# WorldStream：分层规划与持久状态实现开放式3D世界流式生成

> 原文链接：https://robotics.hellobike.com/papers/2026-10-20-2610-28827-0c93b530
> 频道：papers | 更新日期：2026-10-20 | 标签：auto, arxiv, hf-daily-papers
> 核心摘要：提出WorldStream开放式3D世界模型，通过宏微观分层规划器与确定性编译机制，实现超地图边界扩展、自然语言交互及编辑状态持久化。该架构在有限算力预算下支持 tiles 独立规划与无缝连接，为具身智能提供可交互的无限仿真环境。

# WorldStream：分层规划与持久状态实现开放式3D世界流式生成

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[WorldStream: Streaming Interactive 3D Worlds with Hierarchical Planning and Persistent State](https://arxiv.org/abs/2610.28827)（HF Daily Papers · arXiv 2610.28827）

## 核心内容

WorldStream 是一种具备开放Ended能力的3D世界模型，核心突破在于同时实现地形无限扩展、运行时语言指令响应及访问区域编辑状态的精确恢复。该系统解决了传统生成式仿真中地图边界受限、交互非结构化及状态易丢失的问题，为具身智能体提供了在有限计算预算下持续演化、逻辑自洽且支持回溯编辑的高保真训练与验证环境。

技术架构采用基于类型决策模型的分层规划器：Macro 层统筹聚落、生态、交通与水文结构，Micro 层将其特化为具体 tile 语义并约束邻接关系。确定性本地代码将规划编译为地形与实体，配合流式机制实现预测走廊预规划、按需几何编译及后方资源释放。单一世界记录持久化存储种子、对象ID、可变状态及边界承诺，确保返回区域重编译一致性，并支持NPC等有状态实体的实时交互与意图挂载。

## 关键看点

- **分层规划架构**：Macro层定义全局拓扑结构，Micro层处理tile语义与邻接约束，通过类型化决策模型输出结构化状态，保障生成内容的逻辑连贯性。
- **流式资源调度**：采用预测走廊预规划与按需编译策略，仅在移动路径前方生成几何体并及时释放后方资源，实现有界算力下的无限世界扩展。
- **持久化状态管理**：单一世界记录保存规划、种子、对象身份及属性级编辑历史，支持用户返回时基于完整上下文重编译，确保交互与修改的长期一致性。
- **交互式实体系统**：模型可直接操作现有实体与有状态NPC，支持对当前区域编辑及向未规划区域附加意图，实现自然语言驱动的动态世界演化。

## 工程与产业观察

该工作为具身智能仿真提供了关键的“无限世界”基础设施。其分层规划与持久化状态机制，有效解决了生成式环境在长时序任务训练中的一致性难题。对于算法工程师，这种支持编辑回溯与语言交互的流式架构，是构建高保真Sim-to-Real闭环、验证通用机器人长程规划能力的重要参考范式。

## 英文摘要 (Abstract)

We present WorldStream, an open-ended 3D world model with three simultaneous capabilities: it extends playable terrain beyond any pre-specified map, acts on open-ended language in the running world, and restores visited places with their edits. A hierarchical planner over a typed decision model emits structured decision state: Macro fixes settlement, ecology, transport, and hydrological structure across a group of tiles; Micro specializes that contract into tile semantics under measured terrain and accepted neighbor commitments. Deterministic local code compiles accepted plans into terrain, roads, settlements, vegetation, and entities. Streaming separates their reach: plans run ahead along the predicted corridor, compilation only where movement will soon need it, geometry behind the user released. One world record retains plans, seeds, object identities and mutable state, boundary commitments, and edits by attribute, revised as the user advances and edits; returning tiles recompile from it. The same record grounds interaction: the model acts on existing entities, including stateful NPCs that update as they act, edits the present region, or attaches intent to unplanned regions. Together they yield a world expanding under bounded budgets, connected across independently planned tiles, and interactive.

Hugging Face Daily Papers 热度：1 upvotes
作者：Xiaoxiao Ma, Weiyang Jin, Yanwen Ma, Haoran Li, Zhixiang Wei, Songchun Zhang, Yaowei Li, Junhao Zhuang, Zeyue Xue, Yuming Li, Zheming Fu, Haoyang Huang, Nan Duan, Feng Zhao
:::

## 原始参考来源 (Sources)

- [HF Daily Papers · arXiv 2610.28827：WorldStream: Streaming Interactive 3D Worlds with Hierarchical Planning and Persistent State](https://arxiv.org/abs/2610.28827)