Learning from Visual Observation via Offline Pretrained State-to-Go Transformer

TL;DR

This paper introduces a two-stage offline learning framework using a pretrained State-to-Go Transformer to derive policies solely from visual observation data, outperforming baselines and matching reward-based policies in some tasks.

Contribution

The paper presents a novel two-stage framework with a pretrained State-to-Go Transformer for learning from visual data without task-specific information.

Findings

Outperforms baselines on Atari and Minecraft benchmarks.

Achieves performance comparable to reward-based policies in some tasks.

Demonstrates the potential of video-only data for visual reinforcement learning.

Abstract

Learning from visual observation (LfVO), aiming at recovering policies from only visual observation data, is promising yet a challenging problem. Existing LfVO approaches either only adopt inefficient online learning schemes or require additional task-specific information like goal states, making them not suited for open-ended tasks. To address these issues, we propose a two-stage framework for learning from visual observation. In the first stage, we introduce and pretrain State-to-Go (STG) Transformer offline to predict and differentiate latent transitions of demonstrations. Subsequently, in the second stage, the STG Transformer provides intrinsic rewards for downstream reinforcement learning tasks where an agent learns merely from intrinsic rewards. Empirical results on Atari and Minecraft show that our proposed method outperforms baselines and in some tasks even achieves performance comparable to the policy learned from environmental rewards. These results shed light on the potential of utilizing video-only data to solve difficult visual reinforcement learning tasks rather than relying on complete offline datasets containing states, actions, and rewards. The project's website and code can be found at https://sites.google.com/view/stgtransformer.