I3L: Iterative Interactive Imitation Learning for Whole-Body Mobile Manipulation

Key points

• Policies trained offline from teleoperated demonstrations can degrade outside their training distribution.1
• This gap is particularly acute in long-horizon whole-body mobile manipulation, where long horizons amplify compounding errors and coordinated arm, torso, and base motion makes exhaustive demonstration coverage impractical.1
• We evaluate I3L across three robot embodiments and eight simulated and real-world tasks, including four long-horizon whole-body mobile manipulation tasks.1
• Together, these results show that I3L helps close the last-mile gap between offline training and reliable deployment and provides practical insights into collecting and using corrections for long-horizon whole-body mobile manipulation.1

Sources (1)

• [1]I3L: Iterative Interactive Imitation Learning for Whole-Body Mobile Manipulation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 13, 05:28 AM

  We introduce I3L, an integrated hardware and learning system that combines whole-body correction with iterative policy refinement to address this last-mile reliability gap.

  Policies trained offline from teleoperated demonstrations can degrade outside their training distribution.

Extractive summary: sentences quoted from the sources.