Short answer: a robot joint module — often sold as an integrated actuator — is a single unit that combines a motor, a reduction gearbox, an encoder and usually a driver into the joint of a robot. Instead of sourcing and integrating four components per joint, the robot maker buys one tested assembly. Humanoid robots made the category strategic: every humanoid design needs joints by the dozen — the count varies widely by design — so module cost, weight and consistency now decide whether a humanoid can be manufactured at all, not just how well it moves.
What is inside a joint module?
- The motor. Increasingly a frameless torque motor — stator and rotor supplied without a housing, built directly into the joint structure. Our article on frameless torque motors and integrated joints covers that layer in depth.
- The gearbox. Usually a harmonic reducer in smaller joints, a planetary or cycloidal stage where torque and shock loads are higher. The gearbox sets backlash and torsional stiffness — the joint's feel.
- The encoder. Position feedback, often on both the motor side and the output side, so the controller knows where the joint actually is, not just where the motor turned.
- The driver and electronics. The power stage and communications (typically a real-time bus) that let one controller coordinate every joint on the machine.
Rotary and linear modules are different products
Most arm and leg joints rotate, and rotary modules dominate catalogues. But several humanoid designs use linear actuators — a motor driving a screw — for knees, ankles or hips, where a push-pull motion gives high force in a slim package. The two types have different suppliers, different failure modes and different test regimes; a sourcing plan that treats "actuator" as one commodity will miss both. When you read a humanoid announcement, the actuator strategy is the tell: it reveals whether the company is building a research prototype or engineering for manufacture — a theme we return to in our 2026 survey of humanoid robots in warehouses and factories.
Buy the module, or build the joint?
Integration is the trade. Buying modules shortens development, transfers test burden to the supplier, and gives consistent units from day one — at a higher unit price and with your product's performance bounded by a catalogue. Self-integrating components gives design freedom and, at volume, lower cost, but the maker now owns gear-mesh quality, thermal behaviour, encoder calibration and every failure mode in between. Some makers choose the second path as identity: Unitree, for example, is publicly known for developing its own joint units as part of its cost strategy. Precision-drive specialists such as maxon sit on the supply side of the same decision. Neither answer is wrong; the mistake is drifting into self-integration by accident because a catalogue module was 10% too expensive at prototype stage.
What buyers should check
Ask for backlash and torsional stiffness figures measured the same way across vendors; continuous torque at a stated temperature, not only peak torque; encoder arrangement and what happens on power loss; the communications protocol and whether the driver firmware is field-updatable; and rated life under a stated duty cycle, with the test method named. Joint modules live in the directory's core components category — 160 companies as of 10 November 2026 — which is the fastest way to see how deep the specialist supply base now runs.
One friction to plan around: standards are still thin
Unlike servo drives in factory automation, joint modules have not converged on common interfaces. Mounting patterns, shaft connections, bus protocols, voltage classes and connector choices differ from maker to maker, which means a module is rarely a drop-in replacement for another brand's. For buyers, the practical consequences are threefold: qualify a second source early if the joint is strategic; negotiate documentation and firmware-update terms at purchase, while you still have leverage; and treat the supplier's own test data format as part of the product, because your acceptance testing will be built on it. As humanoid volumes grow, interface pressure will increase — large makers standardise supply chains by force — but a 2026 sourcing plan should assume today's fragmentation persists for the life of the current design.
Frequently asked questions
What is a robot joint module?
It is an integrated actuator: a single assembly that combines an electric motor, a reduction gearbox, position encoders and usually the driver electronics for one robot joint. Buying the module means the robot maker receives a tested joint with known backlash, stiffness and torque behaviour, instead of integrating four separate components and owning every interface between them.
How many joint modules does a humanoid robot need?
There is no standard number — it depends entirely on the design. A humanoid needs powered joints at shoulders, elbows, wrists, hips, knees and ankles, plus more if the hands, waist and neck are actuated. Designs range from a dozen-plus powered axes to several dozen. That spread is why reporting a single “joints per humanoid” figure is meaningless without naming the machine.
What is the difference between a joint module and a servo motor?
A servo motor is one component: it produces rotation. A joint module is the working joint built around a motor — gearing to multiply torque and slow the output, encoders to report position, bearings and housing to carry load, and usually the driver electronics. A servo can be part of a joint module, but a joint module is the larger, integration-complete unit.
Who supplies robot joint modules?
Two groups. Component specialists sell modules and their ingredients into the core components supply chain — the GlobleMatch directory lists 160 companies in that category as of 10 November 2026. And some robot makers, Unitree among them, develop their own joint units in-house as a deliberate cost and design strategy rather than buying them.