Repository navigation
新增 G1 RGB 视觉巡检示例,提供感知控制、端到端策略与同步建图导航 - #1085
Closed
usernamekkyy wants to merge 15 commits into
Closed
usernamekkyy wants to merge 15 commits into
usernamekkyy wants to merge 15 commits into
Conversation
The learned semantic controller can still command forward motion when segmentation returns no floor pixels. Stop on this unambiguous failure, add a regression test, and state that the guard is a deployment rule rather than a network capability. Also record the seed 9-12 closed-loop results, the catkin source/install build findings, the MaskController curve limitation, and the development test dependency.
Contributor
Author
kangxin-li
requested changes
Nov 10, 2026
kangxin-li
left a comment
Contributor
There was a problem hiding this comment.
1.页面存在隐藏字符警告与部署状态异常:PR 页面提示 “This branch has not been deployed”,同时出现了 “hidden or bidirectional Unicode text” 警告。需要排查是否无意中引入了不可见字符,并确认分支部署状态是否正常。
2.安全保护是“规则”而非网络能力,测试覆盖不足:PR 承认 “分割无地面像素即停车” 是部署端的规则保护,且 MaskController 没有保存逐轮训练曲线。这意味着该控制器的实际鲁棒性缺乏系统性验证,建议要求补充至少一个回归测试,证明该保护规则在无地面像素时确实生效。
3.端到端近障行为未单独测试:PR 明确说 “端到端近障行为未单独测试”。对于视觉巡检示例来说,近障场景是核心风险点,缺少这一项的验证会降低模块的可信度。
kangxin-li
approved these changes
Nov 10, 2026
Contributor
Author
|
感谢审核!已保留并运行无地面像素的停止保护回归测试(本地 6 项通过),补充端到端策略在三种初始近墙距离下的 MuJoCo 测试及结果,并合入上游处理冲突。三个短时测试未记录端墙接触,测试范围与限制已写入文档。扫描本次文本变更及 PR 正文未发现 Unicode 格式控制字符;分支页面部署状态仍需 GitHub 更新后确认。 |
This was referenced Nov 11, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
修改目的与内容
为课程提供区别于无人机和雷达 MLP 避障的 RGB 人形机器人示例。基于官方 Unitree G1 预训练步态,新增
humanoid_visual_patrol模块,覆盖四项任务:底层步态、G1 模型和 ROS 软件归属上游;本项目独立训练 FloorNet、MaskController、ImageController 和 CostNet,并提供训练/评估代码、4 个部署权重、MkDocs 文档和运行 GIF。
已知限制:测试场景仍为同一走廊模板;FloorNet 测试 IoU 0.714;MaskController 未保存逐轮曲线;无地面保护是规则不是网络能力;端到端近障行为未单独测试;安装空间仅验证构建,运行仍按源码空间。
测试环境与结果
catkin_make与catkin_make install均通过,安装空间不含models/与patrol包。pytest tests -q(带HUMANOID_ASSETS):6 passed。SUCCEEDED,21 次地图更新;神经辅助导航 5/5 航点SUCCEEDED,expanded 47 vs 47、path cost 46 vs 46。运行效果
docs/img/humanoid_visual_patrol/perception_7.gifdocs/img/humanoid_visual_patrol/end2end_7.gifdocs/img/humanoid_visual_patrol/slam_map.png(PR 页面补充分支内图片链接。)
提交前检查
大模型使用声明
使用 OpenAI Codex 辅助方案、编程、调试、实验和文档整理;提交者已复核内容并对提交负责。官方 G1 模型、预训练步态和 ROS 组件归属上游,详见模块
third_party_notices.md。拆分后的独立 PR
本 PR 已拆为以下五项。请在对应 PR 中审核。