Dual-Granularity Contrastive Reward via Generated Episodic Guidance for Efficient Embodied RL

TL;DR

This paper introduces DEG, a novel reward framework that uses generated episodic guidance from video models to improve sample efficiency in embodied RL without human annotations.

Contribution

DEG leverages large video generation models and expert videos to create dense, task-specific rewards, enhancing RL efficiency and stability without extensive supervision.

Findings

DEG accelerates discovery of success in diverse tasks

It improves policy stability and convergence

Effective in both simulation and real-world environments

Abstract

Designing suitable rewards poses a significant challenge in reinforcement learning (RL), especially for embodied manipulation. Trajectory success rewards are suitable for human judges or model fitting, but the sparsity severely limits RL sample efficiency. While recent methods have effectively improved RL via dense rewards, they rely heavily on high-quality human-annotated data or abundant expert supervision. To tackle these issues, this paper proposes Dual-granularity contrastive reward via generated Episodic Guidance (DEG), a novel framework to seek sample-efficient dense rewards without requiring human annotations or extensive supervision. Leveraging the prior knowledge of large video generation models, DEG only needs a small number of expert videos for domain adaptation to generate dedicated task guidance for each RL episode. Then, the proposed dual-granularity reward that balances coarse-grained exploration and fine-grained matching, will guide the agent to efficiently approximate the generated guidance video sequentially in the contrastive self-supervised latent space, and finally complete the target task. Extensive experiments on 18 diverse tasks across both simulation and real-world settings show that DEG can not only serve as an efficient exploration stimulus to help the agent quickly discover sparse success rewards, but also guide effective RL and stable policy convergence independently.