跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.20075 · v1

Prompt Re-Specification and Target-Mask Availability in Mask-Conditioned Visual Reinforcement Learning for Moving-Platform Landing

首次提交 2026-11-15 23:32
版本更新 2026-11-15 23:32
雷达观测 2026-11-17 12:06
arXiv ↗PDF ↗

摘要

Visual reinforcement learning (RL) policies that land a drone on a moving platform can fail when the landing pad looks different from the one used in training. Using binary target masks as the visual input removes most pixel-level appearance information while preserving the spatial representation of the target, and lets a text prompt specify the target. In simulation, we test whether a frozen (fixed-weight) mask-based DrQ-v2 policy can handle a new pad when only the prompt of its fine-tuned SAM 3 detector changes. A banana-themed pad lowered landing success from 75.6% to 12.0%, and SAM 3 rarely detected it under the original prompt. A revised prompt, chosen after testing several candidates, restored success to 68.4% with no weight updates. In a separate experiment with a policy trained on ground-truth masks, filling in masks that were missing while the pad was visible helped landing far more than the tested mask-centroid and area corrections. Tracking the pad over time supplied masks more often and also kept producing them after the pad was confirmed absent from the rendered image, but it did not establish a reliable improvement in landing success. Whether a usable mask reaches the policy at all thus appears to be an important limiting factor for this approach.

作者备注:11 pages, 2 figures, 12 tables

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-15 23:32 · 最新观测版本

Prompt Re-Specification and Target-Mask Availability in Mask-Conditioned Visual Reinforcement Learning for Moving-Platform Landing

Visual reinforcement learning (RL) policies that land a drone on a moving platform can fail when the landing pad looks different from the one used in training. Using binary target masks as the visual input removes most pixel-level appearance information while preserving the spatial representation of the target, and lets a text prompt specify the target. In simulation, we test whether a frozen (fixed-weight) mask-based DrQ-v2 policy can handle a new pad when only the prompt of its fine-tuned SAM 3 detector changes. A banana-themed pad lowered landing success from 75.6% to 12.0%, and SAM 3 rarely detected it under the original prompt. A revised prompt, chosen after testing several candidates, restored success to 68.4% with no weight updates. In a separate experiment with a policy trained on ground-truth masks, filling in masks that were missing while the pad was visible helped landing far more than the tested mask-centroid and area corrections. Tracking the pad over time supplied masks more often and also kept producing them after the pad was confirmed absent from the rendered image, but it did not establish a reliable improvement in landing success. Whether a usable mask reaches the policy at all thus appears to be an important limiting factor for this approach.

查看此版本解读 →查看此版本原文 ↗