Iterative Design for Robotic Assembly (IDfRA)

Official codebase for the paper titled "IDfRA: Self-Verification for Iterative Design in Robotic Assembly" (ICRA 2026)

Setup

Installation

git clone https://github.com/nishkakhendry/iterative_dfra.git
conda env create -f environment.yaml
conda activate idfra

OpenAI API Key

IDfRA uses the GPT-4o for all LLM and VLM functionalities via OpenAI API: create a file named .env in the root directory of the repository and include OPENAI_API_KEY=[your api key]

Running IDfRA

Repository Structure

Directories

• prompting_and_structure/: modified code from BloxNet for GPT prompting, structural functions
• blocksets/: all blocksets used for assembly generation

Scripts

• run_idfra_pipeline.py: runs end-to-end IDfRA pipeline in simulation
• pick_and_place_env.py: custom simulation environment class, modified from SayCan
• robot_suction.py: custom suction gripper class
• utils.py: misc functions for dimension switch, plots, pictures
• vlm_resemblance_evaluation.py: runs VLM-based quantitative assessment

Generating Assembly Structures

To iteratively generate assemblies using IDfRA, run:

python run_idfra_pipeline.py