On the Efficacy of 3D Point Cloud Reinforcement Learning

TL;DR

This paper systematically compares 3D point cloud and 2D visual representations in reinforcement learning, demonstrating that 3D point clouds offer significant advantages in tasks requiring relationship reasoning.

Contribution

It provides a comprehensive analysis of 3D point cloud RL, introduces a robust algorithm, and highlights the benefits of 3D representations over 2D in robotic tasks.

Findings

3D point cloud RL outperforms 2D in relationship reasoning tasks

A robust 3D RL algorithm was developed for robotic manipulation

3D representations provide a beneficial inductive bias in visual RL

Abstract

Recent studies on visual reinforcement learning (visual RL) have explored the use of 3D visual representations. However, none of these work has systematically compared the efficacy of 3D representations with 2D representations across different tasks, nor have they analyzed 3D representations from the perspective of agent-object / object-object relationship reasoning. In this work, we seek answers to the question of when and how do 3D neural networks that learn features in the 3D-native space provide a beneficial inductive bias for visual RL. We specifically focus on 3D point clouds, one of the most common forms of 3D representations. We systematically investigate design choices for 3D point cloud RL, leading to the development of a robust algorithm for various robotic manipulation and control tasks. Furthermore, through comparisons between 2D image vs 3D point cloud RL methods on both minimalist synthetic tasks and complex robotic manipulation tasks, we find that 3D point cloud RL can significantly outperform the 2D counterpart when agent-object / object-object relationship encoding is a key factor.