# PathLift：用于视觉导航先验的深度条件度量提升

> 原文链接：https://robotics.hellobike.com/papers/2026-10-12-2610-16103
> 频道：papers | 更新日期：2026-10-12 | 标签：arxiv, cs.RO
> 核心摘要：针对视觉导航策略只给方向不定距离，PathLift用传感器级深度逐点生成度量路径，并在仿真和真实轮式机器人上验证。

# PathLift：用于视觉导航先验的深度条件度量提升

::: v-pre
> 原文链接：[PathLift: Depth-Conditioned Metric Lifting for Visual-Navigation Priors](https://arxiv.org/abs/2610.16103)（arXiv 2610.16103 · 发布日期：2026-10-12）

**PathLift: Depth-Conditioned Metric Lifting for Visual-Navigation Priors**

针对视觉导航策略只给方向不定距离，PathLift用传感器级深度逐点生成度量路径，并在仿真和真实轮式机器人上验证。

## 核心要点

- 通用视觉导航策略输出无量纲轨迹，部署时还需转换为控制器可跟踪的度量路径。
- PathLift接入RGB-D、双目或投影LiDAR深度，拟合每个路点的幅值与横向残差。
- 相比统一缩放或策略内单目深度，方法面向每个路点进行深度条件校正。
- 在多种视觉导航策略、仿真及真实轮式机器人上，闭环导航表现持续提升。

**验证范围与局限**：验证范围为仿真和一台真实轮式机器人，真实实验使用双目与LiDAR深度。

## 英文摘要 (Abstract)

Generalist visual-navigation policies tell a robot which direction to go but not how far. A separate lifting step is needed at deployment to convert their unitless trajectory into a metric path that a controller can track; existing scalar lifters apply uniform scaling that misses per-waypoint corrections, while in-policy approaches that inject monocular depth into the policy inherit the prediction errors of the depth model near obstacles and at close range. We present PathLift, a per-waypoint depth-conditioned adapter that lifts a pre-trained visual-navigation policy by consuming an explicit, sensor-grade depth source (RGB-D, stereo, or projected LiDAR) and fitting per-waypoint magnitude and lateral residuals to a canonicalized prior. PathLift consistently improves closed-loop navigation across diverse visual-navigation policies in simulation and on a real wheeled robot under both stereo and LiDAR depth. By turning the policy's unitless intent into a depth-feasible metric path in a single feed-forward pass, the adapter lifts real-world success without test-time optimization.

作者：Gwanjun Shin, Hyeongseok Suh, Hongmo Jung, Jongwoo Lim
:::

## 原始参考来源 (Sources)

- [arXiv 2610.16103：PathLift: Depth-Conditioned Metric Lifting for Visual-Navigation Priors](https://arxiv.org/abs/2610.16103)