Drawer Simulation: Digital Reconstruction and Articulation With Environment Realism
February 10, 2026 ยท View on GitHub
This project provides a simulation and data generation pipeline for drawer articulation using Isaac Sim 4.2 and Isaac Lab, with support for large-scale demonstration collection and visual replay.
โ ๏ธ Requirement This code is developed and tested on Isaac Sim 4.2. Please install Isaac Sim 4.2 before running the commands below.
๐ Kitchen Asset Download
Download the reconstructed kitchen assets used in this project from the link below and place them in the appropriate asset directory specified in the task configuration:
๐ Kitchen Assets (Google Drive) https://drive.google.com/drive/folders/17bVju4wAgy6MGNono_AfAqQ7pl0-VQC_?usp=drive_link
๐ Data Collection
./isaaclab.sh -p scripts/workflows/automatic_articulation/random_multi_step.py --task Isaac-Open-Drawer-Franka-IK-Abs-v0 --num_envs=1 --enable_cameras --config_file source/config/task/automatic_articulation/kitchen01.yaml --log_dir logs/kitchen01 --num_demos=50 --save_path raw_data --init_open
๐ Data Conversion (NPZ โ HDF5)
./isaaclab.sh -p scripts/workflows/utils/convert_npz_to_h5py.py --task Isaac-Open-Drawer-Franka-IK-Abs-v0 --num_envs=1 --config_file source/config/task/automatic_articulation/kitchen01.yaml --log_dir logs/kitchen01 --num_demos=2000 --load_path raw_data --save_path raw_data
๐ฅ Replay & Rendering
./isaaclab.sh -p scripts/workflows/automatic_articulation/replay_multi_step.py --task Isaac-Open-Drawer-Franka-IK-Abs-v0 --num_envs=1 --config_file source/config/task/automatic_articulation/kitchen01.yaml --log_dir logs/kitchen02_yunchu --enable_cameras --num_demos=2000 --init_open --load_path raw_data --save_path render_data
๐ Notes
Use --num_envs=1 for stable articulation and debugging.
Camera settings and articulation parameters are defined in the YAML config.
The replay step is recommended to verify articulation quality and data correctness before training.