外观
论文arXiv 2504.160542025-04-22
π0.5模型实现跨机器人开放世界泛化与长程灵巧操作
本文由 AI 根据公开内容整理，可能有误，请以原文为准：$π_{0.5}$: a Vision-Language-Action Model with Open-World Generalization（arXiv 2504.16054）
核心内容
针对VLA模型在实验室外泛化能力不足的问题，研究团队提出π0.5模型。该系统基于π0架构，利用来自多种机器人本体、Web数据及高层语义预测的异构数据进行联合训练。实验首次证明，该端到端学习系统能在全新家庭场景中完成厨房或卧室清洁等长程灵巧操作任务，确立了跨域知识迁移在实现真实世界通用操作中的核心地位。
技术上，π0.5采用混合多模态样本策略，将图像观测、语言指令、物体检测、语义子任务预测与底层动作统一建模。系统通过结合联合训练与多源异构数据流，解决了单一数据集难以覆盖开放世界多样性的瓶颈。这种融合高层语义理解与低层控制信号的架构设计，使模型具备了从非结构化环境到具体物理操作的强泛化能力，突破了传统具身智能的场景边界。
关键看点
- 异构联合训练：整合多机器人本体数据、Web资源及语义预测，通过混合多模态样本实现跨域知识迁移
- 混合多模态输入：统一处理图像、语言、检测框、语义子任务与底层动作，构建端到端通用操作表征
- 开放世界验证：首次在全新家庭环境成功执行厨房/卧室清洁等长程灵巧任务，突破实验室场景限制
- 架构演进路径：基于π0模型升级，证实高层语义预测与低层动作协同是解决野外泛化难题的关键机制
工程与产业观察
π0.5证明了异构数据联合训练是VLA模型走出实验室的有效路径。其将语义子任务预测嵌入端到端控制的范式，为解决具身智能长尾场景泛化提供了工程参考，表明未来通用机器人研发需重点关注跨本体数据融合与高层语义引导的低层控制对齐。
英文摘要 (Abstract)
In order for robots to be useful, they must perform practically relevant tasks in the real world, outside of the lab. While vision-language-action (VLA) models have demonstrated impressive results for end-to-end robot control, it remains an open question how far such models can generalize in the wild. We describe $π_{0.5}$, a new model based on $π_{0}$ that uses co-training on heterogeneous tasks to enable broad generalization. $π_{0.5}$\ uses data from multiple robots, high-level semantic prediction, web data, and other sources to enable broadly generalizable real-world robotic manipulation. Our system uses a combination of co-training and hybrid multi-modal examples that combine image observations, language commands, object detections, semantic subtask prediction, and low-level actions. Our experiments show that this kind of knowledge transfer is essential for effective generalization, and we demonstrate for the first time that an end-to-end learning-enabled robotic system can perform long-horizon and dexterous manipulation skills, such as cleaning a kitchen or bedroom, in entirely new homes.
作者：Physical Intelligence, Kevin Black, Noah Brown, James Darpinian, Karan Dhabalia, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, Manuel Y. Galliker, Dibya Ghosh, Lachy Groom, Ka…
来源
本页事实对应的公开页面，以原文为准。