Diffusion Augmented Agents: A Framework for Efficient Exploration and Transfer Learning

TL;DR

Diffusion Augmented Agents (DAAG) utilize diffusion models and large language models to enhance sample efficiency and transfer learning in reinforcement learning for embodied agents, enabling autonomous experience relabeling and lifelong learning.

Contribution

Introduction of DAAG framework that combines diffusion models and language models for autonomous experience relabeling and improved reinforcement learning efficiency.

Findings

DAAG improves reward detector training efficiency.

DAAG enhances transfer of past experiences to new tasks.

DAAG demonstrates significant sample efficiency gains in robotics simulations.

Abstract

We introduce Diffusion Augmented Agents (DAAG), a novel framework that leverages large language models, vision language models, and diffusion models to improve sample efficiency and transfer learning in reinforcement learning for embodied agents. DAAG hindsight relabels the agent's past experience by using diffusion models to transform videos in a temporally and geometrically consistent way to align with target instructions with a technique we call Hindsight Experience Augmentation. A large language model orchestrates this autonomous process without requiring human supervision, making it well-suited for lifelong learning scenarios. The framework reduces the amount of reward-labeled data needed to 1) finetune a vision language model that acts as a reward detector, and 2) train RL agents on new tasks. We demonstrate the sample efficiency gains of DAAG in simulated robotics environments involving manipulation and navigation. Our results show that DAAG improves learning of reward detectors, transferring past experience, and acquiring new tasks - key abilities for developing efficient lifelong learning agents. Supplementary material and visualizations are available on our website https://sites.google.com/view/diffusion-augmented-agents/