Scaling Behavior Foundation Model for Humanoid Robots
Weishuai Zeng, Kangning Yin, Xiaojie Niu, Shunlin Lu, Weixiang Zhong, Jiahe Chen, Feiyu Jia, Xiao Chen, Zirui Wang, Furui Xu, Ming Zhou, Kailin Li, Weinan Zhang, He Wang, Li Yi, Dahua Lin, Jiangmiao Pang, Jingbo Wang
Why It Matters
What makes this one worth your time
Improving humanoid robot control is crucial for developing generalist embodied agents capable of operating in diverse environments, which has significant implications for robotics applications.
This work advances humanoid robot control through a novel scaling approach for Behavior Foundation Models.
Summary
The paper presents a framework for scaling Behavior Foundation Models (BFMs) in humanoid robots, focusing on the coordination of learning paradigms, data diversity, and model architecture to enhance control fidelity and task generalization.
Key contributions
- Coordination of learning paradigms to reformulate humanoid control problems.
- Strategic synergy between on-policy rollout quantity and reference motion diversity.
- Introduction of the Humanoid Transformer architecture for scalable behavioral representations.
Notable insights
- The integration of motion tracking as a learning paradigm allows for a more holistic approach to humanoid control problems.
- The proposed Humanoid Transformer architecture facilitates the emergence of structured behavioral representations, enhancing model expressiveness.
Possible limitations
- Not stated in the abstract.
Abstract
arXiv:2607.15163v1 Announce Type: cross Abstract: Humanoid control requires natural whole-body coordination, precise real-time responses to control signals, and robust generalization across diverse environmental contexts, making it a cornerstone for generalist embodied agents. Behavior Foundation Models (BFMs) have recently emerged as a promising solution to address these challenges by leveraging large-scale behavioral data to achieve superior expressiveness, versatility and generalization. However, despite growing interest in scaling BFMs to further improve their capabilities, it remains unclear how key factors, including the learning paradigm, behavioral data and model architecture should be coordinated to enable effective scaling. In this work, we revisit the scaling recipe for BFMs and demonstrate that substantial performance gains can be achieved through the coordination of three core components: 1) the learning paradigm of motion tracking that reformulates diverse humanoid control problems as the reproduction of integrated whole-body behaviors in the global frame; 2) the strategic synergy between on-policy rollout quantity and reference motion diversity; and 3) the expressive and scalable model architecture termed Humanoid Transformer that facilitates the natural emergence of structured behavioral representations. Through extensive experiments in both simulation and real-world deployment, we demonstrate that our approach yields significant improvements in control fidelity and task generalization, reducing Mean Per-Keypoint Position Error (MPKPE) on the test set by over 10% in local mode and 82% in global mode compared with existing humanoid controllers. These results establish BFM as a principled and effective foundation for scalable and general-purpose humanoid control.