[{"data":1,"prerenderedAt":345},["ShallowReactive",2],{"project-detail-\u002Fprojects\u002Far-ur5-real":3,"project-detail-nav":94},{"id":4,"title":5,"body":6,"category":56,"cover":57,"description":58,"extension":59,"highlights":60,"images":65,"links":71,"meta":73,"navigation":75,"path":76,"period":77,"portrait":78,"poster":79,"role":80,"seo":81,"stack":82,"status":87,"stem":88,"tags":89,"video":92,"videos":79,"__hash__":93},"projects\u002Fprojects\u002Far-ur5-real.md","真机拓展：Quest 3 AR 透视控制真实 UR5",{"type":7,"value":8,"toc":49},"minimark",[9,13,17,20,43,46],[10,11,12],"h2",{"id":12},"项目简介",[14,15,16],"p",{},"在「Quest 3 VR 远程操控 UR5 数字孪生」纯仿真项目的基础上，独立开发真机控制链路 UR5_Teleop：Quest 3 以 AR 彩色透视模式运行，将头显与手柄位姿映射到真实 UR5 机械臂，在实际工作台上完成遥控运动。该拓展与仿真仓库代码相互独立。",[10,18,19],{"id":19},"技术方案",[21,22,23,31,37],"ul",{},[24,25,26,30],"li",{},[27,28,29],"strong",{},"真机通信","：头显控制位姿经局域网发送至真机控制服务，转换为 UR5 可执行的末端目标。",[24,32,33,36],{},[27,34,35],{},"AR 透视","：利用 Quest 3 彩色透视直接观察真实工作台与机械臂，无需重建场景模型。",[24,38,39,42],{},[27,40,41],{},"控制迁移","：复用仿真中验证的笛卡尔伺服与夹爪连续控制逻辑，缩短真机调试周期。",[10,44,45],{"id":45},"个人收获",[14,47,48],{},"完整走通了控制策略从仿真到真机的迁移过程，体会到 AR 透视在遥操作中降低空间认知负担的作用，也积累了真机安全控制与通信时延处理的经验。",{"title":50,"searchDepth":51,"depth":51,"links":52},"",2,[53,54,55],{"id":12,"depth":51,"text":12},{"id":19,"depth":51,"text":19},{"id":45,"depth":51,"text":45},"具身智能","\u002Fimages\u002Fprojects\u002Freal-hardware-overview.jpg","在纯仿真项目基础上独立拓展真机控制链路 UR5_Teleop，Quest 3 AR 彩色透视下将头显与手柄位姿映射到真实 UR5 完成遥控运动。","md",[61,62,63,64],"在纯仿真项目基础上独立拓展真机控制链路（UR5_Teleop），将 Quest 3 头显与手柄位姿映射到真实 UR5。","Quest 3 AR 彩色透视下保留真实工作台视野，在实际工作台上完成机械臂遥控运动。","真机端复用笛卡尔伺服与连续夹爪控制逻辑，仿真中验证的控制链路可直接向真机迁移。","拓展代码 UR5_Teleop 与仿真仓库相互独立，已开源。",[66,68],{"src":57,"title":67},"真机 UR5、RobotiQ 夹爪与 D405 腕部相机",{"src":69,"title":70},"\u002Fimages\u002Fprojects\u002Freal-d405-mount.jpg","D405 相机法兰安装",{"github":72},"https:\u002F\u002Fgithub.com\u002Fwkcherish\u002FUR5_Teleop.git",{"date":74},"2026-09-01",true,"\u002Fprojects\u002Far-ur5-real","2026.08 – 至今",false,null,"独立开发",{"title":5,"description":58},[83,84,85,86],"Quest 3","Python","UR5","Three.js","进行中","projects\u002Far-ur5-real",[83,35,85,90,91],"UR5_Teleop","遥操作","\u002Fvideos\u002Fprojects\u002Far-ur5.mp4","Piwg_HpfiihFgsBjLUZdewuitikfE9K5-oasSLhNyGM",[95,136,226,308],{"id":4,"title":5,"body":96,"category":56,"cover":57,"description":58,"extension":59,"highlights":127,"images":128,"links":131,"meta":132,"navigation":75,"path":76,"period":77,"portrait":78,"poster":79,"role":80,"seo":133,"stack":134,"status":87,"stem":88,"tags":135,"video":92,"videos":79,"__hash__":93},{"type":7,"value":97,"toc":122},[98,100,102,104,118,120],[10,99,12],{"id":12},[14,101,16],{},[10,103,19],{"id":19},[21,105,106,110,114],{},[24,107,108,30],{},[27,109,29],{},[24,111,112,36],{},[27,113,35],{},[24,115,116,42],{},[27,117,41],{},[10,119,45],{"id":45},[14,121,48],{},{"title":50,"searchDepth":51,"depth":51,"links":123},[124,125,126],{"id":12,"depth":51,"text":12},{"id":19,"depth":51,"text":19},{"id":45,"depth":51,"text":45},[61,62,63,64],[129,130],{"src":57,"title":67},{"src":69,"title":70},{"github":72},{"date":74},{"title":5,"description":58},[83,84,85,86],[83,35,85,90,91],{"id":137,"title":138,"body":139,"category":56,"cover":190,"description":191,"extension":59,"highlights":192,"images":197,"links":212,"meta":213,"navigation":75,"path":215,"period":216,"portrait":78,"poster":79,"role":80,"seo":217,"stack":218,"status":221,"stem":222,"tags":223,"video":50,"videos":79,"__hash__":225},"projects\u002Fprojects\u002Far-ur5-rl.md","Quest 3 VR 远程操控 UR5 数字孪生（纯仿真）",{"type":7,"value":140,"toc":185},[141,143,146,148,180,182],[10,142,12],{"id":12},[14,144,145],{},"搭建可用于交互控制和状态反馈的 UR5 机械臂 MuJoCo 数字孪生仿真。前端基于 Three.js 构建与后端状态相互映射的 3D 可视化界面，后端使用 FastAPI 接收控制输入并驱动仿真；Quest 3 通过局域网 WebXR 进入 VR 后，可用手柄或手势连续控制机械臂与夹爪。本项目为纯仿真工作，真机 AR 拓展为独立项目。",[10,147,19],{"id":19},[21,149,150,156,162,168,174],{},[24,151,152,155],{},[27,153,154],{},"仿真环境","：MuJoCo 中建立 UR5 机械臂、工作台、工具箱与工具模型，配置接触、摩擦与执行器参数。",[24,157,158,161],{},[27,159,160],{},"可视化","：Three.js 实时渲染机械臂状态，与后端通过 WebSocket\u002FHTTP 保持状态同步。",[24,163,164,167],{},[27,165,166],{},"通信层","：FastAPI 提供状态观测、相机画面与控制接口，桥接前端输入与仿真动作。",[24,169,170,173],{},[27,171,172],{},"VR 交互","：Quest 3 利用手柄 6DoF 位姿或 Hand Tracking 手势，经笛卡尔伺服映射为末端执行器目标位姿。",[24,175,176,179],{},[27,177,178],{},"伺服控制","：后端通过 MuJoCo Jacobian 将末端误差换算为关节速度，并串联限位、奇异点降速与碰撞保护。",[10,181,45],{"id":45},[14,183,184],{},"系统理解了数字孪生系统从建模、通信到 VR 交互的完整链路，掌握了 WebXR 开发与基于 Jacobian 的笛卡尔伺服方法，也为后续真机遥操作与模仿学习复现打下基础。",{"title":50,"searchDepth":51,"depth":51,"links":186},[187,188,189],{"id":12,"depth":51,"text":12},{"id":19,"depth":51,"text":19},{"id":45,"depth":51,"text":45},"\u002Fimages\u002Fprojects\u002Fsim-overview.png","基于 MuJoCo 的 UR5 数字孪生仿真，Three.js 可视化与 FastAPI 后端，Quest 3 通过 WebXR 在 VR 中以手柄或手势控制机械臂。",[193,194,195,196],"在 MuJoCo 中搭建 UR5 机械臂、工作台与工具箱、钳子、扳手等工具的高保真仿真场景，FastAPI 后端提供状态观测与控制接口。","Three.js 构建与后端状态实时双向映射的 3D 数字孪生界面，桌面浏览器即可完成调试与观测。","Quest 3 经局域网以 WebXR 进入 VR，手柄或手势通过 6DoF 笛卡尔伺服控制机械臂，Trigger 连续控制夹爪开合。","后端基于 Jacobian 实现 Cartesian servo，并施加速度、加速度、关节限位与奇异点约束，保证仿真中运动平滑安全。",[198,200,203,206,209],{"src":190,"title":199},"MuJoCo 仿真场景：UR5、工作台与工具箱",{"src":201,"title":202},"\u002Fimages\u002Fprojects\u002Fsim-overview-2.png","仿真场景另一视角",{"src":204,"title":205},"\u002Fimages\u002Fprojects\u002Fsim-overview-3.png","工作台后侧视角",{"src":207,"title":208},"\u002Fimages\u002Fprojects\u002Fsim-d405-view.png","D405 腕部相机渲染视角",{"src":210,"title":211},"\u002Fimages\u002Fprojects\u002Fsim-cad-mount.png","相机法兰安装件设计",{},{"date":214},"2026-06-15","\u002Fprojects\u002Far-ur5-rl","2026.03 – 2026.06",{"title":138,"description":191},[219,86,220,83,84],"MuJoCo","FastAPI","已完成","projects\u002Far-ur5-rl",[219,86,220,83,224],"WebXR","kXCTEg8RB2d0GP8OHN0IfhCQg71WeWqr28NSnqb6jHU",{"id":227,"title":228,"body":229,"category":270,"cover":271,"description":272,"extension":59,"highlights":273,"images":79,"links":278,"meta":280,"navigation":75,"path":282,"period":283,"portrait":78,"poster":271,"role":284,"seo":285,"stack":286,"status":221,"stem":290,"tags":291,"video":295,"videos":296,"__hash__":307},"projects\u002Fprojects\u002Flerobot-vla-reproduction.md","LeRobot101：ACT \u002F smolVLA \u002F pi0.5 复现",{"type":7,"value":230,"toc":265},[231,233,236,239,259,262],[10,232,12],{"id":12},[14,234,235],{},"以 Hugging Face LeRobot 框架为主线，系统复现 ACT、smolVLA、pi0.5 三个代表性模型，覆盖模仿学习与视觉-语言-动作两大技术路线。项目打通了从遥操作数据采集、策略训练到仿真评估的完整链路，所有复现代码已整理上传至 GitHub。",[10,237,238],{"id":238},"复现内容",[21,240,241,247,253],{},[24,242,243,246],{},[27,244,245],{},"ACT（Action Chunking with Transformers）","：以动作分块预测替代单步动作回归，配合时间集成（temporal ensembling）平滑执行轨迹，复现数据采集、训练与仿真评估流程。",[24,248,249,252],{},[27,250,251],{},"smolVLA","：基于 SmolVLM 视觉-语言骨干的轻量 VLA 模型，在桌面操作任务上完成端到端复现，理解视觉观测与语言指令到动作 token 的映射方式。",[24,254,255,258],{},[27,256,257],{},"pi0.5","：基于 OpenPI 的预训练 VLA 策略，在自有任务配置上进行微调，验证「语言指令 → 视觉-语言理解 → 动作序列输出」的完整执行链路。",[10,260,261],{"id":261},"评估结果",[14,263,264],{},"在 Robosuite 中以脚本专家生成演示数据，训练行为克隆（BC）策略；在连续 20 轮评估中达到 80% 任务成功率，验证了数据质量与动作分块对策略稳定性的作用。",{"title":50,"searchDepth":51,"depth":51,"links":266},[267,268,269],{"id":12,"depth":51,"text":12},{"id":238,"depth":51,"text":238},{"id":261,"depth":51,"text":261},"机器人学习","\u002Fimages\u002Fprojects\u002Fact-poster.jpg","基于 LeRobot 框架复现 ACT、smolVLA 与 pi0.5 三类模仿学习与视觉-语言-动作模型，打通数据采集、策略训练到仿真评估全流程。",[274,275,276,277],"跟随 LeRobot101 复现 ACT 模型，理解动作分块（Action Chunking）与模仿学习训练流程。","复现 smolVLA 轻量 VLA 模型，在仿真桌面任务上完成数据采集、策略训练与评估。","复现 pi0.5：基于 OpenPI 预训练 VLA 策略进行微调，验证语言指令到机械臂动作的执行链路。","配合 Robosuite 行为克隆实验，从脚本专家数据训练策略，20 轮评估达到 80% 成功率。",{"github":279},"https:\u002F\u002Fgithub.com\u002Fwkcherish\u002Flerobot.git",{"date":281},"2026-05-20","\u002Fprojects\u002Flerobot-vla-reproduction","2025.11 – 2026.05","独立复现",{"title":228,"description":272},[287,288,289,219],"LeRobot","PyTorch","Transformers","projects\u002Flerobot-vla-reproduction",[287,292,251,257,293,294],"ACT","模仿学习","VLA","\u002Fvideos\u002Fprojects\u002Fact.mp4",[297,299,303],{"src":295,"title":298,"poster":271},"ACT 策略仿真任务演示",{"src":300,"title":301,"poster":302},"\u002Fvideos\u002Fprojects\u002Fsmolvla.mp4","smolVLA 桌面任务复现","\u002Fimages\u002Fprojects\u002Fsmolvla-poster.jpg",{"src":304,"title":305,"poster":306},"\u002Fvideos\u002Fprojects\u002Fpi0.5.mp4","pi0.5 微调策略执行","\u002Fimages\u002Fprojects\u002Fpi05-poster.jpg","OBkLJ9rOpfpt66cMl1TkSBaef3E8ALjvOwncZv3t9WU",{"id":309,"title":310,"body":311,"category":321,"cover":322,"description":323,"extension":59,"highlights":324,"images":79,"links":329,"meta":330,"navigation":75,"path":332,"period":333,"portrait":78,"poster":79,"role":80,"seo":334,"stack":335,"status":87,"stem":339,"tags":340,"video":50,"videos":79,"__hash__":344},"projects\u002Fprojects\u002Furgrasp-agent.md","UR5 Grasp Agent：从手眼标定到 GraspNet 抓取",{"type":7,"value":312,"toc":318},[313,315],[10,314,12],{"id":12},[14,316,317],{},"构建一套从硬件标定到抓取策略的 UR5 抓取系统，打通 RealSense 感知、GraspNet 位姿检测与机械臂执行链路。",{"title":50,"searchDepth":51,"depth":51,"links":319},[320],{"id":12,"depth":51,"text":12},"机器人感知","\u002Fimages\u002Fprojects\u002Furgrasp-annotation.png","面向真实 UR5 的抓取智能体，覆盖相机标定、手眼标定、GraspNet 抓取位姿检测与机械臂运动执行全链路。",[325,326,327,328],"完成 RealSense 相机内参标定与眼在手上（eye-in-hand）手眼标定，统一米制与旋转矢量表示。","基于 GraspNet 从 RGB-D 场景检测 6D 抓取位姿，融合桌面物体分割与抓取点生成。","通过 ur_rtde 驱动 UR5 执行抓取运动，构建从视觉感知到运动执行的完整抓取 Agent。","统一相机系、标定板系与机器人基坐标系的米制尺度与旋转矢量表示，控制手眼标定重投影与位姿传递误差。",{},{"date":331},"2025-09-01","\u002Fprojects\u002Furgrasp-agent","2025.12 – 2026.04",{"title":310,"description":323},[84,288,336,337,338],"OpenCV","GraspNet","ur_rtde","projects\u002Furgrasp-agent",[337,85,341,342,343],"手眼标定","RealSense","6D 抓取位姿","YyQkluZioC8fIDzM_lYdoYv_sstvJ1FHw1GVcy43mYU",1791660993166]