# 走出去再回来：Real2Sim2Real 协同训练中的世界与行为 grounding

> 原文链接：https://robotics.hellobike.com/papers/2026-09-30-2610-00821
> 频道：papers | 更新日期：2026-09-30 | 标签：arxiv, cs.RO
> 核心摘要：研究用 Real2Sim2Real 分离世界与行为 grounding 生成协同训练数据，并在动态灵巧抓取分拣任务上验证性能提升。

# 走出去再回来：Real2Sim2Real 协同训练中的世界与行为 grounding

::: v-pre
> 原文链接：[Getting Out and Getting Back: World and Behavior Grounding in Real2Sim2Real Co-Training](https://arxiv.org/abs/2610.00821)（arXiv 2610.00821 · 发布日期：2026-09-30）

**Getting Out and Getting Back: World and Behavior Grounding in Real2Sim2Real Co-Training**

研究用 Real2Sim2Real 分离世界与行为 grounding 生成协同训练数据，并在动态灵巧抓取分拣任务上验证性能提升。

## 核心要点

- 问题在于真实示范稀缺可由仿真扩充，但世界保真度与人类行为相似性对策略表现的影响不清楚。
- 方法区分对齐仿真与真实系统的世界 grounding，以及对齐仿真轨迹与人类动作的行为 grounding，并独立改变两者生成数据。
- 在动态灵巧抓取分拣任务上，完全 grounding 的协同训练将成功率从52%提升到86%，世界与行为 grounding 平均分别带来18和10个百分点提升。
- 潜空间分析显示部署策略像真实来源与仿真来源策略的混合；grounded simulation 在协同训练基础模型时仍有益。

**验证范围与局限**：验证范围为动态灵巧抓取分拣任务；摘要未说明机器人数量或更多任务类型。

## 英文摘要 (Abstract)

Simulation can expand scarce real demonstrations for co-training, yet how world fidelity and similarity to human behavior affect policy performance remains unclear. We distinguish world grounding, which aligns simulation with the real system, and behavior grounding, which aligns simulated trajectories with human motion. We build a real2sim2real pipeline that varies these axes independently to generate data for co-training. On a dynamic dexterous pick-and-sort task, fully grounded co-training raises success from 52% to 86%; averaged across configurations, world grounding improves success by 18 percentage points and behavior grounding by 10. Deployed policies behave like a mixture of real-derived and simulation-derived policies, imitating real demonstrations in covered states and relying on simulated behavior elsewhere, which we examine through latent-space analysis. Together, these results suggest complementary roles: world grounding lets policies use simulated experience beyond real-data coverage, while behavior grounding matters mainly when world grounding is imperfect. Grounded simulation remains beneficial when co-training foundation models.

作者：Samuel Liu, Youngsun Kim, Martin Matak, Gilwoo Lee
:::

## 原始参考来源 (Sources)

- [arXiv 2610.00821：Getting Out and Getting Back: World and Behavior Grounding in Real2Sim2Real Co-Training](https://arxiv.org/abs/2610.00821)