Soft Robotic Sim2Real via Conditional Flow Matching

This repository contains the implementation of our method for bridging the Sim2Real gap in soft robotics using Conditional Flow Matching (CFM). Our approach learns a mapping between simulation and real-world domains to accurately predict force and deformation in soft robotic systems.

Features

• Conditional Flow Matching: Novel application of CFM to bridge the Sim2Real gap in soft robotics
• Multi-Domain Support: Works with both tensile test benchmarks and Fin Ray gripper systems
• Encoder Architecture: Custom encoder for learning contact state representations
• Hysteresis Modeling: Captures complex elastic behaviors including hysteresis and force fluctuations
• Strong Generalization: Effective performance with limited datasets and on unseen interactions

Installation

Prerequisites

• Python 3.8+
• CUDA-capable GPU (recommended)
• PyTorch 1.12+

Setup

1. Clone the repository:

git clone https://github.com/csiro-robotics/Sim2Real_CFM_Soft_robotic.git
cd Sim2Real_CFM_Soft_robotic

2. Install dependencies:

pip install -r requirements.txt

Dataset

Download Dataset

The complete dataset for this project is available through CSIRO Data Access Portal:

Dataset Link: https://data.csiro.au/collection/csiro:65870

Dataset Contents

The dataset includes:

• Tensile Sim2Real data: Experimental and simulated tensile test measurements
• Tensile Sim2Sim data: SOFA and Warp simulation comparisons
• Fin Ray Gripper data: Real-world and simulated gripper force measurements

Usage

Training the CFM Model

Run the training script for tensile test Sim2Real:

cd tensile-sim2real-src
python data_gen_model_train_sim2real_gen.py
python data_gen_model_cfm_extra_condi.py

Contact

For questions and feedback, please contact:

• [Ge Shi] - [ge.shi@csiro.au]

Acknowledgments

This research was conducted at CSIRO's Robotics and Autonomous Systems Group.