IRASim: Learning Interactive Real-Robot Action Simulators

[Project page] [Paper]

Fangqi Zhu^1,2, Hongtao Wu^1†*, Song Guo^2*, Yuxiao Liu¹, Chilam Cheang¹, Tao Kong¹

¹ByteDance Research, ²Hong Kong University of Science and Technology

*Corresponding authors †Project Lead

Introduction.mp4

Scalable robot learning in the real world is limited by the cost and safety issues of real robots. In addition, rolling out robot trajectories in the real world can be time-consuming and labor-intensive. In this paper, we propose to learn an interactive real-robot action simulator as an alternative. We introduce a novel method, IRASim, which leverages the power of generative models to generate extremely realistic videos of a robot arm that executes a given action trajectory, starting from an initial given frame. To validate the effectiveness of our method, we create a new benchmark, IRASim Benchmark, based on three real-robot datasets and perform extensive experiments on the benchmark. Results show that IRASim outperforms all the baseline methods and is more preferable in human evaluations. We hope that IRASim can serve as an effective and scalable approach to enhance robot learning in the real world. To promote research for generative real-robot action simulators, we open-source code, benchmark, and checkpoints.

Installation

To set up the environment, run the following command:

bash scripts/install.sh

Dataset

To download the complete dataset, run:

bash scripts/download.sh

This table lists the download links and file sizes for the RT-1, Bridge, and Language-Table datasets, categorized into train, evaluation, and checkpoints data.

Category	Train	Size	Evaluation	Size	Checkpoints	Size
RT-1	rt1_train_data.tar.gz	86G	rt1_evaluation_data.tar.gz	100G	rt1_checkpoints_data.tar.gz	29G
Bridge	bridge_train_data.tar.gz	31G	bridge_evaluation_data.tar.gz	63G	bridge_checkpoints_data.tar.gz	32G
Language-Table	languagetable_train_data.tar.gz	200G	languagetable_evaluation_data.tar.gz	194G	languagetable_checkpoints_data.tar.gz	34G

The complete dataset structure can be found in dataset_structure.txt.

📢 Update (May 20, 2025)

We are excited to announce that the IRASim dataset is now available on Hugging Face:
🔗 https://huggingface.co/datasets/fangqi/IRASim

To reconstruct the full dataset locally:

1. Download all dataset parts from the Hugging Face page.
2. Use the provided merge.sh script to merge the downloaded files into multiple ZIP archives.
3. Extract each ZIP file separately to access the complete dataset.

Language Table Application

We recommend starting with the Language Table application. This application provides a user-friendly keyboard interface to control the robotic arm in an initial image on a 2D plane:

python3 application/languagetable.py

Training

Below are example scripts for training the IRASim-Frame-Ada model on the RT-1 dataset.

To accelerate training, we recommend encoding videos into latent videos first. Our code also supports direct training by setting pre_encode to false.

Single GPU Training

python3 main.py --config configs/train/rt1/frame_ada.yaml

Multi-GPU Training on a Single Machine

torchrun --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_endpoint {node_address}:{port} --rdzv_id 107 --rdzv_backend c10d main.py --config configs/train/rt1/frame_ada.yaml

Evaluation

Below are example scripts for evaluating the IRASim-Frame-Ada model on the RT-1 dataset.

Short Trajectory Setting

To quantitatively evaluate the model in the short trajectory setting, we first need to generate all evaluation videos.

Generate evaluation videos:

torchrun --nproc_per_node 8 --nnodes 1 --node_rank 0 --rdzv_endpoint {node_address}:{port} --rdzv_id 107 --rdzv_backend c10d main.py --config configs/evaluation/rt1/frame_ada.yaml

We provide an automated script to calculate the metrics of the generated short videos:

python3 evaluate/evaluation_short_script.py

Long Trajectory Setting

Generate all long videos in an autoregressive manner.

Generate the scripts for generating long videos in a multi-process manner:

python3 scripts/generate_command.py

Run:

bash scripts/generate_long_video_rt1_frame_ada.sh

Use the automated script to calculate the metrics of the generated long videos:

python3 evaluate/evaluation_long_script.py

Citation

If you find this code useful in your work, please consider citing

@article{FangqiIRASim2024,
  title={IRASim: Learning Interactive Real-Robot Action Simulators},
  author={Fangqi Zhu and Hongtao Wu and Song Guo and Yuxiao Liu and Chilam Cheang and Tao Kong},
  year={2024},
  journal={arXiv:2406.12802}
}

Acknowledgement

• Our implementation is largely adapted from Latte.
• Our FVD implementation is adapted from stylegan-v.
• Our FID implementation is adapted from pytorch-fid.
• Our RT-1, Bridge, Language-Table datasets are adapted from RT-1, Bridge, open_x_embodiment .

Discussion Group

If you have any questions during the trial, running or deployment, feel free to join our WeChat group discussion! If you have any ideas or suggestions for the project, you are also welcome to join our WeChat group discussion!