arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2511.08922cs.LGcs.AIcs.RO
收藏

基于价值条件优化的扩散策略用于离线强化学习

Diffusion Policies with Value-Conditional Optimization for Offline Reinforcement Learning

College of Intelligence Science and Technology, National University of Defense Technology(智能科学与技术学院,国防科技大学)

Yunchang Ma, Tenglong Liu, Yixing Lan, Xin Yin, Changxin Zhang, Xinglong Zhang, Xin Xu

AI总结:

针对离线强化学习中扩散策略因无差别正则化导致的过度保守问题,提出DIVO方法,通过价值条件优化引导扩散模型生成高质量分布内样本,在D4RL基准上提升了locomotion任务平均回报并在AntMaze领域表现更优。

摘要:

在离线强化学习中,由分布外(OOD)动作导致的价值高估严重限制了策略性能。近年来,扩散模型因其强大的分布匹配能力被用于通过行为策略约束来增强保守性。然而,现有方法常对低质量数据集中的冗余动作进行无差别正则化,导致过度保守以及扩散建模的表达性与效率之间的失衡。为解决这些问题,我们提出了价值条件优化扩散策略(DIVO),这是一种利用扩散模型生成高质量、覆盖广泛的分布内状态-动作样本同时促进高效策略改进的新方法。具体而言,DIVO引入了二元加权机制,利用离线数据集中动作的优势值来引导扩散模型训练。这使得模型能更精确地与数据集分布对齐,同时选择性地扩展高优势动作的边界。在策略改进阶段,DIVO从扩散模型中动态筛选具有高回报潜力的动作,有效引导学习到的策略获得更好性能。该方法在离线强化学习中实现了保守性与探索性之间的关键平衡。我们在D4RL基准上评估了DIVO并与最先进的基线方法进行比较。实验结果表明,DIVO性能更优,在 locomotion 任务中的平均回报显著提升,且在稀疏奖励构成主要挑战的复杂AntMaze领域优于现有方法。

英文摘要:

In offline reinforcement learning, value overestimation caused by out-of-distribution (OOD) actions significantly limits policy performance. Recently, diffusion models have been leveraged for their strong distribution-matching capabilities, enforcing conservatism through behavior policy constraints. However, existing methods often apply indiscriminate regularization to redundant actions in low-quality datasets, resulting in excessive conservatism and an imbalance between the expressiveness and efficiency of diffusion modeling. To address these issues, we propose DIffusion policies with Value-conditional Optimization (DIVO), a novel approach that leverages diffusion models to generate high-quality, broadly covered in-distribution state-action samples while facilitating efficient policy improvement. Specifically, DIVO introduces a binary-weighted mechanism that utilizes the advantage values of actions in the offline dataset to guide diffusion model training. This enables a more precise alignment with the dataset's distribution while selectively expanding the boundaries of high-advantage actions. During policy improvement, DIVO dynamically filters high-return-potential actions from the diffusion model, effectively guiding the learned policy toward better performance. This approach achieves a critical balance between conservatism and explorability in offline RL. We evaluate DIVO on the D4RL benchmark and compare it against state-of-the-art baselines. Empirical results demonstrate that DIVO achieves superior performance, delivering significant improvements in average returns across locomotion tasks and outperforming existing methods in the challenging AntMaze domain, where sparse rewards pose a major difficulty.

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。

↑