CORN: Contact-based Object Representation for Nonprehensile Manipulation of General Unseen Objects

TL;DR

This paper introduces CORN, a contact-based object representation and pretraining method using a patch-based transformer, enabling efficient zero-shot transfer of nonprehensile manipulation policies to unseen objects.

Contribution

It presents a novel contact-based object representation and a scalable pretraining pipeline that improves generalization and efficiency in nonprehensile manipulation tasks.

Findings

Effective zero-shot transfer to real-world objects

Significant reduction in training time and data requirements

Outperforms previous shape representation methods

Abstract

Nonprehensile manipulation is essential for manipulating objects that are too thin, large, or otherwise ungraspable in the wild. To sidestep the difficulty of contact modeling in conventional modeling-based approaches, reinforcement learning (RL) has recently emerged as a promising alternative. However, previous RL approaches either lack the ability to generalize over diverse object shapes, or use simple action primitives that limit the diversity of robot motions. Furthermore, using RL over diverse object geometry is challenging due to the high cost of training a policy that takes in high-dimensional sensory inputs. We propose a novel contact-based object representation and pretraining pipeline to tackle this. To enable massively parallel training, we leverage a lightweight patch-based transformer architecture for our encoder that processes point clouds, thus scaling our training across thousands of environments. Compared to learning from scratch, or other shape representation baselines, our representation facilitates both time- and data-efficient learning. We validate the efficacy of our overall system by zero-shot transferring the trained policy to novel real-world objects. Code and videos are available at https://sites.google.com/view/contact-non-prehensile.