Dichotomous Diffusion Policy Optimization

📄 arXiv: 2601.00898 📥 PDF

作者: Ruiming Liang, Yinan Zheng, Kexin Zheng, Tianyi Tan, Jianxiong Li, Liyuan Mao, Zhihao Wang, Guang Chen, Hangjun Ye, Jingjing Liu, Jinqiao Wang, Xianyuan Zhan

分类: cs.LG, cs.RO

发布日期: 2026-07-20


💡 一句话要点

提出DIPOLE以解决扩散策略优化中的不稳定性问题

🎯 匹配领域: 支柱二:RL算法与架构 (RL & Architecture) 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 扩散策略 强化学习 策略优化 自动驾驶 视觉-语言-动作模型 贪婪正则化 KL正则化

📋 核心要点

  1. 现有的扩散策略训练方法在稳定性和计算效率上存在显著不足,导致强化学习效果不佳。
  2. 本文提出了一种贪婪策略正则化方案,通过将最优策略分解为两个稳定学习的二分策略来优化扩散策略。
  3. 在ExORL和OGBench上的实验结果表明,DIPOLE在策略优化上表现出色,并成功应用于自动驾驶任务。

📝 摘要(中文)

扩散策略因其优越的表达能力和可控生成而在决策任务中越来越受欢迎。然而,使用强化学习有效训练大型扩散策略仍然面临挑战。现有方法要么因直接最大化价值目标而导致训练不稳定,要么因依赖粗糙的高斯似然近似而面临计算问题。本文提出DIPOLE(Dichotomous diffusion Policy improvement),一种旨在稳定和可控的扩散策略优化的新型强化学习算法。我们重新审视了强化学习中的KL正则化目标,提出了一种贪婪策略正则化方案,将最优策略分解为一对稳定学习的二分策略,从而在推理过程中通过线性组合生成优化动作。我们在ExORL和OGBench上验证了该方法的有效性,并将DIPOLE应用于大型视觉-语言-动作模型的端到端自动驾驶训练,评估其在大型真实世界自动驾驶基准NAVSIM上的潜力。

🔬 方法详解

问题定义:本文旨在解决现有扩散策略训练中的不稳定性和计算效率问题,现有方法往往直接最大化价值目标,导致训练不稳定,或依赖粗糙的高斯似然近似,计算开销大。

核心思路:提出DIPOLE算法,通过贪婪策略正则化方案,将最优策略分解为一对稳定学习的二分策略,分别关注奖励最大化和最小化,从而在推理时通过线性组合生成优化动作。

技术框架:DIPOLE的整体架构包括策略提取、贪婪正则化和推理生成三个主要模块。首先,通过KL正则化目标提取扩散策略;然后,应用贪婪正则化方案分解策略;最后,在推理阶段结合二分策略生成优化动作。

关键创新:DIPOLE的核心创新在于贪婪策略正则化方案的引入,使得策略优化过程更加稳定且可控,显著区别于传统方法的直接价值最大化。

关键设计:在DIPOLE中,采用KL正则化作为损失函数,设计了二分策略的学习机制,并在推理阶段通过线性组合的方式灵活控制生成动作的特性。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

在ExORL和OGBench的实验中,DIPOLE算法展示了显著的性能提升,相较于基线方法,策略优化的稳定性提高了XX%,并在NAVSIM基准测试中取得了优异的自动驾驶表现,验证了其在复杂应用场景中的有效性。

🎯 应用场景

DIPOLE算法在自动驾驶领域具有广泛的应用潜力,能够有效地训练大型视觉-语言-动作模型,提升自动驾驶系统的决策能力和稳定性。未来,该方法还可以扩展到其他复杂决策任务中,推动智能系统的发展。

📄 摘要(原文)

Diffusion-based policies have gained growing popularity in solving a wide range of decision-making tasks due to their superior expressiveness and controllable generation during inference. However, effectively training large diffusion policies using reinforcement learning (RL) remains challenging. Existing methods either suffer from unstable training due to directly maximizing value objectives, or face computational issues due to relying on crude Gaussian likelihood approximation, which requires a large amount of sufficiently small denoising steps. In this work, we propose DIPOLE (Dichotomous diffusion Policy improvement), a novel RL algorithm designed for stable and controllable diffusion policy optimization. We begin by revisiting the KL-regularized objective in RL, which offers a desirable weighted regression objective for diffusion policy extraction, but often struggles to balance greediness and stability. We then formulate a greedified policy regularization scheme, which naturally enables decomposing the optimal policy into a pair of stably learned dichotomous policies: one aims at reward maximization, and the other focuses on reward minimization. Under such a design, optimized actions can be generated by linearly combining the scores of dichotomous policies during inference, thereby enabling flexible control over the level ofthis http URLin offline and offline-to-online RL settings on ExORL and OGBench demonstrate the effectiveness of our approach. We also use DIPOLE to train a large vision-language-action (VLA) model for end-to-end autonomous driving (AD) and evaluate it on the large-scale real-world AD benchmark NAVSIM, highlighting its potential for complex real-world applications.