外观
论文arXiv 2610.0659823 天前
SimForcing：将仿真运动先验蒸馏到真实域机器人世界模型中
原文链接：SimForcing: Distilling Simulation Motion Priors into Real-Domain Robot World Models（arXiv 2610.06598 · 发布日期：2026-10-05）
SimForcing: Distilling Simulation Motion Priors into Real-Domain Robot World Models
SimForcing通过仿真运动蒸馏与仿真条件引导，生成真实域机器人视频，并在Bridge等数据集验证。
核心要点
- 异构机器人视频中同时学习精确轨迹响应与真实视觉动态仍具挑战。
- 从仿真教师进行潜在运动蒸馏，在潜空间对齐时序变化以迁移运动先验。
- 采用多块仿真条件、条件丢弃和classifier-free guidance，降低对仿真预测准确性的过度依赖。
- Bridge多指标最佳，InternData-A1与LIBERO显示可用；代码给出GitHub链接。
验证范围与局限：验证包括Bridge、InternData-A1机器人数据集及LIBERO下游策略学习。
英文摘要 (Abstract)
Action-conditioned robot world models must respond precisely to robot trajectories while preserving realistic visual dynamics, yet learning both from heterogeneous robot videos remains challenging. Simulation offers structured motion supervision, but appearance differences hinder direct transfer, and inaccurate simulation predictions can misguide real-video generation. We present SimForcing, a simulation-guided framework that uses simulation both as a source of transferable motion knowledge and as a controllable reference for prediction. First, we transfer motion knowledge from a simulation teacher through latent-motion distillation, aligning temporal changes in latent space to internalize motion priors while mitigating the influence of appearance differences. Second, we introduce multi-block simulation conditioning with condition dropout to exploit predicted simulation trajectories without relying excessively on their accuracy. Our simulation-conditioning classifier-free guidance scheme unifies these two ideas by balancing predictions based on internalized motion knowledge with those additionally guided by simulation latents. The jointly trained student generates both simulation conditions and real-domain videos, requiring no additional world model at inference. On Bridge, SimForcing achieves the best PSNR, SSIM, LPIPS, and FVD among the compared methods without external embodied pretraining. Evaluation on InternData-A1 further supports its applicability across robot datasets. Moreover, using our trained world model to initialize a vision-language-action model improves LIBERO success, suggesting its utility for downstream policy learning. \url{https://github.com/Wang-Xiaodong1899/SimForcing}
作者：Xiaodong Wang, Tianle Li, Chuanxin Song, Junliang Xie, Zhanmi Zhong, Wu, Suiying, Peixi Peng
来源
本页事实对应的公开页面，以原文为准。