传统机器人抓取模型经常和具体机械手绑定,换一种手就要重新训练。8月12日的《Nature Machine Intelligence》发表了一篇文章,提出一种“通用抓取”方案:首先学习物体上哪些位置适合接触和抓取,这与机械手硬件无关;之后再由规划器根据机械手具体形状生成动作。因此,一个模型无需重新训练,就能零样本迁移到 7种不同机械手、2—5根手指。
参考文献:Wang, X., Tam, L.M. & Xu, Q. Learning contact representations in real-world clutter for universal robotic grasping. Nat Mach Intell (2026).
![]()
特别声明:以上内容(如有图片或视频亦包括在内)为自媒体平台“网易号”用户上传并发布,本平台仅提供信息存储服务。
Notice: The content above (including the pictures and videos if any) is uploaded and posted by a user of NetEase Hao, which is a social media platform and only provides information storage services.