[{"data":1,"prerenderedAt":2188},["ShallowReactive",2],{"glossary-list-en":3},[4,184,258,331,429,499,584,668,755,889,952,1049,1124,1197,1264,1426,1519,1628,1705,1842,1926,2033,2118],{"id":5,"title":6,"alternateName":7,"body":8,"description":174,"extension":175,"keywords":176,"meta":177,"navigation":178,"path":179,"seo":180,"stem":181,"updated":182,"__hash__":183},"glossary\u002Fglossary\u002Fen\u002Fcoreless-motor.md","Coreless Motor","空心杯电机",{"type":9,"value":10,"toc":168},"minimark",[11,16,33,38,46,74,77,81,151],[12,13,15],"h1",{"id":14},"what-is-a-coreless-motor","What Is a Coreless Motor?",[17,18,19,20,24,25,28,29,32],"p",{},"A ",[21,22,23],"strong",{},"coreless motor"," (ironless motor) uses a ",[21,26,27],{},"self-supporting cup-shaped winding without an iron core"," as its rotor — the copper cylinder itself spins in the magnetic field. Removing the iron eliminates iron losses and cogging torque, cuts rotor inertia by roughly an order of magnitude versus an iron-core motor of the same size, and pushes the electrical time constant ",[21,30,31],{},"below 1 ms"," for very fast start-stop response.",[34,35,37],"h2",{"id":36},"why-dexterous-hands-use-coreless-motors","Why Dexterous Hands Use Coreless Motors",[17,39,19,40,45],{},[41,42,44],"a",{"href":43},"\u002Fen\u002Fglossary\u002Fdexterous-hand","dexterous hand"," finger has only 10–20 mm of diameter to work with, yet must grasp at high frequency:",[47,48,49,56,62,68],"ul",{},[50,51,52,55],"li",{},[21,53,54],{},"Compact",": a 4–16 mm coreless motor plus a micro planetary gearbox or leadscrew drives one finger;",[50,57,58,61],{},[21,59,60],{},"Low inertia, fast response",": finger open\u002Fclose cycles of several Hz are achievable;",[50,63,64,67],{},[21,65,66],{},"No cogging",": smooth grasp force, better fingertip force control;",[50,69,70,73],{},[21,71,72],{},"Efficient",": typically 70%–90%, important with the limited heat dissipation inside a palm.",[17,75,76],{},"The trade-off: the cup winding dissipates heat poorly and tolerates little overload — continuous torque sits in the mN·m range, so a high gear ratio is mandatory for usable grip force.",[34,78,80],{"id":79},"coreless-vs-frameless-torque-motor-division-of-labor","Coreless vs. Frameless Torque Motor: Division of Labor",[82,83,84,103],"table",{},[85,86,87],"thead",{},[88,89,90,94,97],"tr",{},[91,92,93],"th",{},"Dimension",[91,95,96],{},"Coreless motor",[91,98,99],{},[41,100,102],{"href":101},"\u002Fen\u002Fglossary\u002Fframeless-torque-motor","Frameless torque motor",[104,105,106,118,129,140],"tbody",{},[88,107,108,112,115],{},[109,110,111],"td",{},"Rotor",[109,113,114],{},"Ironless cup winding",[109,116,117],{},"Iron-core stator\u002Frotor kit",[88,119,120,123,126],{},[109,121,122],{},"Typical diameter",[109,124,125],{},"4–40 mm",[109,127,128],{},"40–120 mm",[88,130,131,134,137],{},[109,132,133],{},"Torque scale",[109,135,136],{},"mN·m (needs high ratio)",[109,138,139],{},"N·m (low ratio suffices)",[88,141,142,145,148],{},[109,143,144],{},"Typical location",[109,146,147],{},"Fingers, micro actuators",[109,149,150],{},"Arm\u002Fleg joint actuators",[17,152,153,154,157,158,162,163,167],{},"One-line split: ",[21,155,156],{},"coreless for fingers, frameless torque motors for arm and leg joints",". BXI's ",[41,159,161],{"href":160},"\u002Fen\u002Fmotors\u002Fadvanced-motors","85\u002F70\u002F50-series joint actuators"," take the frameless-motor-plus-planetary ",[41,164,166],{"href":165},"\u002Fen\u002Fglossary\u002Fquasi-direct-drive","quasi-direct-drive"," route for high-torque limbs, complementing coreless-driven hands at the end effector.",{"title":169,"searchDepth":170,"depth":170,"links":171},"",2,[172,173],{"id":36,"depth":170,"text":37},{"id":79,"depth":170,"text":80},"A coreless motor uses an ironless cup winding for low inertia, low cogging, and fast response in dexterous hands and other compact precision drives.","md","coreless motor, ironless motor, coreless motor dexterous hand, coreless vs frameless motor, low inertia motor",{},true,"\u002Fglossary\u002Fen\u002Fcoreless-motor",{"title":6,"description":174},"glossary\u002Fen\u002Fcoreless-motor",null,"e0EHSclkEwdKjY17rtR8FkYZzFMBRsPJGnY5r-DSWqw",{"id":185,"title":186,"alternateName":187,"body":188,"description":251,"extension":175,"keywords":252,"meta":253,"navigation":178,"path":254,"seo":255,"stem":256,"updated":182,"__hash__":257},"glossary\u002Fglossary\u002Fen\u002Fcrossed-roller-bearing.md","Crossed Roller Bearing","交叉滚子轴承",{"type":9,"value":189,"toc":248},[190,194,200,204,211,241],[12,191,193],{"id":192},"what-is-a-crossed-roller-bearing","What Is a Crossed Roller Bearing?",[17,195,19,196,199],{},[21,197,198],{},"crossed roller bearing"," is a precision bearing whose cylindrical rollers alternate orientation at 90° to each other inside a V-shaped raceway. Because adjacent rollers face opposite directions, one bearing simultaneously carries radial loads, bidirectional axial loads, and overturning (moment) loads.",[34,201,203],{"id":202},"why-robot-joints-prefer-it","Why Robot Joints Prefer It",[17,205,206,207,210],{},"Standard deep-groove ball bearings mainly take radial load; covering combined loads usually requires paired bearings and a bigger structure. Robot joint loads are inherently combined — a humanoid hip joint simultaneously supports body weight (axial), leg-swing centrifugal force (radial), and the bending moment of a cantilevered leg (overturning). A crossed roller bearing handles ",[21,208,209],{},"all three load types in a single bearing position",":",[47,212,213,219,230],{},[50,214,215,218],{},[21,216,217],{},"High stiffness",": roller line contact resists deformation better than ball point contact.",[50,220,221,224,225,229],{},[21,222,223],{},"High rotational precision",": a stable reference for output-side measurement by ",[41,226,228],{"href":227},"\u002Fen\u002Fglossary\u002Fdual-absolute-encoder","dual absolute encoders",".",[50,231,232,235,236,240],{},[21,233,234],{},"Space savings",": the single-bearing design keeps ",[41,237,239],{"href":238},"\u002Fen\u002Fglossary\u002Fjoint-motor","joint motors"," thinner and lighter.",[17,242,243,244,247],{},"The BXI ",[41,245,246],{"href":160},"85\u002F70\u002F50-series joint motors"," fit crossed roller bearings at the output across the whole lineup, from load-bearing legs to dexterous arms.",{"title":169,"searchDepth":170,"depth":170,"links":249},[250],{"id":202,"depth":170,"text":203},"A crossed roller bearing alternates cylindrical rollers at 90° in a V-shaped raceway, letting a single bearing carry radial, axial, and moment loads simultaneously — giving robot joints high stiffness and rotational precision.","crossed roller bearing, robot bearing, joint stiffness, moment load",{},"\u002Fglossary\u002Fen\u002Fcrossed-roller-bearing",{"title":186,"description":251},"glossary\u002Fen\u002Fcrossed-roller-bearing","KoMUdgdok7QNxJjSWGk9TwhPtuw-cktDynA8ZmwxPIU",{"id":259,"title":260,"alternateName":261,"body":262,"description":324,"extension":175,"keywords":325,"meta":326,"navigation":178,"path":327,"seo":328,"stem":329,"updated":182,"__hash__":330},"glossary\u002Fglossary\u002Fen\u002Fdegrees-of-freedom.md","Degrees of Freedom (DoF)","自由度",{"type":9,"value":263,"toc":320},[264,268,277,281,301,305,312],[12,265,267],{"id":266},"what-are-degrees-of-freedom","What Are Degrees of Freedom?",[17,269,270,273,274,229],{},[21,271,272],{},"Degrees of freedom (DoF)"," count the independently controllable motion axes of a robot. Each actively driven rotary or prismatic joint counts as one DoF, typically corresponding to one ",[41,275,276],{"href":238},"joint motor",[34,278,280],{"id":279},"dof-determines-capability","DoF Determines Capability",[47,282,283,289,295],{},[50,284,285,288],{},[21,286,287],{},"6 DoF"," is the minimum for a robot arm to reach \"any position + any orientation\" in 3D space (3 for position + 3 for orientation);",[50,290,291,294],{},[21,292,293],{},"7 DoF"," adds a redundant axis, letting the elbow move while the hand stays fixed — dodging obstacles and optimizing posture, exactly the configuration of the human arm;",[50,296,297,300],{},[21,298,299],{},"Humanoid robots"," need whole-body coordination and typically exceed 30 DoF.",[34,302,304],{"id":303},"example-the-31-dof-of-elf-3","Example: The 31 DoF of Elf 3",[17,306,243,307,311],{},[41,308,310],{"href":309},"\u002Fen\u002Frobots\u002Fhumanoid-robot","Elf 3 humanoid robot"," has 31 DoF excluding hands: 6 per leg, 7 per arm, 3 in the waist, 2 in the head. The 6-DoF legs cover walking's hip (3), knee (1), and ankle (2); the 7-DoF arms provide human-like manipulation redundancy; the 3-DoF waist expands the reachable workspace.",[17,313,314,315,319],{},"More DoF means more actuators, more control-bus bandwidth (see ",[41,316,318],{"href":317},"\u002Fen\u002Fglossary\u002Fmit-protocol-can","MIT protocol","), and more weight — making DoF layout one of the central trade-offs in humanoid design.",{"title":169,"searchDepth":170,"depth":170,"links":321},[322,323],{"id":279,"depth":170,"text":280},{"id":303,"depth":170,"text":304},"Degrees of freedom count the independently controllable joint axes of a robot and directly determine its motion capability: a 6-DoF arm reaches arbitrary poses, while humanoid robots typically need 30+ DoF for whole-body coordination.","degrees of freedom, DoF, robot DoF, humanoid robot DoF, redundant DoF",{},"\u002Fglossary\u002Fen\u002Fdegrees-of-freedom",{"title":260,"description":324},"glossary\u002Fen\u002Fdegrees-of-freedom","SkapjgIm8JHQbbVFUk-AwsALlHdvJjGwWAR3WQh8jaY",{"id":332,"title":333,"alternateName":334,"body":335,"description":422,"extension":175,"keywords":423,"meta":424,"navigation":178,"path":425,"seo":426,"stem":427,"updated":182,"__hash__":428},"glossary\u002Fglossary\u002Fen\u002Fdexterous-hand.md","Dexterous Hand","灵巧手",{"type":9,"value":336,"toc":417},[337,341,350,354,362,376,380,400,404],[12,338,340],{"id":339},"what-is-a-dexterous-hand","What Is a Dexterous Hand?",[17,342,19,343,345,346,349],{},[21,344,44],{}," is a multi-degree-of-freedom robotic end-effector modeled on the structure and function of the human hand: coordinated motion of multiple fingers and joints enables grasping, pinching, and in-hand reorientation of objects. Unlike a two-finger gripper with only open\u002Fclose states, a dexterous hand adapts to objects and tools of arbitrary shape, making it the key component for ",[41,347,348],{"href":309},"humanoid robots"," moving toward general-purpose manipulation.",[34,351,353],{"id":352},"degrees-of-freedom-underactuated-vs-fully-actuated","Degrees of Freedom: Underactuated vs Fully Actuated",[17,355,356,357,361],{},"The human hand has roughly 20+ ",[41,358,360],{"href":359},"\u002Fen\u002Fglossary\u002Fdegrees-of-freedom","degrees of freedom","; dexterous-hand design trades off human-likeness against engineering reliability:",[47,363,364,370],{},[50,365,366,369],{},[21,367,368],{},"Underactuated (about 6-12 actively driven DOF)",": fewer motors than joints, with multiple joints coupled to one drive and mechanical compliance conforming to the object. Lower cost, higher reliability, covers most grasping tasks;",[50,371,372,375],{},[21,373,374],{},"Fully actuated (20+ DOF)",": every joint independently controlled, enabling in-hand manipulation and other demanding skills — at the price of complexity, cost, and maintenance, so mostly found on research platforms.",[34,377,379],{"id":378},"actuation-approaches","Actuation Approaches",[47,381,382,388,394],{},[50,383,384,387],{},[21,385,386],{},"Coreless-motor direct\u002Fgeared drive",": exploits the power density of coreless motors to fit actuators inside fingers or the palm — fast response, compact structure;",[50,389,390,393],{},[21,391,392],{},"Tendon drive",": motors sit in the forearm and pull fingers via tendons (like human anatomy), keeping the hand light, but friction and cable elasticity introduce control error;",[50,395,396,399],{},[21,397,398],{},"Linkage drive",": rigid linkages transmit motion with good precision and stiffness, at some cost in DOF layout flexibility.",[34,401,403],{"id":402},"role-in-humanoid-robotics","Role in Humanoid Robotics",[17,405,406,407,411,412,416],{},"The dexterous hand defines what work a humanoid can actually do. In ",[41,408,410],{"href":409},"\u002Fen\u002Fglossary\u002Fembodied-ai","embodied AI"," research, dexterous manipulation data is mostly collected through ",[41,413,415],{"href":414},"\u002Fen\u002Fglossary\u002Fteleoperation","teleoperation"," — data gloves, exoskeletons, VR — to train imitation-learning policies. More hand DOF means a larger learnable task space, but also harder data collection and control, which is why most deployments today still start from grippers or underactuated hands.",{"title":169,"searchDepth":170,"depth":170,"links":418},[419,420,421],{"id":352,"depth":170,"text":353},{"id":378,"depth":170,"text":379},{"id":402,"depth":170,"text":403},"A dexterous hand is a multi-DOF robotic end-effector modeled on the human hand for grasping, tool use, and fine manipulation.","dexterous hand, multi-fingered robot hand, degrees of freedom, humanoid end-effector, robotic manipulation",{},"\u002Fglossary\u002Fen\u002Fdexterous-hand",{"title":333,"description":422},"glossary\u002Fen\u002Fdexterous-hand","z0IGnrbN_DbM3ljry8hdynab2_Dkt4l3k11TDXAHEWg",{"id":430,"title":431,"alternateName":432,"body":433,"description":492,"extension":175,"keywords":493,"meta":494,"navigation":178,"path":495,"seo":496,"stem":497,"updated":182,"__hash__":498},"glossary\u002Fglossary\u002Fen\u002Fdual-absolute-encoder.md","Dual Absolute Encoder","双绝对值编码器",{"type":9,"value":434,"toc":488},[435,439,448,452,459,479,483],[12,436,438],{"id":437},"what-is-a-dual-absolute-encoder","What Is a Dual Absolute Encoder?",[17,440,19,441,444,445,447],{},[21,442,443],{},"dual absolute encoder"," configuration is the encoder architecture of high-end ",[41,446,239],{"href":238},": one absolute encoder on the motor rotor (input side, before reduction) and one on the output flange (output side, after reduction), measuring angle directly at both ends.",[34,449,451],{"id":450},"why-two-encoders","Why Two Encoders?",[17,453,454,455,458],{},"With only a motor-side encoder, output angle must be inferred as \"motor angle ÷ reduction ratio\" — and gearbox backlash, elastic deformation, and assembly tolerance all pull that estimate away from the true joint angle. An output-side encoder ",[21,456,457],{},"measures the real post-reduction angle directly",", so the control loop closes on ground truth:",[47,460,461,467,473],{},[50,462,463,466],{},[21,464,465],{},"Higher precision",": angle errors from backlash and deformation are eliminated.",[50,468,469,472],{},[21,470,471],{},"Zero-calibration startup",": absolute encoders retain position through power loss, so the robot knows every joint pose at power-on with no homing routine.",[50,474,475,478],{},[21,476,477],{},"Redundancy",": input and output angles cross-check each other, catching anomalies such as gearbox slip.",[34,480,482],{"id":481},"implementation","Implementation",[17,484,243,485,487],{},[41,486,246],{"href":160}," pair a magnetic encoder on the input with an inductive encoder on the output — two sensing principles that don't interfere — achieving true dual-encoder closed loop in a compact package.",{"title":169,"searchDepth":170,"depth":170,"links":489},[490,491],{"id":450,"depth":170,"text":451},{"id":481,"depth":170,"text":482},"A dual absolute encoder setup places one absolute encoder at the joint motor input and another at the output, directly measuring true post-reduction joint angle for precise closed-loop control and zero-calibration startup.","dual absolute encoder, absolute encoder, joint angle sensing, zero calibration startup",{},"\u002Fglossary\u002Fen\u002Fdual-absolute-encoder",{"title":431,"description":492},"glossary\u002Fen\u002Fdual-absolute-encoder","G9sod0p6BHMaU_XSx7PUCyIPIdnjXMM1pHPzOdRf-WM",{"id":500,"title":501,"alternateName":502,"body":503,"description":577,"extension":175,"keywords":578,"meta":579,"navigation":178,"path":580,"seo":581,"stem":582,"updated":182,"__hash__":583},"glossary\u002Fglossary\u002Fen\u002Fembodied-ai.md","Embodied AI","具身智能 \u002F Embodied Intelligence",{"type":9,"value":504,"toc":573},[505,509,514,518,528,532,568],[12,506,508],{"id":507},"what-is-embodied-ai","What Is Embodied AI?",[17,510,511,513],{},[21,512,501],{}," (embodied intelligence) is the paradigm in which an intelligent agent has a physical body and accomplishes perception, decision-making, and action through interaction with the real environment. Unlike \"disembodied\" models that only process text or images, embodied AI must close the loop in the physical world: see → understand → act → observe the result → correct.",[34,515,517],{"id":516},"why-humanoids-are-the-primary-platform","Why Humanoids Are the Primary Platform",[17,519,520,521,524,525,527],{},"Human environments — stairs, door handles, tools, workstations — are designed for the human body. The humanoid form lets robots reuse this infrastructure ",[21,522,523],{},"without modifying the environment",", and lets vast human motion data (video, motion capture, ",[41,526,415],{"href":414}," demonstrations) transfer directly into training data.",[34,529,531],{"id":530},"the-stack","The Stack",[47,533,534,546,552,562],{},[50,535,536,539,540,542,543,545],{},[21,537,538],{},"Body",": high-dynamic hardware is the prerequisite — torque-dense ",[41,541,239],{"href":238},", high-frequency control buses, and whole-body ",[41,544,360],{"href":359},";",[50,547,548,551],{},[21,549,550],{},"Perception",": multimodal sensing — vision, depth, IMU, touch;",[50,553,554,557,558,545],{},[21,555,556],{},"Decision",": foundation-model or reinforcement-learning policies, often transferred from simulation via ",[41,559,561],{"href":560},"\u002Fen\u002Fglossary\u002Fsim-to-real","Sim-to-Real",[50,563,564,567],{},[21,565,566],{},"Data",": teleoperated real-robot data collection for imitation learning.",[17,569,243,570,572],{},[41,571,310],{"href":309}," ships with a ROS2 SDK and MuJoCo simulation — an open platform built for embodied-AI research.",{"title":169,"searchDepth":170,"depth":170,"links":574},[575,576],{"id":516,"depth":170,"text":517},{"id":530,"depth":170,"text":531},"Embodied AI is the paradigm where an agent perceives, decides, and acts through a physical body interacting with the real world; humanoid robots are its primary platform, closing the perception–decision–action loop in reality.","embodied AI, embodied intelligence, humanoid robot, perception decision action, agents",{},"\u002Fglossary\u002Fen\u002Fembodied-ai",{"title":501,"description":577},"glossary\u002Fen\u002Fembodied-ai","C6NRbR956x72lE_j5DM_QaS5e_lb5_lRYo0H3qL44lY",{"id":585,"title":586,"alternateName":587,"body":588,"description":661,"extension":175,"keywords":662,"meta":663,"navigation":178,"path":664,"seo":665,"stem":666,"updated":182,"__hash__":667},"glossary\u002Fglossary\u002Fen\u002Fforce-control.md","Force Control (Impedance \u002F Admittance Control)","力控（阻抗控制\u002F导纳控制）",{"type":9,"value":589,"toc":657},[590,594,604,608,636,640],[12,591,593],{"id":592},"what-is-force-control-in-robotics","What Is Force Control in Robotics?",[17,595,596,599,600,603],{},[21,597,598],{},"Force control"," means a robot joint regulates its ",[21,601,602],{},"output force\u002Ftorque"," rather than just its position, giving the robot a controllable \"softness\" when contacting the environment or people. Landing a step, carrying objects with two arms, and human-robot collaboration all depend on it — a position-controlled joint rams through unexpected contact, while a force-controlled joint yields.",[34,605,607],{"id":606},"the-three-approaches","The Three Approaches",[47,609,610,624,630],{},[50,611,612,615,616,620,621,623],{},[21,613,614],{},"Torque control",": command joint torque directly — the lowest-level, fastest loop; the ",[617,618,619],"code",{},"T_ff"," feed-forward torque in the ",[41,622,318],{"href":317}," serves exactly this;",[50,625,626,629],{},[21,627,628],{},"Impedance control",": command a virtual spring-damper (stiffness Kp, damping Kd) so position error maps to torque — the standard for legged robots; one MIT-protocol frame carries Kp\u002FKd and torque feed-forward together;",[50,631,632,635],{},[21,633,634],{},"Admittance control",": measure external force, output position corrections — common on arms with wrist force sensors, bandwidth-limited by the position loop.",[34,637,639],{"id":638},"do-you-need-a-torque-sensor","Do You Need a Torque Sensor?",[17,641,642,643,646,647,649,650,654,655,229],{},"Friction, efficiency variation, and hysteresis make current-based output-torque estimation less accurate in high-ratio joints, so precision applications often use joint or end-effector force sensors. A low-ratio ",[41,644,645],{"href":165},"quasi-direct-drive (QDD)"," joint can estimate output torque from phase current after calibrating motor torque constant, gearbox efficiency, and friction. This supports some proprioceptive and impedance-control tasks, but it does not replace sensing for every high-accuracy or safety-critical application. BXI ",[41,648,161],{"href":160}," support joint control through a 19.5:1 ",[41,651,653],{"href":652},"\u002Fen\u002Fglossary\u002Fplanetary-gearbox","planetary design",", MIT-protocol CAN commands, and ",[41,656,228],{"href":227},{"title":169,"searchDepth":170,"depth":170,"links":658},[659,660],{"id":606,"depth":170,"text":607},{"id":638,"depth":170,"text":639},"Robot force control uses torque, impedance, or admittance control for compliant interaction. Calibrated QDD joints can estimate output torque from phase current.","force control, impedance control, admittance control, torque control, compliant robot control",{},"\u002Fglossary\u002Fen\u002Fforce-control",{"title":586,"description":661},"glossary\u002Fen\u002Fforce-control","52y-BfXDNxdRGaxDRO3W2bWFON4Judq3kr79hmCJNpA",{"id":669,"title":670,"alternateName":671,"body":672,"description":748,"extension":175,"keywords":749,"meta":750,"navigation":178,"path":751,"seo":752,"stem":753,"updated":182,"__hash__":754},"glossary\u002Fglossary\u002Fen\u002Fframeless-torque-motor.md","Frameless Torque Motor","无框力矩电机",{"type":9,"value":673,"toc":744},[674,678,692,696,729,733],[12,675,677],{"id":676},"what-is-a-frameless-torque-motor","What Is a Frameless Torque Motor?",[17,679,19,680,683,684,687,688,691],{},[21,681,682],{},"frameless torque motor"," is delivered as ",[21,685,686],{},"a stator and rotor",", without its own housing, output shaft, or bearings. Both parts integrate directly into the joint structure, reducing redundant packaging and shortening the drivetrain. It is common in ",[41,689,690],{"href":238},"integrated joint actuators"," and humanoid robot joints.",[34,693,695],{"id":694},"why-joint-modules-use-frameless-motors","Why Joint Modules Use Frameless Motors",[47,697,698,708,718],{},[50,699,700,703,704,707],{},[21,701,702],{},"Structural fusion",": the stator is heat-fitted into the joint housing and the rotor couples straight to the ",[41,705,706],{"href":652},"planetary gearbox"," sun gear — the shortest possible drivetrain, shrinking joint volume by roughly 30%;",[50,709,710,713,714,545],{},[21,711,712],{},"Large-diameter pancake form",": torque motors use high pole counts at a large air-gap radius for low-speed, high-torque output — exactly a joint's duty profile, and the source of high ",[41,715,717],{"href":716},"\u002Fen\u002Fglossary\u002Ftorque-density","torque density",[50,719,720,723,724,728],{},[21,721,722],{},"Coaxial hollow bore",": the annular stator\u002Frotor geometry naturally supports ",[41,725,727],{"href":726},"\u002Fen\u002Fglossary\u002Fhollow-shaft-motor","hollow-shaft cable routing"," through the joint's center.",[34,730,732],{"id":731},"inrunner-vs-outrunner","Inrunner vs. Outrunner",[17,734,735,736,738,739,741,742,229],{},"Frameless torque motors come as inrunners (rotor inside — better cooling, lower inertia) and outrunners (rotor outside — more torque per volume, flatter). Dynamic legged robots often pick outrunners for torque density. The BXI ",[41,737,161],{"href":160}," build on a frameless torque motor with a 19.5:1 planetary reduction in a ",[41,740,166],{"href":165}," architecture, completed by ",[41,743,228],{"href":227},{"title":169,"searchDepth":170,"depth":170,"links":745},[746,747],{"id":694,"depth":170,"text":695},{"id":731,"depth":170,"text":732},"A frameless torque motor supplies a stator and rotor for direct integration into a robot joint, reducing redundant housing and drivetrain components.","frameless torque motor, frameless motor, torque motor, integrated actuator module, humanoid robot motor",{},"\u002Fglossary\u002Fen\u002Fframeless-torque-motor",{"title":670,"description":748},"glossary\u002Fen\u002Fframeless-torque-motor","S2_rt4hWBhGbQhmeTrBTXxK41bFkfZ-h2hgu49ITg0s",{"id":756,"title":757,"alternateName":758,"body":759,"description":882,"extension":175,"keywords":883,"meta":884,"navigation":178,"path":885,"seo":886,"stem":887,"updated":182,"__hash__":888},"glossary\u002Fglossary\u002Fen\u002Fharmonic-drive.md","Harmonic Drive (Strain Wave Gear)","谐波减速器",{"type":9,"value":760,"toc":878},[761,765,779,783,856,862,866],[12,762,764],{"id":763},"what-is-a-harmonic-drive","What Is a Harmonic Drive?",[17,766,19,767,770,771,774,775,778],{},[21,768,769],{},"harmonic drive"," (strain wave gear) transmits torque by elastically deforming a thin-walled ",[21,772,773],{},"flexspline"," against a rigid circular spline via an elliptical wave generator, achieving a ",[21,776,777],{},"50–160:1 reduction in a single stage"," with near-zero backlash in a compact, lightweight package. It is one of the three precision robot gearbox families alongside cycloidal (RV) and planetary drives.",[34,780,782],{"id":781},"harmonic-drive-vs-planetary-gearbox","Harmonic Drive vs. Planetary Gearbox",[82,784,785,799],{},[85,786,787],{},[88,788,789,791,794],{},[91,790,93],{},[91,792,793],{},"Harmonic drive",[91,795,796],{},[41,797,798],{"href":652},"Planetary gearbox",[104,800,801,812,823,834,845],{},[88,802,803,806,809],{},[109,804,805],{},"Single-stage ratio",[109,807,808],{},"50–160:1",[109,810,811],{},"3–10:1 (stackable)",[88,813,814,817,820],{},[109,815,816],{},"Backlash",[109,818,819],{},"Near zero",[109,821,822],{},"Small but present",[88,824,825,828,831],{},[109,826,827],{},"Impact tolerance",[109,829,830],{},"Flexspline is shock-sensitive",[109,832,833],{},"Solid gears, robust",[88,835,836,839,842],{},[109,837,838],{},"Backdrivability",[109,840,841],{},"Poor (high friction)",[109,843,844],{},"Good at low ratios",[88,846,847,850,853],{},[109,848,849],{},"Typical joints",[109,851,852],{},"Cobot wrists\u002Felbows, humanoid arms",[109,854,855],{},"Legs, high-dynamic joints",[17,857,858,859,229],{},"One-line verdict: ",[21,860,861],{},"choose harmonic for precision, planetary for impact tolerance and force control",[34,863,865],{"id":864},"how-robots-choose","How Robots Choose",[17,867,868,869,871,872,874,875,877],{},"Positioning accuracy makes harmonic drives the default in industrial and collaborative robot arm joints. But humanoid and quadruped legs must absorb landing impacts and estimate external forces from motor current, where the high friction and flexspline fatigue of a high-ratio harmonic drive fall short — so legs overwhelmingly use the ",[41,870,645],{"href":165}," planetary route. The BXI ",[41,873,161],{"href":160}," use a 19.5:1 planetary design balancing ",[41,876,717],{"href":716}," with backdrivability. Base joints needing extreme torque often use cycloidal (RV) gears — hence the common split: RV at the base, harmonic in the arms, planetary in the legs.",{"title":169,"searchDepth":170,"depth":170,"links":879},[880,881],{"id":781,"depth":170,"text":782},{"id":864,"depth":170,"text":865},"A harmonic drive uses flexspline deformation for high reduction and low backlash. Robot arms commonly use it; dynamic legs may prioritize backdrivability.","harmonic drive, strain wave gear, harmonic drive vs planetary gearbox, robot gearbox, zero backlash",{},"\u002Fglossary\u002Fen\u002Fharmonic-drive",{"title":757,"description":882},"glossary\u002Fen\u002Fharmonic-drive","ocMmVbwr6yESM3uDuoTZB4rUCTghY-jU9rdjZ38vQxQ",{"id":890,"title":891,"alternateName":892,"body":893,"description":945,"extension":175,"keywords":946,"meta":947,"navigation":178,"path":948,"seo":949,"stem":950,"updated":182,"__hash__":951},"glossary\u002Fglossary\u002Fen\u002Fhollow-shaft-motor.md","Hollow Shaft Motor","中空轴电机",{"type":9,"value":894,"toc":941},[895,899,905,909,912,923,928,932],[12,896,898],{"id":897},"what-is-a-hollow-shaft-motor","What Is a Hollow Shaft Motor?",[17,900,19,901,904],{},[21,902,903],{},"hollow shaft motor"," is a motor or actuator whose output shaft has a through-bore along its rotation axis. Unlike solid-shaft designs, cables, hydraulic lines, and sensor harnesses can pass straight through the joint's center of rotation instead of looping around the outside.",[34,906,908],{"id":907},"why-robots-need-hollow-shafts","Why Robots Need Hollow Shafts",[17,910,911],{},"High-DOF robots — such as a 31-DOF humanoid — chain many joints in series. If every joint routes its cables externally:",[47,913,914,917,920],{},[50,915,916],{},"harnesses fatigue and fail where joints flex repeatedly;",[50,918,919],{},"exposed cables limit range of motion and snag on the environment;",[50,921,922],{},"wiring complexity and assembly time grow rapidly with joint count.",[17,924,925,926,229],{},"Routing through the center bore solves all three at once, which is why hollow-shaft designs are central to modern integrated ",[41,927,239],{"href":238},[34,929,931],{"id":930},"engineering-trade-offs","Engineering Trade-offs",[17,933,934,935,937,938,940],{},"A larger bore passes more wiring but complicates the layout of the rotor, encoders, and gearbox. The BXI ",[41,936,246],{"href":160}," provide 6–10 mm hollow bores, using a ",[41,939,706],{"href":652}," to balance torque density against routing space.",{"title":169,"searchDepth":170,"depth":170,"links":942},[943,944],{"id":907,"depth":170,"text":908},{"id":930,"depth":170,"text":931},"A hollow shaft motor has a through-bore along its output axis, letting cables, tubing, and sensor harnesses pass through the joint center — dramatically simplifying robot wiring and expanding joint range of motion.","hollow shaft motor, hollow bore actuator, robot cable routing, joint motor design",{},"\u002Fglossary\u002Fen\u002Fhollow-shaft-motor",{"title":891,"description":945},"glossary\u002Fen\u002Fhollow-shaft-motor","8AXKSNUXkdra2vm9kBh5SEf95zF0eH7tQEh9y--ROE4",{"id":953,"title":954,"alternateName":955,"body":956,"description":1042,"extension":175,"keywords":1043,"meta":1044,"navigation":178,"path":1045,"seo":1046,"stem":1047,"updated":182,"__hash__":1048},"glossary\u002Fglossary\u002Fen\u002Fjoint-motor.md","Joint Motor","关节电机 \u002F Robot Actuator",{"type":9,"value":957,"toc":1038},[958,962,967,971,1013,1017,1031],[12,959,961],{"id":960},"what-is-a-joint-motor","What Is a Joint Motor?",[17,963,19,964,966],{},[21,965,276],{}," (also called a robot joint module or integrated actuator) packs a frameless torque motor, gearbox, encoders, driver, and bearings into one compact unit. Mounted directly at a robot's joints, it outputs controlled torque and rotation — the \"muscle\" of humanoid robots, quadrupeds, and robotic arms.",[34,968,970],{"id":969},"core-components","Core Components",[47,972,973,978,987,995,1004],{},[50,974,975,977],{},[21,976,102],{},": generates raw torque and sets the power ceiling.",[50,979,980,983,984,986],{},[21,981,982],{},"Gearbox",": trades speed for torque; common choices are the ",[41,985,706],{"href":652}," and harmonic drive.",[50,988,989,992,993,229],{},[21,990,991],{},"Encoders",": measure joint angle for closed-loop control; high-end designs use ",[41,994,228],{"href":227},[50,996,997,1000,1001,229],{},[21,998,999],{},"Bearing",": carries loads while preserving rotational precision; load-bearing joints typically use a ",[41,1002,198],{"href":1003},"\u002Fen\u002Fglossary\u002Fcrossed-roller-bearing",[50,1005,1006,1009,1010,1012],{},[21,1007,1008],{},"Driver",": runs current\u002Fvelocity\u002Fposition loops and receives commands over CAN\u002FCANFD buses (see ",[41,1011,318],{"href":317},").",[34,1014,1016],{"id":1015},"key-specifications","Key Specifications",[17,1018,1019,1020,1023,1024,1027,1028,1030],{},"The parameters that matter most are ",[21,1021,1022],{},"rated torque"," (continuous output), ",[21,1025,1026],{},"peak torque"," (short-term ceiling), weight, envelope dimensions, reduction ratio, and communication interface. A humanoid robot typically needs 20–40 joint motors across several torque tiers — the BXI Elf 3, for example, runs 31 ",[41,1029,246],{"href":160}," spanning 25–150 N·m peak torque.",[17,1032,1033,1034,229],{},"For a sizing methodology, see the ",[41,1035,1037],{"href":1036},"\u002Fen\u002Fblog\u002Fjoint-motor-selection-guide","joint motor selection guide",{"title":169,"searchDepth":170,"depth":170,"links":1039},[1040,1041],{"id":969,"depth":170,"text":970},{"id":1015,"depth":170,"text":1016},"A joint motor is an integrated robot actuator combining a frameless motor, gearbox, encoders, and driver in one unit, mounted directly at a robot joint — the core power component of humanoid robots and robotic arms.","joint motor, robot actuator, integrated actuator, robot joint module",{},"\u002Fglossary\u002Fen\u002Fjoint-motor",{"title":954,"description":1042},"glossary\u002Fen\u002Fjoint-motor","x9EEPuA3LvSXMHvXgBzkE8Esmeqaip4sfS4pTLlHaL8",{"id":1050,"title":1051,"alternateName":1052,"body":1053,"description":1117,"extension":175,"keywords":1118,"meta":1119,"navigation":178,"path":1120,"seo":1121,"stem":1122,"updated":182,"__hash__":1123},"glossary\u002Fglossary\u002Fen\u002Fmit-protocol-can.md","MIT Protocol (CAN Motor Control)","MIT 协议 \u002F MIT Mode",{"type":9,"value":1054,"toc":1113},[1055,1059,1065,1069,1076,1095,1099],[12,1056,1058],{"id":1057},"what-is-the-mit-protocol","What Is the MIT Protocol?",[17,1060,1061,1062,1064],{},"The ",[21,1063,318],{}," (MIT Mode) is a CAN-bus motor control protocol originating from the MIT Mini Cheetah open-source legged robot. Its core idea: compress a full control command into one CAN frame carrying five quantities — target position p, target velocity v, position gain Kp, velocity gain Kd, and feed-forward torque τ.",[34,1066,1068],{"id":1067},"hybrid-force-position-control","Hybrid Force-Position Control",[17,1070,1071,1072,1075],{},"The motor computes output torque as ",[617,1073,1074],{},"τ_out = Kp·(p − p_actual) + Kd·(v − v_actual) + τ",". By tuning the gains, one protocol covers three control modes:",[47,1077,1078,1084,1090],{},[50,1079,1080,1083],{},[21,1081,1082],{},"Pure torque control",": Kp = Kd = 0, feed-forward torque only — the basis of high-dynamic force control;",[50,1085,1086,1089],{},[21,1087,1088],{},"Position control",": normal Kp\u002FKd values track position like a servo;",[50,1091,1092,1094],{},[21,1093,628],{},": intermediate gains make the joint behave as a spring-damper for compliant contact.",[34,1096,1098],{"id":1097},"why-it-became-a-de-facto-standard","Why It Became a De Facto Standard",[17,1100,1101,1102,1104,1105,1107,1108,1112],{},"The protocol is open, compact (one frame per command), and force-control-native. Many ",[41,1103,276],{"href":238}," vendors now support it, creating a cross-brand control interface. All BXI ",[41,1106,246],{"href":160}," support MIT-protocol CAN\u002FCANFD, and paired with the ",[41,1109,1111],{"href":1110},"\u002Fen\u002Fmotors\u002Fcontrol-modules","PCIE-CAN control module"," achieve >1000 Hz whole-robot control loops.",{"title":169,"searchDepth":170,"depth":170,"links":1114},[1115,1116],{"id":1067,"depth":170,"text":1068},{"id":1097,"depth":170,"text":1098},"The MIT protocol is a CAN-bus motor control scheme from the MIT Mini Cheetah open-source project, sending position, velocity, feed-forward torque, and gains in a single frame for hybrid force-position control — a de facto standard for robot joint motors.","MIT protocol, CAN motor control, hybrid force position control, motor communication protocol, CANFD",{},"\u002Fglossary\u002Fen\u002Fmit-protocol-can",{"title":1051,"description":1117},"glossary\u002Fen\u002Fmit-protocol-can","Q0AyuTGcRHiD67zY9FjjO_dfgKHkDH9mFB7PkB7eLds",{"id":1125,"title":1126,"alternateName":1127,"body":1128,"description":1190,"extension":175,"keywords":1191,"meta":1192,"navigation":178,"path":1193,"seo":1194,"stem":1195,"updated":182,"__hash__":1196},"glossary\u002Fglossary\u002Fen\u002Fplanetary-gearbox.md","Planetary Gearbox","行星减速器",{"type":9,"value":1129,"toc":1186},[1130,1134,1139,1143,1174,1178,1181],[12,1131,1133],{"id":1132},"what-is-a-planetary-gearbox","What Is a Planetary Gearbox?",[17,1135,19,1136,1138],{},[21,1137,706],{}," is a gear train built from a central sun gear (input), several planet gears, and an outer ring gear. The planets spin on their own axes while orbiting the sun — like planets around a star, hence the name. It converts a motor's high speed into low-speed, high-torque output.",[34,1140,1142],{"id":1141},"structural-advantages","Structural Advantages",[47,1144,1145,1151,1160,1166],{},[50,1146,1147,1150],{},[21,1148,1149],{},"Load sharing",": torque is split across multiple planet gears, giving high torque density and impact tolerance.",[50,1152,1153,1156,1157,1159],{},[21,1154,1155],{},"Coaxial input\u002Foutput",": the compact cylindrical form factor suits integrated ",[41,1158,239],{"href":238}," naturally.",[50,1161,1162,1165],{},[21,1163,1164],{},"High efficiency",": single-stage efficiency typically exceeds 95%, higher than harmonic drives.",[50,1167,1168,1171,1172,229],{},[21,1169,1170],{},"Hollow-friendly",": a through-bore at the center enables ",[41,1173,727],{"href":726},[34,1175,1177],{"id":1176},"planetary-vs-harmonic","Planetary vs. Harmonic",[17,1179,1180],{},"Harmonic drives offer large ratios (50–160) and zero backlash but lower efficiency, limited stiffness, and higher cost. Planetary gearboxes have smaller ratios (roughly 3–10 per stage, ~20 with compounding) and slight backlash, but win on efficiency, shock tolerance, and cost — decisive advantages for high-dynamic humanoid leg joints.",[17,1182,243,1183,1185],{},[41,1184,246],{"href":160}," all use a 19.5-ratio planetary design with a uniform 100 rpm rated output speed, so sizing reduces to picking the right torque tier.",{"title":169,"searchDepth":170,"depth":170,"links":1187},[1188,1189],{"id":1141,"depth":170,"text":1142},{"id":1176,"depth":170,"text":1177},"A planetary gearbox uses a sun gear, planet gears, and a ring gear with multiple simultaneous tooth contacts and coaxial input\u002Foutput — delivering high torque density, compactness, and efficiency for robot joint motors.","planetary gearbox, planetary reducer, reduction ratio, robot gearbox, harmonic drive comparison",{},"\u002Fglossary\u002Fen\u002Fplanetary-gearbox",{"title":1126,"description":1190},"glossary\u002Fen\u002Fplanetary-gearbox","MMMfDWhs9jXejonqa7muAjm4xvqSAKEEFnKVsAI7_GE",{"id":1198,"title":1199,"alternateName":1200,"body":1201,"description":1257,"extension":175,"keywords":1258,"meta":1259,"navigation":178,"path":1260,"seo":1261,"stem":1262,"updated":182,"__hash__":1263},"glossary\u002Fglossary\u002Fen\u002Fquasi-direct-drive.md","Quasi-Direct Drive (QDD) Actuator","准直驱执行器",{"type":9,"value":1202,"toc":1253},[1203,1207,1217,1221,1227,1238,1242],[12,1204,1206],{"id":1205},"what-is-a-quasi-direct-drive-actuator","What Is a Quasi-Direct Drive Actuator?",[17,1208,19,1209,1212,1213,1216],{},[21,1210,1211],{},"quasi-direct drive (QDD) actuator"," pairs a large-diameter, high-torque motor with a ",[21,1214,1215],{},"low reduction ratio"," gearbox (typically 5–20), an approach popularized by legged-robot projects such as MIT Cheetah. \"Quasi-direct\" means close to direct drive (no reduction) while keeping a small amount of gearing.",[34,1218,1220],{"id":1219},"the-key-property-backdrivability","The Key Property: Backdrivability",[17,1222,1223,1224,210],{},"The higher the reduction ratio, the harder it is to drive the motor backwards from the output (reflected friction and inertia scale with the ratio squared). A low ratio keeps the joint ",[21,1225,1226],{},"backdrivable",[47,1228,1229,1232,1235],{},[50,1230,1231],{},"external impacts are absorbed by the motor yielding rather than the gears fighting them, protecting the drivetrain during landings and collisions;",[50,1233,1234],{},"output torque can be estimated from motor current after calibrating torque constant, efficiency, and friction, reducing reliance on dedicated joint torque sensors in some applications;",[50,1236,1237],{},"force-control bandwidth stays high — essential for jumping and running.",[34,1239,1241],{"id":1240},"qdd-vs-high-ratio-drives","QDD vs. High-Ratio Drives",[17,1243,1244,1245,1247,1248,1250,1251,1012],{},"High-ratio harmonic and cycloidal drives typically provide high output torque and holding capability but have more backdrive resistance. QDD favors dynamic response and compliant control, so it is common in dynamic legged robots. BXI ",[41,1246,246],{"href":160}," use a 19.5-ratio ",[41,1249,653],{"href":652}," with MIT-protocol control (see ",[41,1252,318],{"href":317},{"title":169,"searchDepth":170,"depth":170,"links":1254},[1255,1256],{"id":1219,"depth":170,"text":1220},{"id":1240,"depth":170,"text":1241},"A QDD actuator pairs a high-torque motor with a low ratio of about 5–20, balancing torque density, backdrivability, and current-based torque estimation.","quasi-direct drive, QDD, backdrivability, proprioceptive actuator, legged robot actuator",{},"\u002Fglossary\u002Fen\u002Fquasi-direct-drive",{"title":1199,"description":1257},"glossary\u002Fen\u002Fquasi-direct-drive","xlu_Wie4IMlkFcek9r6jcdjXKsYAwjvxowbVz6lwUOg",{"id":1265,"title":1266,"alternateName":1267,"body":1268,"description":1419,"extension":175,"keywords":1420,"meta":1421,"navigation":178,"path":1422,"seo":1423,"stem":1424,"updated":182,"__hash__":1425},"glossary\u002Fglossary\u002Fen\u002Frv-reducer.md","RV Reducer (Cycloidal Reducer)","RV 减速器",{"type":9,"value":1269,"toc":1415},[1270,1274,1293,1297,1392,1399,1403],[12,1271,1273],{"id":1272},"what-is-an-rv-reducer","What Is an RV Reducer?",[17,1275,1276,1277,1280,1281,1284,1285,1288,1289,1292],{},"An ",[21,1278,1279],{},"RV reducer"," (a member of the ",[21,1282,1283],{},"cycloidal"," gear family) uses a ",[21,1286,1287],{},"two-stage compound structure: a planetary front stage feeding a cycloidal pin-wheel output stage",". Because the cycloidal disc engages many pins simultaneously, it achieves ",[21,1290,1291],{},"30–200:1 ratios"," with the highest torsional rigidity and shock overload capacity (commonly 2–5x rated) among the three precision reducer families, covering tens to thousands of N·m. The name comes from Nabtesco's RV series and is now generic for the architecture.",[34,1294,1296],{"id":1295},"harmonic-vs-rv-vs-planetary-the-three-way-split","Harmonic vs. RV vs. Planetary: The Three-Way Split",[82,1298,1299,1316],{},[85,1300,1301],{},[88,1302,1303,1305,1307,1312],{},[91,1304,93],{},[91,1306,1279],{},[91,1308,1309],{},[41,1310,793],{"href":1311},"\u002Fen\u002Fglossary\u002Fharmonic-drive",[91,1313,1314],{},[41,1315,798],{"href":652},[104,1317,1318,1330,1344,1355,1368,1379],{},[88,1319,1320,1323,1326,1328],{},[109,1321,1322],{},"Single-unit ratio",[109,1324,1325],{},"30–200:1",[109,1327,808],{},[109,1329,811],{},[88,1331,1332,1335,1338,1341],{},[109,1333,1334],{},"Rigidity \u002F shock",[109,1336,1337],{},"High",[109,1339,1340],{},"Flexspline is relatively shock-sensitive",[109,1342,1343],{},"Good",[88,1345,1346,1348,1351,1353],{},[109,1347,816],{},[109,1349,1350],{},"~1 arcmin",[109,1352,819],{},[109,1354,822],{},[88,1356,1357,1360,1363,1365],{},[109,1358,1359],{},"Weight \u002F size",[109,1361,1362],{},"Large and heavy",[109,1364,54],{},[109,1366,1367],{},"Moderate",[88,1369,1370,1372,1375,1377],{},[109,1371,838],{},[109,1373,1374],{},"Poor",[109,1376,1374],{},[109,1378,844],{},[88,1380,1381,1383,1386,1389],{},[109,1382,849],{},[109,1384,1385],{},"Industrial robot base\u002Fshoulder",[109,1387,1388],{},"Wrists, humanoid arms",[109,1390,1391],{},"Humanoid\u002Fquadruped legs",[17,1393,1394,1395,1398],{},"A common application pattern is ",[21,1396,1397],{},"RV at industrial robot bases, harmonic drives in lightweight arms, and low-ratio planetary gears in backdrivable legs",". The final choice still depends on load, precision, mass, and shock requirements.",[34,1400,1402],{"id":1401},"why-rv-reducers-are-less-common-in-humanoid-legs","Why RV Reducers Are Less Common in Humanoid Legs",[17,1404,1405,1406,1408,1409,1411,1412,1414],{},"A six-axis industrial robot is bolted to the floor, so its base joint benefits from RV rigidity and load capacity. A humanoid must carry every joint it uses: at comparable torque levels, an RV solution is typically heavier than a low-ratio planetary design, reducing whole-robot ",[41,1407,717],{"href":716},". Higher friction and lower backdrivability also increase uncertainty in current-based output-torque estimation. Humanoid legs that prioritize low mass and compliant control therefore often use planetary ",[41,1410,645],{"href":165}," actuators. BXI's ",[41,1413,161],{"href":160}," use a 19.5:1 planetary design with up to 150 N·m peak torque.",{"title":169,"searchDepth":170,"depth":170,"links":1416},[1417,1418],{"id":1295,"depth":170,"text":1296},{"id":1401,"depth":170,"text":1402},"An RV reducer combines planetary and cycloidal stages for high rigidity, load capacity, and shock tolerance in industrial robot base and shoulder joints.","RV reducer, cycloidal reducer, RV reducer vs harmonic drive, industrial robot gearbox, robot reducer",{},"\u002Fglossary\u002Fen\u002Frv-reducer",{"title":1266,"description":1419},"glossary\u002Fen\u002Frv-reducer","Wu3pV6VVSkc_eKz3EM0ck72XmXgM9pVSnQhqAASnEHQ",{"id":1427,"title":1428,"alternateName":1429,"body":1430,"description":1512,"extension":175,"keywords":1513,"meta":1514,"navigation":178,"path":1515,"seo":1516,"stem":1517,"updated":182,"__hash__":1518},"glossary\u002Fglossary\u002Fen\u002Freflected-inertia.md","Reflected Inertia","反射惯量",{"type":9,"value":1431,"toc":1508},[1432,1436,1454,1458,1461,1481,1487,1491],[12,1433,1435],{"id":1434},"what-is-reflected-inertia","What Is Reflected Inertia?",[17,1437,1438,1441,1442,1445,1446,1449,1450,1453],{},[21,1439,1440],{},"Reflected inertia"," is the motor rotor's inertia as seen at the gearbox output side: ",[21,1443,1444],{},"J_reflected = J_rotor × N²",", where N is the gear ratio. Inertia scales with the ",[21,1447,1448],{},"square"," of the ratio — a 100:1 joint presents ",[21,1451,1452],{},"10,000x"," the rotor inertia at its output, while a 10:1 joint presents only 100x.",[34,1455,1457],{"id":1456},"why-high-ratio-joints-feel-stiff-and-break-under-impact","Why High-Ratio Joints Feel Stiff and Break Under Impact",[17,1459,1460],{},"Anything pushing on the joint output must also accelerate that squared-up rotor inertia:",[47,1462,1463,1469,1475],{},[50,1464,1465,1468],{},[21,1466,1467],{},"Poor impact tolerance",": at foot touchdown or collision, the impulse arrives faster than any controller can react, and the energy lands directly on gear teeth and flexsplines — a leading failure mode of high-ratio joints in legs;",[50,1470,1471,1474],{},[21,1472,1473],{},"Hard to backdrive",": reflected inertia plus friction makes the joint nearly immovable by hand, killing passive compliance;",[50,1476,1477,1480],{},[21,1478,1479],{},"Limited force-control bandwidth",": to fake softness, the controller must first fight the huge equivalent inertia, capping active-compliance bandwidth and fidelity.",[17,1482,1483,1484],{},"Quotable one-liner: ",[21,1485,1486],{},"double the gear ratio, quadruple the reflected inertia — the physical reason leg joints avoid high ratios.",[34,1488,1490],{"id":1489},"how-qdd-keeps-reflected-inertia-low","How QDD Keeps Reflected Inertia Low",[17,1492,1493,1494,1496,1497,1500,1501,1504,1505,1507],{},"This is the core trade of the ",[41,1495,645],{"href":165}," approach: use a large-diameter, high-",[41,1498,1499],{"href":716},"torque-density"," motor so the gear ratio can stay low (typically 6–20:1), cutting reflected inertia by ",[21,1502,1503],{},"1–2 orders of magnitude"," versus a harmonic solution (50–160:1). The joint becomes naturally backdrivable, impacts are absorbed softly, and external torque can be estimated from motor current alone. BXI's ",[41,1506,161],{"href":160}," pick a 19.5:1 planetary ratio as exactly this engineering balance between output torque and reflected inertia.",{"title":169,"searchDepth":170,"depth":170,"links":1509},[1510,1511],{"id":1456,"depth":170,"text":1457},{"id":1489,"depth":170,"text":1490},"Reflected inertia is rotor inertia multiplied by gear ratio squared at the output. It directly affects joint backdrivability and impact response.","reflected inertia, gear ratio squared, QDD reflected inertia, backdrivability, joint actuator inertia",{},"\u002Fglossary\u002Fen\u002Freflected-inertia",{"title":1428,"description":1512},"glossary\u002Fen\u002Freflected-inertia","8-6CpndTia9NmkL5OpMklo45cRXWGU1NDmnWQHRnTGQ",{"id":1520,"title":1521,"alternateName":1522,"body":1523,"description":1621,"extension":175,"keywords":1622,"meta":1623,"navigation":178,"path":1624,"seo":1625,"stem":1626,"updated":182,"__hash__":1627},"glossary\u002Fglossary\u002Fen\u002Frobot-data-collection.md","Robot Data Collection","具身智能数据采集 \u002F 数采",{"type":9,"value":1524,"toc":1616},[1525,1529,1543,1547,1550,1573,1581,1585,1602,1605,1609],[12,1526,1528],{"id":1527},"what-is-robot-data-collection","What Is Robot Data Collection?",[17,1530,1531,1534,1535,1537,1538,1542],{},[21,1532,1533],{},"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 ",[41,1536,415],{"href":414},". This data is the raw material for training ",[41,1539,1541],{"href":1540},"\u002Fen\u002Fglossary\u002Fvla-model","VLA models"," and imitation-learning policies: what a model can learn is bounded by what was collected.",[34,1544,1546],{"id":1545},"the-leader-follower-teleoperation-pipeline","The Leader-Follower Teleoperation Pipeline",[17,1548,1549],{},"Leader-follower arms are a common real-robot collection setup:",[1551,1552,1553,1564,1567,1570],"ol",{},[50,1554,1555,1556,1559,1560,1563],{},"The operator moves a lightweight ",[21,1557,1558],{},"leader"," arm; the ",[21,1561,1562],{},"follower"," arm mirrors its joint motion in real time;",[50,1565,1566],{},"The system records multi-camera images, follower joint states, and action commands at a fixed rate (typically 30-50 Hz or higher);",[50,1568,1569],{},"Each completed task is saved as one trajectory (episode); a task usually needs tens to hundreds of episodes;",[50,1571,1572],{},"Data is filtered and aligned, then fed into the training pipeline.",[17,1574,1575,1576,1580],{},"When leader and follower correspond joint-to-joint, demonstration is intuitive and can reduce online kinematic-mapping complexity. The BXI UpperBody 1 ",[41,1577,1579],{"href":1578},"\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.",[34,1582,1584],{"id":1583},"real-robot-data-vs-simulation-data","Real-Robot Data vs Simulation Data",[47,1586,1587,1593],{},[50,1588,1589,1592],{},[21,1590,1591],{},"Real-robot data"," captures true friction, deformation, lighting, and sensor noise, so trained policies deploy directly — but collection is labor-intensive and costly;",[50,1594,1595,1598,1599,1601],{},[21,1596,1597],{},"Simulation data"," scales cheaply and in parallel, but differs from reality and must be bridged via ",[41,1600,561],{"href":560}," techniques.",[17,1603,1604],{},"In practice the two are mixed: simulation provides scale, real data anchors the true distribution.",[34,1606,1608],{"id":1607},"how-imitation-learning-consumes-the-data","How Imitation Learning Consumes the Data",[17,1610,1611,1612,1615],{},"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 ",[41,1613,1614],{"href":409},"embodied-AI"," policies.",{"title":169,"searchDepth":170,"depth":170,"links":1617},[1618,1619,1620],{"id":1545,"depth":170,"text":1546},{"id":1583,"depth":170,"text":1584},{"id":1607,"depth":170,"text":1608},"Robot data collection records paired observations and actions, often through teleoperation, to train VLA models and imitation-learning policies.","robot data collection, leader-follower teleoperation, imitation learning data, embodied AI data, robot demonstrations",{},"\u002Fglossary\u002Fen\u002Frobot-data-collection",{"title":1521,"description":1621},"glossary\u002Fen\u002Frobot-data-collection","Ty5EvZhhCY5tmbSlQ0S873D9P5kygXy35DEriyy-UGw",{"id":1629,"title":561,"alternateName":1630,"body":1631,"description":1698,"extension":175,"keywords":1699,"meta":1700,"navigation":178,"path":1701,"seo":1702,"stem":1703,"updated":182,"__hash__":1704},"glossary\u002Fglossary\u002Fen\u002Fsim-to-real.md","仿真到实机迁移",{"type":9,"value":1632,"toc":1694},[1633,1637,1642,1646,1653,1676,1680],[12,1634,1636],{"id":1635},"what-is-sim-to-real","What Is Sim-to-Real?",[17,1638,1639,1641],{},[21,1640,561],{}," is the approach of training robot control policies in a physics simulator (such as MuJoCo or Isaac Gym\u002FLab) and then deploying them on real hardware. Training on real robots is slow, expensive, and breaks hardware; simulation runs thousands of robots in parallel at faster-than-real-time speed, which is what makes reinforcement learning of humanoid walking practical at all.",[34,1643,1645],{"id":1644},"the-core-challenge-the-reality-gap","The Core Challenge: The Reality Gap",[17,1647,1648,1649,1652],{},"Simulation never matches reality exactly — friction, latency, motor characteristics, and sensor noise all differ. This mismatch is the ",[21,1650,1651],{},"sim-to-real gap",". The main mitigations:",[47,1654,1655,1661,1667],{},[50,1656,1657,1660],{},[21,1658,1659],{},"Domain randomization",": randomly perturb simulation parameters (mass, friction, latency) during training, forcing the policy to become robust;",[50,1662,1663,1666],{},[21,1664,1665],{},"System identification",": measure the real robot's motor response and inertial parameters accurately and write them back into the simulator;",[50,1668,1669,1672,1673,1675],{},[21,1670,1671],{},"Actuator modeling",": model the torque-speed behavior of the ",[41,1674,239],{"href":238}," explicitly — often the deciding factor for legged-robot transfer.",[34,1677,1679],{"id":1678},"hardware-requirements","Hardware Requirements",[17,1681,1682,1683,1686,1687,1689,1690,1693],{},"Deployed policies command joints at hundreds of Hz, demanding high-bandwidth force control (see ",[41,1684,1685],{"href":165},"quasi-direct drive"," and the ",[41,1688,318],{"href":317},") and low-latency buses. The BXI ",[41,1691,1692],{"href":309},"Elf 3 humanoid"," ships with a MuJoCo environment and ROS2 SDK, and its >1000 Hz PCIE-CANFD control architecture supports the full simulation-to-hardware workflow.",{"title":169,"searchDepth":170,"depth":170,"links":1695},[1696,1697],{"id":1644,"depth":170,"text":1645},{"id":1678,"depth":170,"text":1679},"Sim-to-Real is the approach of training robot control policies at scale in physics simulation and then transferring them to real robots — the dominant training paradigm for humanoid locomotion today.","sim-to-real, simulation to reality, reinforcement learning robotics, domain randomization, MuJoCo, Isaac",{},"\u002Fglossary\u002Fen\u002Fsim-to-real",{"title":561,"description":1698},"glossary\u002Fen\u002Fsim-to-real","r1zVLzbVErZiDCC98F5liP8bIuQEPdAdUZGonwnDjak",{"id":1706,"title":1707,"alternateName":1708,"body":1709,"description":1835,"extension":175,"keywords":1836,"meta":1837,"navigation":178,"path":1838,"seo":1839,"stem":1840,"updated":182,"__hash__":1841},"glossary\u002Fglossary\u002Fen\u002Fsix-axis-force-torque-sensor.md","Six-Axis Force\u002FTorque Sensor","六维力传感器",{"type":9,"value":1710,"toc":1831},[1711,1715,1725,1729,1754,1758,1764,1825],[12,1712,1714],{"id":1713},"what-is-a-six-axis-forcetorque-sensor","What Is a Six-Axis Force\u002FTorque Sensor?",[17,1716,19,1717,1720,1721,1724],{},[21,1718,1719],{},"six-axis force\u002Ftorque sensor"," (6-axis F\u002FT sensor) measures ",[21,1722,1723],{},"three orthogonal forces (Fx, Fy, Fz) and three torques (Mx, My, Mz)"," at one point, typically using strain gauges or capacitive elements. Range, accuracy, and sample rate should be confirmed from the selected sensor's datasheet.",[34,1726,1728],{"id":1727},"where-it-goes-and-what-it-does","Where It Goes and What It Does",[47,1730,1731,1742,1748],{},[50,1732,1733,1736,1737,1741],{},[21,1734,1735],{},"Wrist",": between the arm flange and the tool, for precision assembly, polishing\u002Fdeburring, and hand-guided teaching — any ",[41,1738,1740],{"href":1739},"\u002Fen\u002Fglossary\u002Fforce-control","force-control"," task that needs true end-effector contact forces;",[50,1743,1744,1747],{},[21,1745,1746],{},"Ankle",": on humanoid feet, measuring ground reaction force and center of pressure (ZMP\u002FCoP), a key input for walking balance;",[50,1749,1750,1753],{},[21,1751,1752],{},"Test rigs",": torque calibration of joint actuators, impact testing.",[34,1755,1757],{"id":1756},"do-you-still-need-one-if-the-joint-does-current-based-force-control","Do You Still Need One If the Joint Does Current-Based Force Control?",[17,1759,1760,1763],{},[41,1761,1762],{"href":165},"Quasi-direct-drive (QDD)"," joints can estimate output torque from phase current after motor and drivetrain calibration. This can support joint-level compliance, but accuracy depends on friction, temperature, efficiency, and model error. A practical comparison is:",[82,1765,1766,1779],{},[85,1767,1768],{},[88,1769,1770,1773,1776],{},[91,1771,1772],{},"Scenario",[91,1774,1775],{},"Current-based estimation (QDD)",[91,1777,1778],{},"External 6-axis F\u002FT sensor",[104,1780,1781,1792,1803,1814],{},[88,1782,1783,1786,1789],{},[109,1784,1785],{},"Leg impact absorption, whole-body compliance",[109,1787,1788],{},"Useful control feedback; accuracy depends on calibration",[109,1790,1791],{},"Fit according to control and safety needs",[88,1793,1794,1797,1800],{},[109,1795,1796],{},"Precision assembly",[109,1798,1799],{},"Does not directly measure six-axis tool forces",[109,1801,1802],{},"Usually the better fit",[88,1804,1805,1808,1811],{},[109,1806,1807],{},"Foot ZMP \u002F ground reaction force",[109,1809,1810],{},"Indirect estimate only",[109,1812,1813],{},"Direct and reliable",[88,1815,1816,1819,1822],{},[109,1817,1818],{},"Hardware and integration",[109,1820,1821],{},"No dedicated joint sensor added",[109,1823,1824],{},"Adds sensing, calibration, and cabling",[17,1826,858,1827,1830],{},[21,1828,1829],{},"QDD current estimation supports joint-level compliance, while six-axis F\u002FT sensors directly measure tool or foot loads",". They can be combined according to accuracy and safety requirements.",{"title":169,"searchDepth":170,"depth":170,"links":1832},[1833,1834],{"id":1727,"depth":170,"text":1728},{"id":1756,"depth":170,"text":1757},"A six-axis force\u002Ftorque sensor measures three forces and three torques at one point for robot wrists, ankles, assembly, and contact control.","six-axis force torque sensor, 6-axis F\u002FT sensor, robot force sensor, wrist force sensor, force control",{},"\u002Fglossary\u002Fen\u002Fsix-axis-force-torque-sensor",{"title":1707,"description":1835},"glossary\u002Fen\u002Fsix-axis-force-torque-sensor","s99H8Bfrjpt7KTgLy5ajRVE4ecSwo3DeY2UJA_S0vkk",{"id":1843,"title":1844,"alternateName":1845,"body":1846,"description":1919,"extension":175,"keywords":1920,"meta":1921,"navigation":178,"path":1922,"seo":1923,"stem":1924,"updated":182,"__hash__":1925},"glossary\u002Fglossary\u002Fen\u002Fteleoperation.md","Teleoperation","遥操作",{"type":9,"value":1847,"toc":1915},[1848,1852,1857,1861,1882,1886,1906],[12,1849,1851],{"id":1850},"what-is-teleoperation","What Is Teleoperation?",[17,1853,1854,1856],{},[21,1855,1844],{}," is the technique of a human operator controlling a robot's motion remotely in real time, with the robot executing tasks on site. The operator side (leader) captures human motion intent; the robot side (follower) reproduces it and streams back visual and other feedback, closing a human-in-the-loop control loop.",[34,1858,1860],{"id":1859},"two-kinds-of-value","Two Kinds of Value",[1551,1862,1863,1869],{},[50,1864,1865,1868],{},[21,1866,1867],{},"Immediate utility",": for tasks where autonomy isn't ready, teleoperation makes robots useful now — hazardous-environment work, remote assembly, live demonstrations.",[50,1870,1871,1874,1875,1878,1879,1881],{},[21,1872,1873],{},"Data collection",": the observation–action trajectories produced by teleoperation are ",[21,1876,1877],{},"the highest-quality training data for imitation learning",". Today's mainstream ",[41,1880,1614],{"href":409}," training pipelines run on large-scale teleoperated demonstrations.",[34,1883,1885],{"id":1884},"common-forms","Common Forms",[47,1887,1888,1894,1900],{},[50,1889,1890,1893],{},[21,1891,1892],{},"Kinematically-matched leader arms",": a small leader arm mirroring the follower's structure — direct mapping, high precision;",[50,1895,1896,1899],{},[21,1897,1898],{},"VR \u002F motion capture",": headsets and controllers, or full-body mocap, driving whole-body humanoid motion;",[50,1901,1902,1905],{},[21,1903,1904],{},"Shared autonomy",": the human gives high-level commands while the robot handles balance and trajectories.",[17,1907,243,1908,1911,1912,1914],{},[41,1909,1910],{"href":1578},"UpperBody 1 dual-arm platform"," ships with a plug-and-play operator console, and the ",[41,1913,1692],{"href":309}," supports both teleoperation and autonomous modes for manipulation data collection.",{"title":169,"searchDepth":170,"depth":170,"links":1916},[1917,1918],{"id":1859,"depth":170,"text":1860},{"id":1884,"depth":170,"text":1885},"Teleoperation is real-time remote control of a robot by a human operator — both a practical deployment mode for humanoid robots today and the core method for collecting real-robot demonstration data to train embodied-AI models.","teleoperation, robot teleoperation, demonstration data collection, leader-follower control, imitation learning data",{},"\u002Fglossary\u002Fen\u002Fteleoperation",{"title":1844,"description":1919},"glossary\u002Fen\u002Fteleoperation","a0mBNQK0YBS6Xa9_ZlYjuzw1VfgVmXYcR5HIf1XT21A",{"id":1927,"title":1928,"alternateName":1929,"body":1930,"description":2026,"extension":175,"keywords":2027,"meta":2028,"navigation":178,"path":2029,"seo":2030,"stem":2031,"updated":182,"__hash__":2032},"glossary\u002Fglossary\u002Fen\u002Ftorque-density.md","Torque Density (N·m\u002Fkg)","扭矩密度",{"type":9,"value":1931,"toc":2022},[1932,1936,1950,1954,1957,1974,2005,2009],[12,1933,1935],{"id":1934},"what-is-torque-density","What Is Torque Density?",[17,1937,1938,1941,1942,1945,1946,1949],{},[21,1939,1940],{},"Torque density"," is the torque an actuator can deliver per kilogram of its own mass, expressed in ",[21,1943,1944],{},"N·m\u002Fkg",", quoted either at peak or at rated torque. It is an important ",[41,1947,1948],{"href":238},"robot joint motor"," metric because actuators mounted in a limb become payload for every joint upstream.",[34,1951,1953],{"id":1952},"how-to-compute-and-compare-it","How to Compute and Compare It",[17,1955,1956],{},"Torque density = output torque ÷ total actuator mass (gearbox and encoders included). Always align the basis before comparing:",[47,1958,1959,1965,1971],{},[50,1960,1961,1964],{},[21,1962,1963],{},"Peak vs. rated",": marketing figures are usually peak-basis, often 3× or more above the rated-basis number;",[50,1966,1967,1970],{},[21,1968,1969],{},"Driver included or not",": integrated actuator modules include the drive electronics — bare-motor figures are not comparable;",[50,1972,1973],{},"Current integrated joint actuators typically reach 30–70 N·m\u002Fkg peak; top products exceed 100 N·m\u002Fkg.",[17,1975,1976,1977,1980,1981,1985,1986,1989,1990,1994,1995,1999,2000,2004],{},"For the BXI ",[41,1978,1979],{"href":160},"85\u002F70\u002F50 series"," (peak basis): the ",[41,1982,1984],{"href":1983},"\u002Fen\u002Fmotors\u002Fbxi8515-19","BXI8515-19"," delivers 150 N·m ÷ 1.4 kg ≈ ",[21,1987,1988],{},"107 N·m\u002Fkg",", the ",[41,1991,1993],{"href":1992},"\u002Fen\u002Fmotors\u002Fbxi7010-19","BXI7010-19"," 50 N·m ÷ 0.8 kg ≈ 63 N·m\u002Fkg, the ",[41,1996,1998],{"href":1997},"\u002Fen\u002Fmotors\u002Fbxi5018-19","BXI5018-19"," ≈ 64 N·m\u002Fkg, and the ",[41,2001,2003],{"href":2002},"\u002Fen\u002Fmotors\u002Fbxi5014-19","BXI5014-19"," 50 N·m\u002Fkg.",[34,2006,2008],{"id":2007},"where-high-torque-density-comes-from","Where High Torque Density Comes From",[17,2010,2011,2012,2014,2015,2017,2018,2021],{},"A large-diameter, high-pole-count outrunner torque motor provides the base torque; a ",[41,2013,706],{"href":652}," multiplies it at low added mass. The ",[41,2016,645],{"href":165}," approach uses a low-to-medium reduction ratio to balance torque density against backdrivability, while ",[41,2019,2020],{"href":1311},"harmonic drives"," can quote higher density figures at the cost of impact tolerance and backdrivability.",{"title":169,"searchDepth":170,"depth":170,"links":2023},[2024,2025],{"id":1952,"depth":170,"text":1953},{"id":2007,"depth":170,"text":2008},"Torque density measures actuator torque per unit mass. It helps compare joint weight, payload capacity, and dynamic performance.","torque density, Nm\u002Fkg, robot actuator metric, torque-to-weight ratio, joint motor selection",{},"\u002Fglossary\u002Fen\u002Ftorque-density",{"title":1928,"description":2026},"glossary\u002Fen\u002Ftorque-density","zX6KLQ8QPigYB3Sofy6MHXyd_zOCs9z5y3uva-gWOeY",{"id":2034,"title":2035,"alternateName":2036,"body":2037,"description":2111,"extension":175,"keywords":2112,"meta":2113,"navigation":178,"path":2114,"seo":2115,"stem":2116,"updated":182,"__hash__":2117},"glossary\u002Fglossary\u002Fen\u002Fvla-model.md","VLA Model","VLA模型 \u002F Vision-Language-Action Model",{"type":9,"value":2038,"toc":2106},[2039,2043,2051,2055,2058,2062,2071,2075,2095],[12,2040,2042],{"id":2041},"what-is-a-vla-model","What Is a VLA Model?",[17,2044,19,2045,2048,2049,229],{},[21,2046,2047],{},"VLA model"," (Vision-Language-Action model) is an end-to-end robot foundation model: it takes camera images and a natural-language instruction (e.g. \"put the cup in the drawer\") as input and directly outputs robot actions — joint angles, end-effector poses, or gripper commands. By compressing perception, language understanding, and motion generation into one network, VLA has become one of the most watched policy architectures in ",[41,2050,410],{"href":409},[34,2052,2054],{"id":2053},"relationship-to-llms","Relationship to LLMs",[17,2056,2057],{},"VLA models are typically built on top of vision-language models (VLMs): they inherit the semantic and commonsense knowledge of large language models, then replace or extend the output head with action tokens or continuous actions. A useful mental model: an LLM predicts the next word; a VLA predicts the next action — it is the language-model family extended into the physical world.",[34,2059,2061],{"id":2060},"why-real-robot-data-is-essential","Why Real-Robot Data Is Essential",[17,2063,2064,2065,2067,2068,2070],{},"The internet holds vast text and images, but almost no paired \"image + instruction → joint action\" data. VLA training therefore depends heavily on real-robot demonstrations, mostly collected via ",[41,2066,415],{"href":414},". Simulation data can add scale, but must cross the ",[41,2069,561],{"href":560}," gap. Data diversity — across tasks, scenes, and embodiments — often matters more for generalization than parameter count.",[34,2072,2074],{"id":2073},"representative-work","Representative Work",[47,2076,2077,2083,2089],{},[50,2078,2079,2082],{},[21,2080,2081],{},"RT-2"," (Google DeepMind): co-trains a VLM with robot action tokens, demonstrating that web-scale knowledge transfers to manipulation;",[50,2084,2085,2088],{},[21,2086,2087],{},"OpenVLA",": an open-source VLA trained on large open robot datasets, designed for community fine-tuning;",[50,2090,2091,2094],{},[21,2092,2093],{},"π0"," (Physical Intelligence): uses flow matching to generate continuous actions, targeting a general cross-embodiment manipulation policy.",[17,2096,2097,2098,2101,2102,2105],{},"Deploying and fine-tuning VLA on real hardware requires a platform that executes high-rate action commands reliably — BXI's ",[41,2099,2100],{"href":309},"humanoid robot"," and ",[41,2103,2104],{"href":1578},"dual-arm platform"," both expose ROS2 interfaces suited to VLA research.",{"title":169,"searchDepth":170,"depth":170,"links":2107},[2108,2109,2110],{"id":2053,"depth":170,"text":2054},{"id":2060,"depth":170,"text":2061},{"id":2073,"depth":170,"text":2074},"A Vision-Language-Action model maps camera observations and language instructions to robot actions for manipulation and embodied-AI tasks.","VLA model, vision language action model, robot foundation model, end-to-end robot policy, embodied AI model",{},"\u002Fglossary\u002Fen\u002Fvla-model",{"title":2035,"description":2111},"glossary\u002Fen\u002Fvla-model","NFikZ8HZmgXboc8Yz0BrE7ZWSYyddapOZmEH4UV-Tt0",{"id":2119,"title":2120,"alternateName":2121,"body":2122,"description":2181,"extension":175,"keywords":2182,"meta":2183,"navigation":178,"path":2184,"seo":2185,"stem":2186,"updated":182,"__hash__":2187},"glossary\u002Fglossary\u002Fen\u002Fworld-model.md","World Model","世界模型",{"type":9,"value":2123,"toc":2177},[2124,2128,2135,2139,2158,2162,2171],[12,2125,2127],{"id":2126},"what-is-a-world-model","What Is a World Model?",[17,2129,2130,2131,2134],{},"In robotics, a ",[21,2132,2133],{},"world model"," is an internal predictive model of environment dynamics learned by an agent: given the current observation (images, state) and an intended action, it predicts what the environment will look like next. With a world model, a robot can roll out the consequences of actions \"in its head\" — imagine first, act second — instead of trial-and-erroring every step in the real world.",[34,2136,2138],{"id":2137},"relationship-to-vla-and-reinforcement-learning","Relationship to VLA and Reinforcement Learning",[47,2140,2141,2149],{},[50,2142,2143,2148],{},[21,2144,2145,2146],{},"Versus ",[41,2147,1541],{"href":1540},": a VLA maps observations and instructions directly to actions — a reactive policy. A world model explicitly captures \"action → consequence\" and supports planning and rollouts. The two are complementary: a world model gives a VLA the ability to imagine and verify.",[50,2150,2151,2154,2155,2157],{},[21,2152,2153],{},"Versus reinforcement learning",": in model-based RL, the policy trains largely by \"dreaming\" inside the learned world model, requiring far fewer real interactions and cutting the cost of real-robot sampling. This parallels training in a simulator followed by ",[41,2156,561],{"href":560}," transfer — a world model is essentially a differentiable \"neural simulator\" learned from data.",[34,2159,2161],{"id":2160},"why-it-is-a-2026-embodied-ai-focus","Why It Is a 2026 Embodied-AI Focus",[17,2163,2164,2165,2167,2168,2170],{},"First, progress in video generation shows large models can absorb substantial physical regularities from massive video corpora — a foundation for general world models. Second, real-robot data is scarce and expensive, so the field wants world models to amplify limited ",[41,2166,415],{"href":414}," data by generating and evaluating additional virtual experience. Third, ",[41,2169,410],{"href":409}," has high safety stakes: rehearsing inside an internal model before executing on hardware is a natural way to reduce risk.",[17,2172,2173,2174,2176],{},"Whatever the technical route, world models must ultimately be validated on real hardware. BXI's ",[41,2175,2100],{"href":309}," ships with MuJoCo simulation and a ROS2 SDK, supporting the full loop of in-model training followed by real-robot validation.",{"title":169,"searchDepth":170,"depth":170,"links":2178},[2179,2180],{"id":2137,"depth":170,"text":2138},{"id":2160,"depth":170,"text":2161},"A robotics world model predicts future environment states from the current state and an action, supporting planning, simulation, and policy training.","world model, embodied AI, predictive model, model-based reinforcement learning, robot learning",{},"\u002Fglossary\u002Fen\u002Fworld-model",{"title":2120,"description":2181},"glossary\u002Fen\u002Fworld-model","M9XOHg398s6hJe7UWdj2CPKScagecDXq33iVYNAg-_U",1789463605035]