# SADP：从基础模型生成示教中学习的面向长时程操作的子目标感知 Diffusion Policy

> 原文链接：https://robotics.hellobike.com/papers/2026-10-09-2605-16871
> 频道：papers | 更新日期：2026-10-09 | 标签：arxiv, cs.RO
> 核心摘要：SADP利用基础模型生成子目标标注示教，训练扩散策略，并在RLBench仿真和UR5e真机评估。

# SADP：从基础模型生成示教中学习的面向长时程操作的子目标感知 Diffusion Policy

::: v-pre
> 原文链接：[SADP: Subgoal-Aware Diffusion Policy for Long-Horizon Manipulation Learned From Foundation Model Generated Demonstrations](https://arxiv.org/abs/2605.16871)（arXiv 2605.16871 · 发布日期：2026-10-09）

**SADP: Subgoal-Aware Diffusion Policy for Long-Horizon Manipulation Learned From Foundation Model Generated Demonstrations**

SADP利用基础模型生成子目标标注示教，训练扩散策略，并在RLBench仿真和UR5e真机评估。

## 核心要点

- 长时程操作需协调多个中间子目标并判断切换时机，常规模仿学习缺少显式建模。
- SADP自动生成带子目标标注的示教，以任务级和子目标级自然语言描述条件化动作生成。
- 轻量辅助头预测继续分数，用于在线子目标切换，并支持阶段级进度监控。
- 实验显示其保持有竞争力的任务表现，同时暴露时间对齐的子目标执行信号。

**验证范围与局限**：验证范围包括RLBench仿真和UR5e机器人真实世界评估。

## 英文摘要 (Abstract)

Long-horizon robot manipulation requires policies to coordinate multiple intermediate subgoals and determine when to advance between them. However, most imitation learning methods are trained solely on task-level demonstrations, without explicitly modeling the active subgoal or its execution progress. This limitation is further exacerbated by the scarcity of subgoal-level supervision in standard robot learning datasets, which makes explicit subgoal-conditioned control and online transition modeling difficult to learn. To address this issue, this paper proposes Subgoal-Aware Diffusion Policy (SADP), a framework that leverages foundation models to autonomously generate subgoal-annotated demonstrations and trains diffusion policies on these datasets. SADP structures policy execution around explicit natural-language subgoals by conditioning action generation on both task-level and subgoal-level descriptions. A lightweight auxiliary head further predicts a continuation score that drives online subgoal switching and supports stage-level progress monitoring. Experiments in RLBench simulations and real-world evaluations on a UR5e robot demonstrate that SADP maintains competitive task performance while exposing temporally aligned subgoal-level execution signals for progress monitoring. These results show that explicit subgoal progression can be incorporated into a diffusion policy without degrading task-level performance.

作者：Site Hu, Takato Horii
:::

## 原始参考来源 (Sources)

- [arXiv 2605.16871：SADP: Subgoal-Aware Diffusion Policy for Long-Horizon Manipulation Learned From Foundation Model Generated Demonstrations](https://arxiv.org/abs/2605.16871)