arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2406.00645cs.LGcs.AIcs.CV
收藏

FuRL:视觉-语言模型作为强化学习的模糊奖励

FuRL: Visual-Language Models as Fuzzy Rewards for Reinforcement Learning

McGill University(麦吉尔大学);Horizon Robotics

Yuwei Fu, Haichao Zhang, Di Wu, Wei Xu, Benoit Boulet

AI总结:

本文研究利用预训练视觉-语言模型作为在线强化学习中的奖励,针对稀疏奖励任务中的奖励不对齐问题,提出轻量级微调方法FuRL,通过微调VLM表征与接力RL提升基线智能体性能,并在Meta-world基准上验证了有效性。

摘要:

在本研究中,我们探讨如何将预训练视觉-语言模型(VLM)用于在线强化学习(RL)。特别地,我们聚焦于带有预定义文本任务描述的稀疏奖励任务。我们首先发现在RL任务中将VLM作为奖励应用时存在的奖励不对齐问题。为解决此问题,我们基于奖励对齐和接力RL,引入了一种名为模糊VLM奖励辅助RL(FuRL)的轻量级微调方法。具体而言,我们通过微调VLM表征并利用接力RL来避免局部最优,从而提升了SAC/DrQ基线智能体在稀疏奖励任务上的性能。在Meta-world基准任务上的大量实验证明了所提方法的有效性。代码获取地址:https://github.com/fuyw/FuRL。

英文摘要:

In this work, we investigate how to leverage pre-trained visual-language models (VLM) for online Reinforcement Learning (RL). In particular, we focus on sparse reward tasks with pre-defined textual task descriptions. We first identify the problem of reward misalignment when applying VLM as a reward in RL tasks. To address this issue, we introduce a lightweight fine-tuning method, named Fuzzy VLM reward-aided RL (FuRL), based on reward alignment and relay RL. Specifically, we enhance the performance of SAC/DrQ baseline agents on sparse reward tasks by fine-tuning VLM representations and using relay RL to avoid local minima. Extensive experiments on the Meta-world benchmark tasks demonstrate the efficacy of the proposed method. Code is available at: https://github.com/fuyw/FuRL.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑