ForeTime-VLA: Causal Future-Token Distillation from a World Action Model for Conveyor-Belt Manipulation
作者: Siyuan Ma, Yutian Zhang, Boshi Zhang, Qinglian Wu, Jiaqi Zhai, Dong Wei, Xiaojin Huang
分类: cs.AI, cs.RO
发布日期: 2026-08-21
备注: 8 pages, 5 figures. Introduces ForeTime-VLA, a causal future-token distillation method for conveyor-belt manipulation from a frozen world action model teacher
💡 一句话要点
提出ForeTime-VLA以解决动态物体操控中的未来预测问题
🎯 匹配领域: 支柱一：机器人控制 (Robot Control) 支柱二：RL算法与架构 (RL & Architecture) 支柱九：具身大模型 (Embodied Foundation Models)
关键词: 动态物体操控 未来预测 视觉-语言-动作 世界动作模型 蒸馏训练 机器人抓取 因果推理
📋 核心要点
- 现有的视觉-语言-动作策略在动态物体操控中面临预测接触事件的挑战，通常依赖当前观察进行微调，缺乏对未来状态的有效预测。
- 本文提出ForeTime-VLA，通过从冻结的Fast-WAM教师中蒸馏出未来感知的表示，结合历史帧信息，增强了对动态操控的预测能力。
- 在实验中，ForeTime-VLA在真实机器人评估中实现了81.1%的静态和58.9%的慢速抓取成功率，超越了对比基线，显示出显著的性能提升。
📝 摘要（中文）
操控移动物体需要预测接触事件的策略，但现有的视觉-语言-动作（VLA）策略通常仅基于当前观察进行微调。世界动作模型（WAMs）能够学习预测动态，但在部署时运行视频规模的教师模型或显式想象未来帧的成本较高。本文提出了ForeTime-VLA，这是一种稠密的pi0.5策略，从冻结的Fast-WAM派生教师中蒸馏出未来感知的、动作等价的表示，同时在推理时保持因果性。实验表明，ForeTime-VLA在动态操控任务中显著提高了抓取成功率。
🔬 方法详解
问题定义：本文旨在解决动态物体操控中对未来接触事件的预测问题。现有方法通常仅依赖当前观察，导致在复杂动态环境中的表现不足。
核心思路：ForeTime-VLA通过从冻结的Fast-WAM教师中蒸馏出未来感知的表示，结合历史帧信息，增强了对未来状态的预测能力，从而提高了操控策略的有效性。
技术框架：该方法的整体架构包括离线和在线两个阶段。离线阶段将当前和未来的视频潜在表示压缩为64维目标；在线阶段则通过八帧历史编码器预测该目标，并结合操控阶段和归一化的过渡时间进行决策。
关键创新：ForeTime-VLA的主要创新在于因果未来令牌蒸馏，允许在不依赖世界模型教师的情况下，提升动态操控的性能。这一设计使得模型在推理时保持因果性，避免了高昂的计算成本。
关键设计：训练过程中保留了原始的流匹配动作目标，并引入了余弦、关系几何、阶段、过渡时间和动作等价性等多个目标函数，以增强模型的学习效果。
🖼️ 关键图片
📊 实验亮点
在实验中，ForeTime-VLA在真实机器人评估中实现了81.1%的静态抓取成功率和58.9%的慢速抓取成功率，分别比对比基线提高了12.2和22.2个百分点。此外，在不同带速下，ForeTime-VLA完成了44/90次抓取，而对比方法仅完成了23/90次，显示出显著的性能优势。
🎯 应用场景
ForeTime-VLA的研究成果在动态物体操控领域具有广泛的应用潜力，尤其是在工业自动化、物流和机器人抓取等场景中。通过提高对未来状态的预测能力，该方法能够显著提升机器人在复杂环境中的操作效率和成功率，推动智能机器人技术的发展。
📄 摘要（原文）
Manipulating moving objects requires a policy to anticipate contact events, yet vision-language-action (VLA) policies are commonly fine-tuned from the current observation alone. World action models (WAMs) learn predictive dynamics, but running a video-scale teacher or explicitly imagining future frames at deployment is costly. We introduce ForeTime-VLA, a dense pi0.5 policy that distills a future-aware, action-equivalent representation from a frozen Fast-WAM-derived teacher while remaining causal at inference. Offline, current and future video latents are compressed into a whitened 64-D target. Online, an eight-frame history encoder predicts this target together with manipulation phase and normalized time-to-transition. Four future tokens and one phase token condition the VLM prefix, while the predicted future and transition horizon condition the action expert. Training retains the original flow-matching action target and adds cosine, relational geometry, phase, time-to-transition, and action-equivalence objectives. On a deduplicated conveyor-belt dataset, we compare 40k-step checkpoints on 768 matched windows per split. Test MAE decreases from 0.134119 to 0.130593 (2.63%; paired-bootstrap 95% CI: 0.82-4.48% improvement), and test L2 decreases by 3.02%, at a 2.46-2.93% latency cost. In quantitative real-robot evaluation, ForeTime-VLA achieves 81.1% stationary and 58.9% slow-moving grasp success, exceeding the next-best reference by 12.2 and 22.2 percentage points, respectively. Across three belt speeds, it completes 44/90 grasps versus 23/90 for pi0.5, including 11/30 versus 2/30 at fast speed. The agreement between offline orientation gains and reduced real-robot contact-pose failures supports causal future-token distillation as an effective way to improve dynamic manipulation without deploying the world-model teacher.