Hugging Face 将人形机器人接入 LeRobot
原标题：Bringing Humanoids to LeRobot
AI 摘要
Hugging Face 博客介绍如何将人形机器人接入 LeRobot 全栈开源机器人学习库，以 Unitree G1 为平台，采用 OpenHLM 方案：π0.5 策略根据语言、相机观测和机器人状态预测 SONIC 潜在运动 token，再由运行在 G1 上的快速全身控制器解码为 29 自由度动作。团队通过 VR 遥操作采集约 100 条、71 分钟的全身演示数据，微调 pi05_base 得到抓放罐头的策略，并开源了 3D 打印硬件改装件。文章还展示了用 LoRA + PPO 在 SONIC 上训练避球策略的实验。
正文节选
A π0.5 policy predicts SONIC latent motion tokens from language prompt, robot state, and camera observations. A fast controller running on the G1 decodes those tokens into stable 29-DoF whole-body motion. This post shows how we connect open-source hardware, teleoperation, datasets, and policy learning in LeRobot to build that workflow on the Unitree G1. You’ll learn how demonstrations become a π0.5 policy that predicts SONIC motion tokens, and how a fast whole-body controller turns those tokens