# ContactDP：用于紧配合插入任务的接触引导 Diffusion Policy

> 原文链接：https://robotics.hellobike.com/papers/2026-09-20-2609-23800
> 频道：papers | 更新日期：2026-09-20 | 标签：arxiv, cs.RO
> 核心摘要：针对高精度连接器插入中的接触不确定性，ContactDP融合视觉、触觉与力矩信息，并在工业级连接器任务上验证。

# ContactDP：用于紧配合插入任务的接触引导 Diffusion Policy

::: v-pre
> 原文链接：[ContactDP: Contact-Guided Diffusion Policy for Tight Insertion Tasks](https://arxiv.org/abs/2609.23800)（arXiv 2609.23800 · 发布日期：2026-09-20）

**ContactDP: Contact-Guided Diffusion Policy for Tight Insertion Tasks**

针对高精度连接器插入中的接触不确定性，ContactDP融合视觉、触觉与力矩信息，并在工业级连接器任务上验证。

## 核心要点

- 高精度连接器插入受紧公差、接触中部分可观测和遮挡歧义带来的多模态不确定性影响。
- ContactDP联合腕部RGB、指尖触觉和腕部力-力矩测量，推断接触状态并生成校正运动。
- 学习策略与混合位置-力控制器配合，在接触下提供顺应的低层交互与稳定执行。
- 在多种工业级连接器插入任务上，较仅视觉扩散策略在性能、可靠性和泛化上更优。

**验证范围与局限**：验证范围为一组工业级连接器插入任务，涵盖不同连接器几何、抓取条件和初始未对准。

## 英文摘要 (Abstract)

High-precision connector insertion remains challenging for robotic systems due to tight mechanical tolerances, partial observability during contact, and multimodal uncertainty arising from occlusion and contact ambiguity. Successful insertion requires closed-loop contact guidance that continuously integrates global alignment cues with local contact feedback to produce stable corrective actions under interaction. In this work, we present ContactDP (Contact-Guided Diffusion Policy for Tight Insertion Tasks), a multimodal diffusion-policy framework for contact-rich insertion. ContactDP jointly integrates wrist RGB observations, fingertip tactile sensing, and wrist-mounted force-torque measurements to infer contact state and generate temporally consistent corrective motions during insertion. To ensure stable execution under contact, the learned policy operates together with a hybrid position-force controller that provides compliant low-level interaction. We evaluate our approach on a suite of industrial-grade connector insertion tasks with varying connector geometries, grasp conditions, and initial misalignment. Across all tasks, ContactDP significantly outperforms vision-only diffusion policies for performance, reliability and generalization.

作者：Chengyi Xing, Shaoxiong Yao, Diego Romeres, Devesh K. Jha
:::

## 原始参考来源 (Sources)

- [arXiv 2609.23800：ContactDP: Contact-Guided Diffusion Policy for Tight Insertion Tasks](https://arxiv.org/abs/2609.23800)