# I-BFM：通过无监督强化学习实现奖励条件化的鲁棒人形交互

> 原文链接：https://robotics.hellobike.com/papers/2026-10-05-2610-06129
> 频道：papers | 更新日期：2026-10-05 | 标签：arxiv, cs.RO
> 核心摘要：I-BFM用前向-后向表征与无监督强化学习学习人形-物体接触动力学，在多种交互任务和Unitree G1真机上验证。

# I-BFM：通过无监督强化学习实现奖励条件化的鲁棒人形交互

::: v-pre
> 原文链接：[I-BFM: Reward-Conditioned Robust Humanoid Interaction via Unsupervised Reinforcement Learning](https://arxiv.org/abs/2610.06129)（arXiv 2610.06129 · 发布日期：2026-10-05）

**I-BFM: Reward-Conditioned Robust Humanoid Interaction via Unsupervised Reinforcement Learning**

I-BFM用前向-后向表征与无监督强化学习学习人形-物体接触动力学，在多种交互任务和Unitree G1真机上验证。

## 核心要点

- BFM已支持多样全身控制，但扩展到物理交互仍具挑战。
- I-BFM学习人形、物体与接触的耦合动力学，避免依赖任务特定策略或参考跟踪。
- 给定下游任务奖励，同一策略可由潜在命令条件化，闭环执行交互且无需任务特定优化。
- 单一策略可搬运、推、踢，并支持目标到达、运动跟踪、风格控制和长时程任务串联。

**验证范围与局限**：验证范围包括多种人形-物体交互任务和Unitree G1真机实验。

## 英文摘要 (Abstract)

Behavioral foundation models (BFMs) have recently shown that a single humanoid policy can support diverse whole-body control, but extending such generality to physical interaction remains challenging. We introduce I-BFM, to our knowledge the first BFM for humanoid-object interaction. Rather than relying on task-specific policies or reference tracking, I-BFM learns a shared representation of the coupled dynamics among the humanoid, objects, and their contacts using forward-backward representations and unsupervised reinforcement learning. Given a downstream task reward, the same policy can be directly conditioned on a latent command to execute closed-loop interaction without task-specific policy optimization. To improve interaction control over different time scales, we further train the policy with both short-horizon interaction targets and longer-horizon goal targets. A single I-BFM policy performs carrying, pushing, and kicking, while also supporting goal reaching, motion tracking, stylistic control, and long-horizon task chaining. More importantly, it remains effective after large deviations from nominal execution: on Carry, I-BFM achieves 94.3% nominal success and retains 89.3% success after robot falls, compared with 1.3% for a planning-based baseline. Real-world experiments on a Unitree G1 further demonstrate diverse loco-manipulation behaviors, rapid recovery from interaction failures and external disturbances, and task chaining without task-specific retraining.

作者：Ziqi Han, Yitang Li, Junhan Sun, Fanrong Dong, Yaojie Shen, Lei Ye, Zetong Jing, Yongqi Zhang, Yiming Zhang, Xue Wang
:::

## 原始参考来源 (Sources)

- [arXiv 2610.06129：I-BFM: Reward-Conditioned Robust Humanoid Interaction via Unsupervised Reinforcement Learning](https://arxiv.org/abs/2610.06129)