# 物理经验驱动具身智能体持续空间世界模型构建与预测优化

> 原文链接：https://robotics.hellobike.com/papers/2026-11-04-2611-04016-4474b13e
> 频道：papers | 更新日期：2026-11-04 | 标签：auto, arxiv
> 核心摘要：研究提出跨片段空间变化记忆机制，使具身智能体在动态迷宫中通过物理交互持续更新世界模型。实验证实结构化经验能显著减少路径修正需求并迁移至未见场景，揭示了物理反馈对修正拓扑认知与动作预测失配的关键作用。

# 物理经验驱动具身智能体持续空间世界模型构建与预测优化

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[Physical Experience Builds Continual Spatial World Models](https://arxiv.org/abs/2611.04016)（arXiv 2611.04016）

## 核心内容

针对具身智能体在持续运行中面临环境变迁与旧知识冲突的难题，arXiv论文提出基于物理经验的持续空间世界模型构建方法。研究通过物理接地的动态迷宫实验，验证了智能体利用重复物理交互而非仅靠静态记忆来适应环境变化的有效性，证明了持续积累的经验能显著提升未来预测与规划能力，且该优势依赖于历史变化对未来状态的信息量。

技术方案引入跨片段空间变化记忆（Cross-episode Spatial Change Memory）学习环境变换关系，并结合状态-动作-结果体验细化运动预测。测试表明，结构化经验不仅降低了导航中的路线修正频率，还能泛化至未经历的空间转换；当先前决策失效时，预测性记忆能触发行为调整。物理测试进一步揭示，正确的拓扑认知不等于正确的动作预测，唯有累积的物理交互经验才能有效消除模型动力学内的预测失配。

## 关键看点

- **跨片段记忆**：提出Cross-episode Spatial Change Memory机制，学习环境变量间的转换关系，使世界模型具备持续更新能力而非单纯持久化存储
- **经验泛化验证**：结构化物理经验显著减少未来路径修正需求，并能成功迁移至训练阶段未遇到的空间转换场景，提升模型鲁棒性
- **行为自适应机制**：预测性记忆仅在先前决策因环境变化失效时触发行为改变，若原决策仍正确则保持行为不变，避免过度反应
- **物理交互必要性**：实验证实正确拓扑结构无法保证动作预测准确，必须依赖累积的物理交互经验来修正模型动力学中的预测偏差

## 工程与产业观察

该研究为具身智能体从“静态部署”转向“终身学习”提供了理论支撑。工程上提示开发者，构建可进化的世界模型不能仅依赖离线数据或纯视觉拓扑，必须设计允许智能体在部署后通过物理交互持续校准动力学的闭环机制，以解决仿真与现实、记忆与感知之间的长尾失配问题。

## 英文摘要 (Abstract)

World models are increasingly used to support prediction and planning in embodied agents, but the knowledge is usually established before deployment and then carried into later encounters. Continual operation faces a different problem. A familiar environment may change, and the agent must use previous experience without allowing old expectations to override current physical evidence. This study asks what repeated physical experience must contribute for a spatial world model to become progressively more useful rather than simply more persistent. A physics-grounded dynamic maze separates recurring spatial change from the continuous consequences of action. Cross-episode Spatial Change Memory learns relations among environmental transformations, while state-action-outcome experience refines predictions of motion. Structured experience progressively reduces the need for future route correction and transfers to spatial transitions not encountered during learning. The advantage disappears when past changes contain no information about future ones. Continuous navigation further shows that predictive memory changes behavior when a previous decision becomes invalid but leaves behavior unchanged when the decision remains correct. Physical tests expose a complementary limit: correct topology does not guarantee correct action prediction, and accumulated physical experience reduces this mismatch within the modeled dynamics. The results characterize continual world knowledge as experience that improves future prediction and action beyond stored episodes while remaining revisable by the physical outcomes it is intended to predict.

作者：Yuhang Sun
:::

## 原始参考来源 (Sources)

- [arXiv 2611.04016：Physical Experience Builds Continual Spatial World Models](https://arxiv.org/abs/2611.04016)