Learning to Drive by Watching YouTube Videos: Action-Conditioned Contrastive Policy Pretraining

TL;DR

This paper introduces a novel contrastive pretraining method for visuomotor policies in driving tasks, leveraging uncurated YouTube videos to improve reinforcement and imitation learning performance.

Contribution

It proposes a new action-conditioned contrastive pretraining approach using pseudo labels from videos, enhancing policy learning without extensive environment interactions.

Findings

Outperforms previous unsupervised and ImageNet pretraining methods

Significantly improves downstream reinforcement learning results

Effective in imitation learning scenarios

Abstract

Deep visuomotor policy learning, which aims to map raw visual observation to action, achieves promising results in control tasks such as robotic manipulation and autonomous driving. However, it requires a huge number of online interactions with the training environment, which limits its real-world application. Compared to the popular unsupervised feature learning for visual recognition, feature pretraining for visuomotor control tasks is much less explored. In this work, we aim to pretrain policy representations for driving tasks by watching hours-long uncurated YouTube videos. Specifically, we train an inverse dynamic model with a small amount of labeled data and use it to predict action labels for all the YouTube video frames. A new contrastive policy pretraining method is then developed to learn action-conditioned features from the video frames with pseudo action labels. Experiments show that the resulting action-conditioned features obtain substantial improvements for the downstream reinforcement learning and imitation learning tasks, outperforming the weights pretrained from previous unsupervised learning methods and ImageNet pretrained weight. Code, model weights, and data are available at: https://metadriverse.github.io/ACO.