: SOFA simulation demos included224 225---226 227*Dataset Version: 1.0 | Last Updated: October 2025*