# PEARS：面向触觉操作的物理先验引导高效适应，通过失败推理和 Diffusion Steering 实现

> 原文链接：https://robotics.hellobike.com/papers/2026-10-06-2610-08784
> 频道：papers | 更新日期：2026-10-06 | 标签：arxiv, cs.RO
> 核心摘要：PEARS面向预训练策略在OOD触觉操作中的退化，结合物理先验失败诊断、力位控制与扩散引导RL，并在仿真和真实任务验证。

# PEARS：面向触觉操作的物理先验引导高效适应，通过失败推理和 Diffusion Steering 实现

::: v-pre
> 原文链接：[PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation](https://arxiv.org/abs/2610.08784)（arXiv 2610.08784 · 发布日期：2026-10-06）

**PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation**

PEARS面向预训练策略在OOD触觉操作中的退化，结合物理先验失败诊断、力位控制与扩散引导RL，并在仿真和真实任务验证。

## 核心要点

- 预训练机器人策略部署遇到OOD条件时性能可能显著下降，而RL后训练交互成本高。
- PFR模块用VLM中的物理先验，从视觉结果和触觉历史诊断失败并更新接触力边界。
- 高频混合力位控制器执行力边界，触觉条件扩散引导RL调整冻结策略的潜在噪声。
- 仿真成功率较最强基线提升12.4至37.4个百分点，达阈值交互次数最多降53.2%；项目网站已公开。

**验证范围与局限**：验证范围包括仿真，以及真实世界Whiteboard Erasing和Pipette Liquid Aspiration任务。

## 英文摘要 (Abstract)

Pretrained robotic policies can suffer substantial performance degradation under out-of-distribution (OOD) conditions encountered during deployment, motivating post-training through real-world interaction. However, reinforcement-learning (RL)-based post-training typically requires substantial environment interactions, a burden that is especially significant in manipulation, where each trial can be slow, costly, or destructive. Therefore, we present PEARS, a physics-prior-guided hybrid RL framework for sample-efficient online adaptation of pretrained policies with tactile feedback. After each episode, its physics-guided force reasoning (PFR) module uses physical priors encoded in a vision-language model (VLM) to diagnose failures from the visual outcome and tactile interaction history and update task-appropriate contact-force bounds. A high-frequency hybrid force-position controller then enforces these bounds during contact. Complementarily, tactile-conditioned diffusion steering reinforcement learning adjusts the latent noise of the frozen flow-matching policy to correct errors in free-space motion and contact timing without updating the base model. In simulation, PEARS improves success rates by 12.4-37.4 percentage points over the strongest per-task baselines. PEARS also reduces the number of interaction episodes required for a certain success threshold by up to 53.2% relative to the fastest baseline. In real-world experiments, PEARS achieves success rates of 95% on Whiteboard Erasing and 90% on Pipette Liquid Aspiration. These results show that combining the PFR module with policy steering can accelerate adaptation while reducing costly interactions. The project website is available at https://song-kun.github.io/pears.

作者：Kun Song, Yiming Wang, Yilin Chen, Tianyi Ding, Jiaxin Tian, Tianqi Gong, Daolin Ma, Jia Pan
:::

## 原始参考来源 (Sources)

- [arXiv 2610.08784：PEARS: Physical-Prior-Guided Efficient Adaptation via Failure Reasoning and Diffusion Steering for Tactile Manipulation](https://arxiv.org/abs/2610.08784)