$Ψ_0$：一种面向通用人形机器人运动- manipulation 的开放基础模型
$Ψ_0$: An Open Foundation Model Towards Universal Humanoid Loco-Manipulation
AI总结：
本文提出$Ψ_0$模型，通过分阶段训练和异构数据利用，实现高效的人形机器人运动- manipulation 任务，仅用800小时视频和30小时机器人数据即超越基线性能。
摘要：
我们介绍了$Ψ_0$（Psi-Zero），一种开放的基础模型，旨在解决具有挑战性的人形机器人运动- manipulation 任务。尽管现有方法通常尝试通过在大量多样的人类和人形机器人数据上共同训练来解决这一根本问题，但我们认为这种策略是次优的，因为人类和人形机器人之间存在根本的运动学和运动差异。因此，数据效率和模型性能仍然不够满意，尽管数据量相当大。为了解决这一挑战，\ours\;将学习过程解耦，以最大限度地利用异构数据源。具体而言，我们提出了一种分阶段的训练范式，具有不同的学习目标：首先，我们自回归地预训练一个VLM骨干网络，在大规模的以自身为中心的人类视频上，以获得具有通用性的视觉-动作表示。然后，我们后训练一个基于流的动作专家，在高质量的人形机器人数据上学习精确的机器人关节控制。我们的研究进一步识别了一个关键但常常被忽视的数据配方：与那些随着嘈杂互联网剪辑或异质跨身体机器人数据集扩增的方法不同，我们证明了在高质量的以自身为中心的人类 manipulation 数据上预训练，然后在特定领域的真实世界人形机器人轨迹上后训练，可以产生更优的性能。广泛的现实世界实验表明，\ours\仅使用约800小时的人类视频数据和30小时的真实世界机器人数据，就能在多个任务上以超过40%的总体成功率超越在超过10倍数据上预训练的基线。我们将开源整个生态系统，包括数据处理和训练管道、人形基础模型和实时动作推理引擎。
英文摘要：
We introduce $Ψ_0$ (Psi-Zero), an open foundation model to address challenging humanoid loco-manipulation tasks. While existing approaches often attempt to address this fundamental problem by co-training on large and diverse human and humanoid data, we argue that this strategy is suboptimal due to the fundamental kinematic and motion disparities between humans and humanoid robots. Therefore, data efficiency and model performance remain unsatisfactory despite the considerable data volume. To address this challenge, \ours\;decouples the learning process to maximize the utility of heterogeneous data sources. Specifically, we propose a staged training paradigm with different learning objectives: First, we autoregressively pre-train a VLM backbone on large-scale egocentric human videos to acquire generalizable visual-action representations. Then, we post-train a flow-based action expert on high-quality humanoid robot data to learn precise robot joint control. Our research further identifies a critical yet often overlooked data recipe: in contrast to approaches that scale with noisy Internet clips or heterogeneous cross-embodiment robot datasets, we demonstrate that pre-training on high-quality egocentric human manipulation data followed by post-training on domain-specific real-world humanoid trajectories yields superior performance. Extensive real-world experiments demonstrate that \ours\ achieves the best performance using only about 800 hours of human video data and 30 hours of real-world robot data, outperforming baselines pre-trained on more than 10$\times$ as much data by over 40\% in overall success rate across multiple tasks. We will open-source the entire ecosystem to the community, including a data processing and training pipeline, a humanoid foundation model, and a real-time action inference engine.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
