Skip to content

论文arXiv 2411.013492024-11-02

研究证实人形机器人扩散策略训练需大规模多样化数据集 ​

本文由 AI 根据公开内容整理,可能有误,请以原文为准:The Role of Domain Randomization in Training Diffusion Policies for Whole-Body Humanoid Control(arXiv 2411.01349)

核心内容 ​

针对扩散策略在人形机器人全身控制中的适用性问题,研究者利用IsaacGym仿真环境开展系统性评估。通过在不同域随机化条件下训练对抗性运动先验代理生成合成示教数据,对比了不同规模与多样性数据集对策略拟合的影响。实验结果表明,尽管扩散策略能够实现稳定的双足行走行为,但在运动控制场景下,其成功训练所需的数据量级和分布广度远超常规机械臂操作任务,即便在简单场景中该数据饥渴特性依然显著。

技术方案采用对抗性运动先验作为合成数据生成器,通过调节域随机化参数构建差异化训练集,以解耦数据多样性与策略性能的关系。研究聚焦于从人类演示或合成数据中蒸馏全身控制策略的瓶颈,揭示了当前扩散模型在处理高维动态 locomotion 任务时的样本效率短板。该发现强调了在具身智能算法迁移过程中,不能简单套用操作领域的经验,必须针对运动学特性重构数据采集与增强管线。

关键看点 ​

  • 数据需求差异:扩散策略用于人形运动控制时,所需数据集规模与多样性显著高于操作任务,简单场景亦不例外
  • 合成数据生成:基于IsaacGym环境利用对抗性运动先验代理,配合域随机化技术批量生成高质量合成示教数据
  • 策略验证结论:扩散策略可实现人形机器人稳定行走,但训练成功率高度依赖数据集的覆盖广度与样本数量
  • Sim-to-Real启示:揭示了从高维人类演示数据蒸馏全身控制策略的难点,指明数据增强是提升样本效率的关键

工程与产业观察 ​

该研究为人形机器人模仿学习提供了关键工程基准:扩散策略在locomotion领域存在显著的数据规模门槛。开发者在复用操作类算法时,需重新评估数据管线,优先通过域随机化扩大合成数据分布,而非单纯增加模型参数,这对解决真实世界人形机器人运动策略的泛化难题具有直接指导意义。

英文摘要 (Abstract) ​

Humanoids have the potential to be the ideal embodiment in environments designed for humans. Thanks to the structural similarity to the human body, they benefit from rich sources of demonstration data, e.g., collected via teleoperation, motion capture, or even using videos of humans performing tasks. However, distilling a policy from demonstrations is still a challenging problem. While Diffusion Policies (DPs) have shown impressive results in robotic manipulation, their applicability to locomotion and humanoid control remains underexplored. In this paper, we investigate how dataset diversity and size affect the performance of DPs for humanoid whole-body control. In a simulated IsaacGym environment, we generate synthetic demonstrations by training Adversarial Motion Prior (AMP) agents under various Domain Randomization (DR) conditions, and we compare DPs fitted to datasets of different size and diversity. Our findings show that, although DPs can achieve stable walking behavior, successful training of locomotion policies requires significantly larger and more diverse datasets compared to manipulation tasks, even in simple scenarios.

作者:Oleg Kaidanov, Firas Al-Hafez, Yusuf Suvari, Boris Belousov, Jan Peters

来源

本页事实对应的公开页面,以原文为准。

  1. arXiv 2411.01349:The Role of Domain Randomization in Training Diffusion Policies for Whole-Body Humanoid Control访问 2026-10-23