为人形机器人像素到动作策略迁移打开仿真到现实之门
Opening the Sim-to-Real Door for Humanoid Pixel-to-Action Policy Transfer
论文展示了仅凭 RGB 感知、完全在仿真中训练的人形机器人全身 loco-manipulation 策略如何零样本迁移到真实铰接物体（如多种门）上；其中分阶段重置探索与基于 GRPO 的微调可用于缓解视觉部分可观测性、提升闭环一致性，对做视觉-动作策略 sim-to-real 的研究者是可复用的训练配方。
原文依据 the first humanoid sim-to-real policy capable of diverse articulated loco-manipulation from pure RGB perception
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论近期，GPU 加速的逼真仿真进展为机器人学习开辟了一条可扩展的数据生成路径，其中大规模的物理与视觉随机化使策略能够泛化到精心策划的环境之外。基于这些进展，我们开发了一个教师-学生-自举学习框架，用于基于视觉的人形机器人移动操作，并以铰接物体交互作为具有代表性的高难度基准。我们的方法引入了分阶段重置探索策略，以稳定长时程特权策略的训练；并设计了基于 GRPO 的微调流程，旨在缓解部分可观测性并提升仿真到现实强化学习中的闭环一致性。完全在合成仿真数据上训练得到的策略，在多种铰接物体（包括多种门型）上实现了鲁棒的零样本性能，并在相同的全身控制栈下，任务完成时间最多优于人类遥操作者 31.7%。这是首个能够仅凭纯 RGB 感知完成多样化铰接物体移动操作的人形机器人仿真到现实策略。
英文原文摘要 1116 字符
Recent progress in GPU-accelerated, photorealistic simulation has opened a scalable data-generation path for robot learning, where massive physics and visual randomization allow policies to generalize beyond curated environments. Building on these advances, we develop a teacher-student-bootstrap learning framework for vision-based humanoid loco-manipulation, using articulated-object interaction as a representative high-difficulty benchmark. Our approach introduces a staged-reset exploration strategy that stabilizes long-horizon privileged-policy training, and a GRPO-based fine-tuning procedure designed to mitigate partial observability and improve closed-loop consistency in sim-to-real RL. Trained entirely on synthetic simulation data, the resulting policy achieves robust zero-shot performance across diverse articulated objects--including multiple door types--and outperforms human teleoperators by up to 31.7% in task completion time under the same whole-body control stack. This represents the first humanoid sim-to-real policy capable of diverse articulated loco-manipulation from pure RGB perception.