[{"data":1,"prerenderedAt":130},["ShallowReactive",2],{"\u002Fen\u002Fglossary\u002Frobot-data-collection":3},{"id":4,"title":5,"alternateName":6,"body":7,"description":120,"extension":121,"keywords":122,"meta":123,"navigation":124,"path":125,"seo":126,"stem":127,"updated":128,"__hash__":129},"glossary\u002Fglossary\u002Fen\u002Frobot-data-collection.md","Robot Data Collection","具身智能数据采集 \u002F 数采",{"type":8,"value":9,"toc":113},"minimark",[10,15,34,39,42,66,74,78,98,101,105],[11,12,14],"h1",{"id":13},"what-is-robot-data-collection","What Is Robot Data Collection?",[16,17,18,22,23,28,29,33],"p",{},[19,20,21],"strong",{},"Robot data collection"," (embodied-AI data collection) is the process of synchronously recording paired data — observations (camera images, joint states) plus actions (joint commands) — while a robot performs a task, typically demonstrated by a human via ",[24,25,27],"a",{"href":26},"\u002Fen\u002Fglossary\u002Fteleoperation","teleoperation",". This data is the raw material for training ",[24,30,32],{"href":31},"\u002Fen\u002Fglossary\u002Fvla-model","VLA models"," and imitation-learning policies: what a model can learn is bounded by what was collected.",[35,36,38],"h2",{"id":37},"the-leader-follower-teleoperation-pipeline","The Leader-Follower Teleoperation Pipeline",[16,40,41],{},"Leader-follower arms are a common real-robot collection setup:",[43,44,45,57,60,63],"ol",{},[46,47,48,49,52,53,56],"li",{},"The operator moves a lightweight ",[19,50,51],{},"leader"," arm; the ",[19,54,55],{},"follower"," arm mirrors its joint motion in real time;",[46,58,59],{},"The system records multi-camera images, follower joint states, and action commands at a fixed rate (typically 30-50 Hz or higher);",[46,61,62],{},"Each completed task is saved as one trajectory (episode); a task usually needs tens to hundreds of episodes;",[46,64,65],{},"Data is filtered and aligned, then fed into the training pipeline.",[16,67,68,69,73],{},"When leader and follower correspond joint-to-joint, demonstration is intuitive and can reduce online kinematic-mapping complexity. The BXI UpperBody 1 ",[24,70,72],{"href":71},"\u002Fen\u002Frobots\u002Frobotic-arms","dual-arm teleoperation platform"," uses a dual-leader, dual-follower architecture and can integrate with ROS2 recording workflows; available interfaces and data formats depend on the project configuration.",[35,75,77],{"id":76},"real-robot-data-vs-simulation-data","Real-Robot Data vs Simulation Data",[79,80,81,87],"ul",{},[46,82,83,86],{},[19,84,85],{},"Real-robot data"," captures true friction, deformation, lighting, and sensor noise, so trained policies deploy directly — but collection is labor-intensive and costly;",[46,88,89,92,93,97],{},[19,90,91],{},"Simulation data"," scales cheaply and in parallel, but differs from reality and must be bridged via ",[24,94,96],{"href":95},"\u002Fen\u002Fglossary\u002Fsim-to-real","Sim-to-Real"," techniques.",[16,99,100],{},"In practice the two are mixed: simulation provides scale, real data anchors the true distribution.",[35,102,104],{"id":103},"how-imitation-learning-consumes-the-data","How Imitation Learning Consumes the Data",[16,106,107,108,112],{},"The ALOHA\u002FACT line of work trains policies directly on leader-follower bimanual trajectories: ACT (Action Chunking with Transformers) predicts a short chunk of future actions as a unit. The required demonstration count depends on task difficulty, data quality, and the desired generalization range. These methods make teleoperated data collection plus imitation learning a practical route for training ",[24,109,111],{"href":110},"\u002Fen\u002Fglossary\u002Fembodied-ai","embodied-AI"," policies.",{"title":114,"searchDepth":115,"depth":115,"links":116},"",2,[117,118,119],{"id":37,"depth":115,"text":38},{"id":76,"depth":115,"text":77},{"id":103,"depth":115,"text":104},"Robot data collection records paired observations and actions, often through teleoperation, to train VLA models and imitation-learning policies.","md","robot data collection, leader-follower teleoperation, imitation learning data, embodied AI data, robot demonstrations",{},true,"\u002Fglossary\u002Fen\u002Frobot-data-collection",{"title":5,"description":120},"glossary\u002Fen\u002Frobot-data-collection",null,"Ty5EvZhhCY5tmbSlQ0S873D9P5kygXy35DEriyy-UGw",1785156467453]