README.md

ContactInsert

This repository has the training and deployment code as well as trained models for the paper

"Zero-Shot Transfer of Haptics-based Object Insertion Policies" - Samarth Brahmbhatt, Ankur Deka, Andrew Spielberg, and Matthias Müller, ICRA 2023.

The code is divided into directories.

• Simulation training:
  • rl (entry point): Simulation environment and training framework, including reproduction of the ICRA 2023 paper.
  • robosuite: Fork of robosuite which implements policy inference delay randomization.
• Real robot deployment:
  • ros_controllers (entry point): ROS package implementing operational space control for the Frank Emika Panda robot.
  • ros_drivers_utils: Soft Robotics mGrip gripper ROS driver, and some other utilities.
  • arduino: Soft Robotics mGrip gripper control circuit and Arduino code.

Code contributors

• Samarth Brahmbhatt
• Ankur Deka
• Andrew Spielberg

Citation

@InProceedings{Brahmbhatt_2023_ICRA,
author = {Brahmbhatt, Samarth and Deka, Ankur and Spielberg, Andrew and M{\"u}ller, Matthias},
title = {Zero-Shot Transfer of Haptics-based Object Insertion Policies},
booktitle = {International Conference on Robotics and Automation (ICRA)},
month = {June},
year = {2023}
}