Robotics

Tendon Drive

1982ActivePublished: 26 June 2026Updated: 26 June 2026Published

Key innovation

Relocating motors away from distal links (fingers, wrists) into the proximal body and transmitting force through tendons — Dyneema cables, steel wires or pneumatic muscles — which radically reduces the mass and inertia of moving segments of a manipulator.

How it works

Step 1: a motor (DC, BLDC, McKibben pneumatic muscle or a servo) is mounted in the robot's body or forearm. Step 2: a drum/pulley is attached to the motor shaft, with a cable wound onto it. Step 3: the cable runs through guides (Bowden tubes or a pulley system) along the manipulator's structure to the distal joint. Step 4: at the joint, the cable terminates on one side of the rotation axis of a phalanx — pulling the cable generates torque. Step 5: because a cable can only pull, every joint needs either an antagonistic pair (two cables on opposite sides of the axis), or one cable with a return spring, or N+1 routing (N joints driven by N+1 cables with shared paths). Step 6: encoders on the motor shafts plus force sensors (Hall, strain gauges, BioTac) in the fingers close position and force control loops, with compensation calibrated for cable stretching and routing friction.

Problem solved

Direct drive (a motor in every joint) in a humanoid hand faces three design barriers: (1) a motor with meaningful torque does not fit inside a phalanx — the finger becomes thicker than a human one, (2) the distally-mounted motor mass increases moment of inertia, damping speed and raising energy cost, (3) motors inside a phalanx are exposed to impact and vibration, shortening lifespan. Tendon Drive solves all three — motors sit in the safe, large body/forearm, fingers are light and slender, and the cable transmission itself is passive and tolerant of momentary overloads (the cable slips or stretches instead of snapping a gearbox).

Components

Proximal ActuatorProduces drive force without loading distal moving segments.

A motor physically remote from the moved joint — in the body, arm or forearm. Most often a BLDC with planetary gearbox (electric Shadow Hand variant), DC with reducer (research designs) or a McKibben pneumatic muscle (Shadow Air Muscle Hand, Festo BionicSoftHand).

BLDC + planetary gearboxStandard for new humanoids (Tesla, Sanctuary).

DC + cycloidal/harmonicResearch configurations.

McKibben pneumatic muscleVery high power-to-weight ratio and compliance, but requires compressed air.

Official

Tendon (Cable)Transfers force from actuator to joint; its stretch determines control bandwidth and precision.

Passive transmission element: steel cable, UHMWPE polymer (Dyneema, Spectra), aramid (Kevlar) or carbon composite. Material choice balances breaking strength, axial stiffness, cyclic fatigue resistance and routing friction.

Official

Routing SystemDefines the force-transfer kinematics — often the largest source of inaccuracy in the system.

A set of pulleys, Bowden tubes, guides and anchor points through which the cable travels from actuator to joint. Routing determines inter-joint coupling (coupled vs decoupled), efficiency (cable-routing friction) and packaging.

Bowden cableCable inside a smooth tube; allows flexible routing but introduces friction.

Pulley networkPulley network with minimal friction at the cost of more complex geometry.

Official

Antagonistic Pair / ReturnEnables bidirectional joint control with a unidirectional actuator.

Because a cable can only pull, every joint needs a mechanism to extend in both directions. The antagonistic variant (two cables on opposite sides of the axis, like biceps-triceps) offers active control of stiffness and force in both directions. The return-spring variant is simpler and lighter but gives up passive stiffness in one direction.

Sensing LayerCloses the position and force control loop despite imperfections of the passive transmission.

Motor-shaft encoders (shaft angle, not joint angle — a notorious pitfall), Hall-effect or potentiometers in joints (actual joint position), strain gauges/load cells on cables (tension), optionally tactile sensors (BioTac, Syntouch). The sensing layer compensates for cable stretch and friction.

Implementation

Reference implementations

Shadow Dexterous Hand

— · Shadow Robot Company (London, UK)

Official

OpenAI Learning Dexterity (sim-to-real RL on Shadow Hand)

Python (MuJoCo + PyTorch) · OpenAI

Official

DLR Hand II

— · DLR Robotics and Mechatronics Center

Official

Implementation pitfalls

Cable stretch under loadHigh

Dyneema stretches by 1-3% under nominal tension; the motor encoder reports shaft position, not joint position. Without compensation, finger position error grows with force.

Fix:A second encoder at the joint plus a stretch model (Hooke + nonlinearity) calibrated per unit. The cable can also be pre-loaded above the working range.

Routing friction and hysteresisHigh

In long/curved Bowden paths, cable-guide friction creates hysteresis: the same joint angle requires different force depending on direction of motion. Impossible to compensate with open-loop control alone.

Fix:End-of-cable force sensors + impedance control; pulley networks instead of Bowden; PTFE/Teflon liners; dry molybdenum lubricant.

Cable wear and breakageMedium

UHMWPE cables withstand static breakage well but are vulnerable to cyclic fatigue (especially on sharp pulley bends). Mean time between replacements in the Shadow Hand is ~10⁶-10⁷ cycles.

Fix:Quick-change cable design; real-time tension monitoring; cable redundancy for critical joints.

Unintended inter-joint couplingMedium

N+1 schemes and shared routing save actuators but make motion of one joint change the cable length serving another. Without a model this causes drift.

Fix:Explicit transmission Jacobian in the controller; prefer decoupled routing for critical paths; per-unit parameter identification.

Evolution

Original paper · 1983 · Journal of Mechanisms, Transmissions, and Automation in Design (ASME), Vol. 105. · J. Kenneth Salisbury

Kinematic and Force Analysis of Articulated Mechanical Hands

J. Kenneth Salisbury, John J. Craig

1962

Belgrade Hand — first advanced tendon-driven prosthesis

Inflection point

Rajko Tomović and Miodrag Rakić (University of Belgrade) demonstrate a 5-finger prosthesis with mechanical tendon control — the progenitor of modern prosthetic hands.

1982

Stanford/JPL Hand — analytical formalisation

Inflection point

Kenneth Salisbury (Stanford) and JPL design a 3-finger tendon-driven hand with a 3-D kinematic and force model. The analytical foundations used to this day are established.

Articulated Hands: Force Control and Kinematic Issues (paper)

1997

Shadow Robot Company founded in London

Foundation of the company that will turn tendon drive from research prototype into a commercial anthropomorphic-hand product — the Shadow Dexterous Hand becomes the industry standard for manipulation research.

2018

OpenAI Learning Dexterity — RL on Shadow Hand

Inflection point

OpenAI demonstrates reinforcement learning solving the Rubik's cube on a Shadow Dexterous Hand. For the first time, AI rivals humans in controlling a tendon-driven manipulator.

Learning Dexterous In-Hand Manipulation (paper)

2023

Tesla Optimus Gen 2 — tendon drive in a mass-market humanoid

Inflection point

Tesla shows the Optimus humanoid hand with 11 degrees of freedom per hand and tendon drive. Tendon drive enters the consumer humanoid segment.

2025

Sanctuary AI Phoenix v8 and Shadow Hand 3.0 — VLA-ready generation

Vendors release 20+ DOF hands with dense tactile sensing and interfaces designed for VLA policy learning. Tendon drive becomes the de facto standard for humanoid hands trained from teleoperation.

Hyperparameters (configurable axes)

Tendon materialCritical

Drives stiffness, breaking strength, cyclic life and friction.

Stalowa linkaHigh stiffness, low stretch but prone to fatigue and kinking.

UHMWPE (Dyneema/Spectra)High breaking strength, light, but relatively elastic — requires a compensation model.

Aramid (Kevlar)Compromise between steel and UHMWPE.

Actuation schemeCritical

Determines actuator count per DOF and whether the joint has active stiffness control.

Para antagonistyczna2N actuators; full stiffness control.

Pull-springN actuators + return springs; simpler, lighter.

N+1 routingN+1 actuators for N DOFs with shared paths; demands careful coupling-decoupling.

Actuator typeHigh

Determines bandwidth, torque, compliance and power-supply scheme.

BLDC + przekładnia planetarnaDefault choice for new humanoids.

Mięsień pneumatyczny McKibbenVery high power-to-weight but requires compressed air.

Routing topologyHigh

Drives inter-joint coupling and friction.

BowdenFlexible routing, higher friction.

Pulley networkLow friction, stiffer geometry.

Computational complexity

Time complexity: Pasmo sterowania ograniczone do ~50-200 Hz (kabel UHMWPE) lub ~5-20 Hz (mięsień pneumatyczny). Space complexity: Liczba aktuatorów ≥ liczba stopni swobody (typowo 1.0-1.5×, dla schematu N+1 wynosi N+1).

Compute bottleneck

Transmission model identification and compensation

The main bottleneck is not computation but identification of the cable-stretch model, friction hysteresis and inter-joint coupling for a given configuration. This requires per-unit calibration, and the model degrades as the cable wears.

Depends on

Rozciągliwość kablaZłożoność routingu

Execution paradigm

Primary mode

Dense

All tendons are active in parallel — classic dense control. The conditional aspect only appears when the policy decides which fingers to engage.

Activation pattern

All paths active

Additional modes

Conditional

Routing mechanism

'Routing' in this concept is the literal physical guidance of the cable from motor to joint — unlike in MoE, where routing is a learned function. Here it is hard-wired into the hardware geometry.