外观
论文arXiv 2610.0796922 天前
EmbodiedSmith：递归自改进仿真框架生成具身数据提升策略泛化
本文由 AI 根据公开内容整理，可能有误，请以原文为准：EmbodiedSmith: Scaling Embodied Data through Recursive Self-Improvement Flywheel in Simulation（arXiv 2610.07969）
核心内容
针对现有仿真管线资产预定义、场景与任务割裂及物理支持受限等痛点，研究团队提出EmbodiedSmith框架。该方案核心在于构建递归自改进（RSI）闭环，将资产、场景与任务生成统一于单一语言驱动管线中，旨在解决机器人基础模型训练中高质量数据稀缺与评估环境不可靠的问题，为具身智能预训练提供可扩展的仿真引擎。
技术架构采用Agentic Refinement Loop机制，使场景生成预判任务需求，任务反馈指导场景编辑，实现双向迭代优化并提升长程任务成功率。系统原生支持移动操作臂、人形机器人及灵巧手本体，覆盖柔性物体与流体交互等复杂物理现象。下游策略实验表明，该框架生成的多样化数据能有效增强模型在未见场景中的泛化性能。
关键看点
- RSI闭环机制：构建场景与任务双向迭代优化回路，突破预定义资产限制，显著提升长程复杂任务生成的成功率与逻辑一致性
- 多模态物理支持：原生适配人形机器人与灵巧手本体，涵盖柔性物体及流体交互等高难度物理仿真，拓展具身行为数据边界
- 统一生成管线：整合资产、场景与任务生成模块，支持语言驱动定制化创建，解决传统仿真中场景与任务语义割裂问题
- 泛化验证有效：下游策略训练实验证实，基于该框架生成的高多样性数据可直接转化为模型在未知环境中的泛化能力提升
工程与产业观察
该工作验证了“数据-仿真”递归自改进范式的有效性，解决了具身大模型训练中合成数据与真实任务分布偏移的难题。其统一的生成管线与复杂物理支持能力，为Sim-to-Real迁移提供了高置信度验证环境，对构建通用具身基础模型的数据飞轮具有重要工程参考价值。
英文摘要 (Abstract)
Scaling robotic foundation models requires diverse training data and reliable evaluation environments. Simulation offers a scalable solution, yet existing generation pipelines remain constrained by predefined assets and skills, a disconnect between scene generation and task generation, and limited support for complex embodiments and physics. We introduce EmbodiedSmith, a framework for scalable embodied data generation through recursive self-improvement (RSI). EmbodiedSmith unifies asset, scene, and task generation in a pipeline that supports autonomous creation and language-driven customization. Its core is an agentic refinement loop: scene generation anticipates downstream task requirements, while task generation guides targeted scene edits, allowing scenes and tasks to iteratively improve one another. This joint refinement improves task generation success, including for long-horizon tasks. The framework further supports mobile manipulators, humanoids, and dexterous hands, as well as interactions involving deformable objects and fluids, broadening the range of behaviors and physical phenomena represented in generated data. Together, these capabilities provide a flexible simulation engine for both robot pretraining and evaluation. Extensive experiments validate the quality, diversity, and generation efficiency of the resulting data, while downstream policy experiments demonstrate that increased data diversity improves generalization.
作者：Yikai Qin, Yifei Deng, Mingjian Liang, Wenxuan Song, Zepeng Lin, Zhiyi Jiang, Jiajun Fu, Qiao Sun, Huashuo Lei, Xicheng Gong, Jiayi Chen, Han Zhao, Shuanghao Bai, Pengxiang Ding, Pengwei Wang, Haoang
来源
本页事实对应的公开页面，以原文为准。