Visual Reinforcement Learning with Self-Supervised 3D Representations
Preprint
Abstract
A prominent approach to visual Reinforcement Learning (RL) is to learn an internal state representation using self-supervised methods, which has the potential benefit of improved sample-efficiency and generalization through additional learning signal and inductive biases. However, while the real world is inherently 3D, prior efforts have largely been focused on leveraging 2D computer vision techniques as auxiliary self-supervision. In this work, we present a unified framework for self-supervised learning of 3D representations for motor control. Our proposed framework consists of two phases: a pretraining phase where a deep voxel-based 3D autoencoder is pretrained on a large object-centric dataset, and a finetuning phase where the representation is jointly finetuned together with RL on in-domain data. We empirically show that our method enjoys improved sample efficiency in simulated manipulation tasks compared to 2D representation learning methods. Additionally, our learned policies transfer zero-shot to a real robot setup with only approximate geometric correspondence, and successfully solve motor control tasks that involve grasping and lifting from a single, uncalibrated RGB camera. Code and videos are available at https://yanjieze.com/3d4rl/ .Keywords
All Related Versions
This publication has 14 references indexed in Scilit:
- Look Closer: Bridging Egocentric and Third-Person Views With Transformers for Robotic ManipulationIEEE Robotics and Automation Letters, 2022
- Video Autoencoder: self-supervised disentanglement of static 3D structure and motion2021
- Common Objects in 3D: Large-Scale Learning and Evaluation of Real-life 3D Category Reconstruction2021
- Generalization in Reinforcement Learning by Soft Data Augmentation2021
- DROID: Minimizing the Reality Gap Using Single-Shot Human DemonstrationIEEE Robotics and Automation Letters, 2021
- 6-PACK: Category-level 6D Pose Tracker with Anchor-Based Keypoints2020
- Learning Precise 3D Manipulation from Multiple Uncalibrated Cameras2020
- Multi-View PointNet for 3D Scene Understanding2019
- Learning Spatial Common Sense With Geometry-Aware Recurrent Networks2019
- Least-squares estimation of transformation parameters between two point patternsIEEE Transactions on Pattern Analysis and Machine Intelligence, 1991