Event
Proposes tactile in-context learning system Tactile-ETA
1 reports · 1 sources
Yubang Wang et al. propose the Tactile-ETA system, which integrates tactile experience into embodied task agents through in-context learning, enabling fine manipulation without updating model parameters. The agent combines expert demonstrations with current visual and tactile observations and proprioception, and selects actions through a general-purpose robot tool. In four simulated manipulation tasks, adding tactile observations to the same two expert demonstrations raised the average success rate from 15.75% to 24.00%; on a real-robot task, demonstrations with aligned vision and touch raised the success rate from 0% to 60%. Experiments also show that performance depends on the sensory content and representation of the demonstrations, and different agent models differ markedly on the Lift Bottle task.
Report timeline
- Tactile-ETA Achieves Fine Manipulation Through Tactile In-Context Learning
The Tactile-ETA system brings tactile experience into embodied task agents through in-context learning, completing fine manipulation without updating model parameters. Across four simulated manipulation tasks, adding tactile observations to the same two expert demonstrations raised average success from 15.75% to 24.00%; on a real robot task, vision- and tactile-aligned demonstrations raised success from 0% to 60%. Experiments also show that performance depends on the sensory content and representation of the demonstrations, and different agent models differ markedly on the Lift Bottle task.
arXiv cs.RO ·