Orienting Novel 3D Objects Using Self-Supervised Learning of Rotation Transforms

TL;DR

This paper introduces a self-supervised deep learning approach to estimate 3D object rotations from depth images, enabling automated object orientation in packing and assembly tasks with high accuracy in simulation and real-world experiments.

Contribution

A novel self-supervised training method for neural networks to predict 3D rotations from depth images, facilitating autonomous object re-orientation.

Findings

Median rotation error of 1.47° in simulation

Median error of 4.2° on physical objects

Successful rotation of unseen objects up to 30°

Abstract

Orienting objects is a critical component in the automation of many packing and assembly tasks. We present an algorithm to orient novel objects given a depth image of the object in its current and desired orientation. We formulate a self-supervised objective for this problem and train a deep neural network to estimate the 3D rotation as parameterized by a quaternion, between these current and desired depth images. We then use the trained network in a proportional controller to re-orient objects based on the estimated rotation between the two depth images. Results suggest that in simulation we can rotate unseen objects with unknown geometries by up to 30{\deg} with a median angle error of 1.47{\deg} over 100 random initial/desired orientations each for 22 novel objects. Experiments on physical objects suggest that the controller can achieve a median angle error of 4.2{\deg} over 10 random initial/desired orientations each for 5 objects.