BEHAVIOR Robot Suite: Streamlining Real-World Whole-Body Manipulation for Everyday Household Activities

Yunfan Jiang, Ruohan Zhang, Josiah Wong, Chen Wang, Yanjie Ze, Hang Yin, Cem Gokmen, Shuran Song, Jiajun Wu, Li Fei-Fei

[Website] [arXiv] [PDF] [Doc] [Robot Code] [Training Data]

---

We introduce the BEHAVIOR Robot Suite (BRS), a comprehensive framework for learning whole-body manipulation to tackle diverse real-world household tasks. BRS addresses both hardware and learning challenges through two key innovations: WB-VIMA and JoyLo.

WB-VIMA is an imitation learning algorithm designed to model whole-body actions by leveraging the robot’s inherent kinematic hierarchy. A key insight behind WB-VIMA is that robot joints exhibit strong interdependencies—small movements in upstream links (e.g., the torso) can lead to large displacements in downstream links (e.g., the end-effectors). To ensure precise coordination across all joints, WB-VIMA conditions action predictions for downstream components on those of upstream components, resulting in more synchronized whole-body movements. Additionally, WB-VIMA dynamically aggregates multi-modal observations using self-attention, allowing it to learn expressive policies while mitigating overfitting to proprioceptive inputs.

Getting Started

Tip

🚀 Check out the doc for detailed installation and usage instructions!

To train a WB-VIMA policy, simply run the following command:

python3 main/train/train.py data_dir=<HDF5_PATH> \
    bs=<BS> \
    arch=wbvima \
    task=<TASK_NAME> \
    exp_root_dir=<EXP_ROOT_DIR> \
    gpus=<NUM_GPUS> \
    use_wandb=<USE_WANDB> \
    wandb_project=<WANDB_PROJECT>

To deploy a WB-VIMA policy on the real robot, simply run the following command:

python3 main/rollout/<TASK_NAME>/rollout_async.py --ckpt_path <CKPT_PATH> --action_execute_start_idx <IDX>

Check out Our Paper

Our paper is posted on arXiv. If you find our work useful, please consider citing us!

@article{jiang2025brs,
title = {BEHAVIOR Robot Suite: Streamlining Real-World Whole-Body Manipulation for Everyday Household Activities},
author = {Yunfan Jiang and Ruohan Zhang and Josiah Wong and Chen Wang and Yanjie Ze and Hang Yin and Cem Gokmen and Shuran Song and Jiajun Wu and Li Fei-Fei},
year = {2025},
journal = {arXiv preprint arXiv: 2503.05652}
}

License

This codebase is released under the MIT License.