原帖引用 · AI 翻译
Rohan Paul@rohanpaul_ai
英伟达发布了 GR00T 2,这是一款可控制机械臂、机器人手和人形机器人的开放世界动作模型。
机器人通常通过将每一帧摄像头画面与精确电机指令配对的记录来学习,而这些记录只能由人操控真实机器人获得。
普通人类活动视频便宜得多,也丰富得多,但没有电机指令,因此早期模型几乎无法利用它。
GR00T 2 通过预训练缓解了这一问题:训练数据包括 402,730 小时的人类头戴摄像头视频、10,184 小时的手持夹爪演示,但遥操作机器人数据只有 23,500 小时。
原文
Nvidia released GR00T 2, an open world-action model that runs arms, robot hands and humanoids. Robots normally learn from recordings that pair each camera frame with exact motor commands, and those only come from a person steering a real robot. Ordinary human video is far cheaper and more plentiful, but it has no motor commands, so earlier models could barely use it. GR00T 2 eases that by pretraining on 402,730 hours of human head-camera video plus 10,184 hours of handheld-gripper demos, but only 23,500 teleoperated robot hours.
· UTC+8