佐治亚理工学院徐丹飞：别被「视觉生成」骗了，视频预测≠机器人规划｜RSS 2026
Merged summary
TL;DR — 徐丹飞指出，逼真的视频预测并不等于可执行的机器人规划；其组合式世界模型通过模块化技能、因子图和自适应时序注意力缩小“视频—动作鸿沟”。相关讨论也强调，机器人能力不仅取决于更大的模型与数据，顺应性硬件同样能显著简化感知和控制。
- 因子图可在测试时组合独立学习的技能，并施加空间、时间及多机械臂约束。
- “Temporal Ratio”衡量策略对预测未来帧与当前观测的关注比例：接近阶段侧重未来轨迹，接触和抓取阶段转向当前视觉与触觉反馈。
- 视频数据远多于机器人动作数据，造成视频模型泛化能力与动作执行能力之间的差距。
- RBO欠驱动硅胶手利用形态顺应性和环境接触完成稳健抓取，相同开环指令即可适应不同形状及易碎物体。
- 当被动顺应性不足时，新原型仅补充低成本声学传感，从而继续降低复杂感知和控制的需求。
注： 两则来源侧重点明显不同：一则介绍徐丹飞的视频预测与机器人规划研究，另一则讨论RSS 2026获奖的RBO气动手，现有摘要不足以确认二者属于同一项工作。
Sources (2)
佐治亚理工学院徐丹飞：别被「视觉生成」骗了，视频预测≠机器人规划｜RSS 2026
TL;DR - Georgia Tech’s Danfei Xu argues that realistic video prediction does not constitute robotic planning. His compositional world-model approach uses modular skills, factor graphs, and adaptive temporal attention to bridge predicted futures and executable actions.
- Factor graphs compose independently learned skills at test time while enforcing spatial, temporal, and multi-arm constraints.
- A “video-action gap” arises because abundant video data supports stronger generalization than scarce robot-action data.
- The Temporal Ratio measures whether control attends more to predicted future frames or the current observation.
- Policy guidance emphasizes future trajectories during reaching, then current visual and tactile feedback during contact and grasping.
RSS 2026 时间检验奖：形态即计算，具身智能的“逻辑刹车”已至
TL;DR - RSS 2026 honored the low-cost RBO pneumatic hand with its Test of Time Award, highlighting how compliant hardware can perform robust grasping without complex sensing or control. The work challenges embodied-AI approaches that rely primarily on larger models and more training data.
- The underactuated silicone hand adapts its grasp through physical compliance, effectively using morphology as computation.
- It deliberately exploits contact with tables and other environmental constraints rather than treating all collisions as errors.
- Identical open-loop commands can grasp objects of different shapes and fragility, reducing sensing and control requirements.
- Newer prototypes add inexpensive acoustic sensing only when passive compliance alone is insufficient.