Zero-Shot Transfer of Haptics-based Object Insertion Policies

Abstract

Humans naturally exploit haptic feedback during contact-rich tasks like loading a dishwasher or stocking a bookshelf. Current robotic systems focus on avoiding unexpected contact. In this paper we train a contact-exploiting manipulation policy in simulation for the contact-rich household task of loading plates into a slotted holder, which transfers without any fine-tuning to the real robot. We investigate various factors necessary for this zero-shot transfer, like time delay modeling, memory representation, and domain randomization. Our policy transfers with minimal sim-to-real gap and significantly outperforms heuristic and learnt baselines. It also generalizes to plates of different sizes and weights.

Resources

arXiv: 2301.12587

Video

Citation

@inproceedings{brahmbhatt2023zero,
  title     = {Zero-Shot Transfer of Haptics-based Object Insertion Policies},
  author    = {Brahmbhatt, Samarth and Deka, Ankur and Spielberg, Andrew and M{\"{u}}ller, Matthias},
  booktitle = {IEEE International Conference on Robotics and Automation (ICRA)},
  year      = {2023}
}