Skip to content

论文arXiv 2611.236203 天前

冻结世界模型反向推理实现动作自校正,机械臂到达误差降至3-5mm ​

本文由 AI 根据公开内容整理,可能有误,请以原文为准:How Uncertain Is My Action? Self-Correcting Visual Planning with Frozen World Models(arXiv 2611.23620)

核心内容 ​

针对冻结潜在世界模型在部署时无法感知执行动作与指令偏差的问题,研究提出一种自校正视觉规划框架。该方法无需重新训练模型,即可在PushT、TwoRoom及OGBench-Cube等基准任务中显著提升性能。实验显示,其偏移检测AUROC与理论预测偏差小于0.05,PushT任务成功率从0.650提升至0.862,Cube任务从0.378升至0.432,达到获知真实偏移的Oracle水平,并优于五种主流测试时自适应方法。

技术核心在于将冻结世界模型反向运行作为滤波器:对假设的动作偏移生成预期观测,与实际观测比对得出仅含偏移分量的“偏移分数”,避免残差能量混淆模型自身误差。该分数支持局部最优检测、Cramer-Rao误差界估计及主导生成器识别。在真实机械臂上,该机制成功识别出关节编码器未记录的基座偏移,使到达误差从36mm降至3-5mm;在随机密封偏移测试中,误差恢复率达92%-94%,验证了其在物理系统中的鲁棒性。

关键看点 ​

  • 自校正机制:反向运行冻结世界模型预测偏移下的预期观测,构建纯偏移分数以分离模型误差,触发报警后补偿规划器
  • 仿真性能跃升:PushT任务成功率由0.650提至0.862,Cube任务由0.378升至0.432,持平Oracle且优于五种TTA方法
  • 真机误差抑制:识别编码器未检出的基座关节偏移,机械臂到达误差从36mm降至3-5mm,误差恢复率达92%-94%
  • 统计检测精度:偏移检测遵循局部最优检验,AUROC实测值与理论预测偏差小于0.05,并提供Cramer-Rao误差界估计

工程与产业观察 ​

该工作证明冻结世界模型可通过反向推理充当动作接口不确定性滤波器,在不微调前提下解决具身智能部署中的执行-感知失配问题。其为低成本提升存量模型鲁棒性提供新范式,尤其适用于硬件老化或标定漂移场景,对降低真机运维成本具有直接工程价值。

英文摘要 (Abstract) ​

A frozen latent world model combined with sampling-based model predictive control is a visual planner. At deployment, the executed action can depart from the commanded one, and the frozen model has no way to notice. We show that the model can be run in reverse: for any hypothesized actuation shift, it predicts what it should have observed under that shift, and comparing this prediction with what was actually observed quantifies the shift and lets the planner correct itself. Unlike the residual energy, which mixes the shift with the model's own error, the score keeps only the part of the residual that a shift can produce. This shift score detects the shift with a locally optimal test, estimates it with Cramer-Rao error bars, names the dominant generator, and compensates the planner once an alarm fires. On PushT, TwoRoom and OGBench-Cube with a frozen latent world model, measured detection follows the prediction to within 0.05 AUROC. Self-correction raises success from 0.650 to 0.862 on PushT and from 0.378 to 0.432 on the harder Cube task, levels with an oracle that is told the true shift at the same block, better than five common test-time adaptation methods. On a real robotic arm, the score identifies a base-joint offset that no joint encoder registers and reduces the reach error from 36 mm to 3-5 mm, and under shifts drawn at random and sealed before the session, estimates recover 92-94% of the error reduction achieved with the true shift. The broader implication is that a frozen world model can act as a filter by predicting the future, testing action-interface uncertainty and validity, and finally correcting failures.

作者:Siqing Li, Luming Yang, Yijia Jiao, Junze Deng, Alper Yilmaz

基于本页技术方向关联的整机参数、核心零部件、实测报告与同赛道选型横评。

外部来源

本页事实对应的公开页面,以原文为准。(点击将离开本站)

  1. arXiv 2611.23620:How Uncertain Is My Action? Self-Correcting Visual Planning with Frozen World Models ↗访问 2026-11-19