# CPNav框架：保形预测校准VLM决策提升无图导航成功率

> 原文链接：https://robotics.hellobike.com/papers/2026-11-10-2611-10366-3a593cc0
> 频道：papers | 更新日期：2026-11-10 | 标签：auto, arxiv
> 核心摘要：针对VLM导航决策过置信问题，CPNav结合碰撞感知规划与保形预测校准动作评分。实测SR提升24.9%，SPL提升15.2%，提供有限样本风险保证，代码已开源。

# CPNav框架：保形预测校准VLM决策提升无图导航成功率

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[CPNav: Calibrated Vision-Language Navigation for Object-Goal Search in Unseen Environments](https://arxiv.org/abs/2611.10366)（arXiv 2611.10366）

## 核心内容

UCLA研究团队提出CPNav无图物体目标导航框架，解决视觉语言模型（VLM）在未知环境中动作决策未校准及过度自信的核心痛点。该方案通过离线保形预测（Conformal Prediction）对VLM输出进行统计校准，并在在线推理中结合树状探索回溯机制。在Hiwonder MentorPi机器人真机部署中，相比未校准基线，CPNav将任务成功率（SR）提升24.9%，路径长度加权成功率（SPL）提升15.2%；在AI Habitat仿真环境HM3D与MP3D数据集上，SR分别相对提升6.2%和12.4%。

技术架构上，CPNav从RGB-D输入构建可通行性地图并生成无碰撞运动原语，利用双VLM分别执行目标检测与动作打分。其核心创新在于模型无关的校准层：利用带回溯的树状探索获取Oracle动作作为真值，通过保形预测离线计算校准参数，为用户提供指定风险水平下的有限样本性能保证。在线阶段，校准后的VLM分数指导动作选择，同时记录替代路径以支持动态回退。该框架不依赖特定VLM架构，项目代码已在GitHub开源。

## 关键看点

- **真机性能验证**：在Hiwonder MentorPi机器人上实测，SR与SPL较未校准基线分别提升24.9%和15.2%
- **统计安全保证**：采用保形预测离线校准VLM动作评分，提供用户指定风险水平下的有限样本理论保证
- **鲁棒探索机制**：集成树状结构化探索历史与回溯策略，在线记录替代动作以应对VLM决策失效场景
- **通用即插即用**：校准层与具体VLM解耦，可包裹任意视觉语言模型策略，无需重新训练底层模型

## 工程与产业观察

该工作为具身智能VLM落地提供了关键的“安全阀”工程范式。通过保形预测将大模型语义推理转化为具有统计保证的控制信号，有效缓解了端到端模型在物理交互中的幻觉与过置信风险。对于缺乏高精度先验地图的场景，这种模型无关的校准后处理机制比单纯追求模型精度更具工程性价比，是连接开放世界大模型与封闭安全约束的重要技术桥梁。

## 英文摘要 (Abstract)

Directing a robot to locate a target object in unseen environments remains a central challenge in embodied AI. Vision-Language Model (VLM)-based methods provide rich semantic reasoning but often produce uncalibrated and overconfident action decisions. We introduce CPNav, a navigation framework that combines collision-aware planning with statistically calibrated VLM action scoring for map-free object-goal navigation. From RGB-D input, CPNav constructs a navigability map, generates collision-free motion primitives, and queries two VLMs for goal detection and action scoring. The resulting scores are calibrated offline via conformal prediction (CP) using oracle actions obtained through a tree-structured exploration procedure with backtracking, providing finite-sample guarantees at a user-specified risk level. The CP layer is model-agnostic and can wrap any VLM policy. Online, calibrated VLM scores guide action selection, while a tree-structured exploration history records alternative actions and allows backtracking. We evaluate CPNav in both simulation and real-world hardware settings. In AI Habitat simulation, compared with the uncalibrated baseline method, CPNav achieves relative SR and SPL improvements of 6.2% and 11.4% on HM3D v0.2, and 12.4% and 8.8% on MP3D, respectively. In real-world deployment on a modified Hiwonder MentorPi robot, the corresponding improvements in SR and SPL are 24.9% and 15.2%, respectively. Code is available at https://github.com/practice-lab-ucla/CPnav.

作者：Qizhao Chen, Yuanhong Zeng, Shoh Nishino, Anushri Dixit
:::

## 原始参考来源 (Sources)

- [arXiv 2611.10366：CPNav: Calibrated Vision-Language Navigation for Object-Goal Search in Unseen Environments](https://arxiv.org/abs/2611.10366)