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

**Authors:** Yunfan Jiang, Ruohan Zhang, Josiah Wong, Chen Wang, Yanjie Ze, Hang Yin, Cem Gokmen, Shuran Song, Jiajun Wu, Li Fei-Fei  
**arXiv:** [2503.05652](https://arxiv.org/abs/2503.05652)  
**Date:** 2025-03-07  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2503.05652v1/

## Key points

- The BEHAVIOR Robot Suite (BRS) is a framework for real-world whole-body manipulation in household activities.
- Supported by ONR MURI N00014-24-1-2748 funding.
- Integrates a bimanual, wheeled robot with a 4-DoF torso to tackle household task challenges.
- Features a cost-effective whole-body teleoperation interface and a novel algorithm for learning visuomotor policies.
- System complexity from hardware design complicates visuomotor policy learning.
- Tested on five challenging household tasks emphasizing core capabilities like long-range navigation, object interaction, and confined space manipulation.
- Open-sourced project led by authors Yunfan Jiang, Ruohan Zhang, Josiah Wong, Chen Wang, Yanjie Ze, Hang Yin, Cem Gokmen, Shuran Song, Jiajun Wu,and Li Fei-Fei.
- Limitations include training WB-VIMA with specific robot data on R1 robot and potential exploration of multi-embodiment data transfer benefits.

## Summary

The BEHAVIOR Robot Suite (BRS) is a comprehensive framework designed to streamline real-world whole-body manipulation for everyday household activities. The project receives support from ONR MURI N00014-24-1-2748 funding. The BRS framework integrates a bimanual, wheeled robot with a 4-DoF torso to address the challenges of household tasks. It includes a cost-effective whole-body teleoperation interface for data collection and a novel algorithm for learning whole-body visuomotor policies. Achieving these capabilities requires careful hardware design, but the resulting system complexity further complicates visuomotor policy learning. To evaluate its effectiveness, the system is tested on five challenging household tasks that emphasize core capabilities and introduce additional complexities such as long-range navigation, interaction with articulated and deformable objects, and manipulation in confined spaces. The integrated robotic embodiment, data collection interface, and learning framework of BRS mark a significant step toward enabling real-world whole-body manipulation for everyday household tasks. The project is open-sourced at https://behavior-robot-suite.github.io/ and led by authors Yunfan Jiang, Ruohan Zhang, Josiah Wong, Chen Wang, Yanjie Ze, Hang Yin, Cem Gokmen, Shuran Song, Jiajun Wu,and Li Fei-Fei. Additionally highlighted are limitations related to training WB-VIMA with specific robot data on R1 robot and potential exploration of multi-embodiment data transfer benefits.