From Wearable Interfaces to Dexterous Policies: Contact Shifts and Tactile Representations

Key points

• Interface geometry can influence both how a task is performed and what tactile observations are recorded for learning.1
• We study two versions of a DexUMI-family exoskeleton that share the same robot command definition, mapping procedure, and tactile module type but differ in hand-side geometry.1
• Lid twisting primarily exhibits a change in contact location, whereas egg carton opening primarily exhibits a change in contact occurrence.1
• These results show that wearable collection hardware is part of the data-generation process for robot learning.1

Sources (1)

• [1]From Wearable Interfaces to Dexterous Policies: Contact Shifts and Tactile Representations
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 03:22 AM

  Unlike conventional teleoperation, wearable interfaces allow operators to collect dexterous demonstrations through their own hand motions while directly interacting with task objects.

  Interface geometry can influence both how a task is performed and what tactile observations are recorded for learning.

Extractive summary: sentences quoted from the sources.