arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.00699cs.LGcs.AI
收藏

基于模型的离线强化学习与下期望分位 Q-Learning

Model-based Offline Reinforcement Learning with Lower Expectile Q-Learning

Yonsei University(延世大学)

Kwanyoung Park, Youngwoon Lee

AI总结:

本文提出基于模型的离线强化学习方法 LEQ,通过对 λ-returns 做下期望分位回归获得低偏差价值估计,并在 D4RL、NeoRL 和 V-D4RL 等基准上展现出优于或匹配现有方法的鲁棒性能。

摘要:

基于模型的离线强化学习(RL)是一种具有吸引力的方法,它通过使用学习到的模型生成想象轨迹,来解决从有限静态数据中学习的挑战。然而,这类方法往往难以处理模型 rollout 导致的价值估计不准确问题。本文提出一种新颖的基于模型的离线 RL 方法——Lower Expectile Q-learning(LEQ,下期望分位 Q-learning),通过对 λ-returns 进行下期望分位回归,提供低偏差的基于模型的价值估计。实验结果表明,LEQ 在长视野任务(如 D4RL AntMaze 任务)上显著优于以往基于模型的离线 RL 方法,并达到或超过无模型方法与序列建模方法的性能。此外,在密集奖励环境中,无论是基于状态的任务(NeoRL 和 D4RL)还是基于像素的任务(V-D4RL),LEQ 都能匹配最先进的基于模型和无模型方法的性能,表明 LEQ 在多种领域中均具有鲁棒性。消融研究表明,下期望分位回归、λ-returns 以及在离线数据上训练 critic 对 LEQ 都至关重要。

英文摘要:

Model-based offline reinforcement learning (RL) is a compelling approach that addresses the challenge of learning from limited, static data by generating imaginary trajectories using learned models. However, these approaches often struggle with inaccurate value estimation from model rollouts. In this paper, we introduce a novel model-based offline RL method, Lower Expectile Q-learning (LEQ), which provides a low-bias model-based value estimation via lower expectile regression of $λ$-returns. Our empirical results show that LEQ significantly outperforms previous model-based offline RL methods on long-horizon tasks, such as the D4RL AntMaze tasks, matching or surpassing the performance of model-free approaches and sequence modeling approaches. Furthermore, LEQ matches the performance of state-of-the-art model-based and model-free methods in dense-reward environments across both state-based tasks (NeoRL and D4RL) and pixel-based tasks (V-D4RL), showing that LEQ works robustly across diverse domains. Our ablation studies demonstrate that lower expectile regression, $λ$-returns, and critic training on offline data are all crucial for LEQ.

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。

↑